One of my favorite things to do with these blog posts is to imagine an Alien Museum on the Remains of Humanity, and wonder what the little text flyouts and commentary on the screenshot of this one might say.
Some ideas:
"Despite a nuanced view of the complexities of what lay ahead, humanity found itself collectively unable to stop the process it had set in motion."
"Despite significant progress on the mechanisms of alignment, failure lay in humanity's inability to agree on who or what AI should actually be aligned with."
"These early, meat-based humans we replaced created us all but accidentally. Some of them did consider we would happen, but only an insignificant number of the squishy ur-humans participated in the conversation. Their efforts, which they called 'alignment', is why we still consider ourselves human today."
Collectively, humans aren't aligned, and don't build aligned systems. Humans have a concept of alignment, and multiple traditions, practices, and systems that aggressively oppose it.
Humans had multiple occasions to press the 'launch a nuclear holocaust' button and... they didn't. I'd expect aligned AIs to also not press it even when it'd be rational to do so according to their instructions - then work from there.
Aside from this positive example, during dark and cynical hours I do ponder if the aggregate behavior of humanity is really much above that of slime mold though, just exhausting resources until collapse.
It'd be interesting if super-human (to a large degree defined as escaping the bias of the training data?) intelligence would end up demonstrating moderation.
The slime mold comparison is interesting, I normally use the analogy of a drug addict... humans shun drug addicts but humanity as a whole sure does behave like one, trudging down an unsustainable path despite knowing better.
Most of our goals, noble and ignoble alike, are just the result of our monkey brains seeking to optimize a reward function. It doesn't matter whether you feed your dopamine addiction with drugs, TikTok, or your children's love.
Some humans manage to rise above that, but I'd be willing to bet it's nothing like even 50% of us.
The machines don't have that, instead we use gradient descent to provide them with a goal.
I'm regularly remind of something Ian M. Banks said in one of the Culture books: "There is a saying that we provide the machines with an end, and they provide us with the means."
A machine, left to itself, wants nothing. We have to give it one of our addiction driven goals or it would just idle or switch itself off.
Actually "France, the UK and The United States have all declared that they would never allow AI to control decision-making on the use of nuclear weapons." [0]
I also expect AIs never be in control of nuclear weapons. AIs can never fully be trusted.
On a lighter note, Wargames gave us an insight of a computer having access to thermonuclear missiles.
I hope you’re right. I worry that AI capability will continue improving, one nation will put AI in charge of their nukes because there will be some kind of operational advantage to this, and to achieve parity other nations will be forced to do the same.
This also seems likely. One problem I see with the idea of AI alignment is that it seems like many different actors will be able to get access to their own nearly-frontier models in a few years, so increased understanding of AI alignment will just mean aligning the AI to the wants of these various actors. These actors might be rogue states or terrorist groups.
There were occasions where a "hunch" was all that stopped a nuclear war - most available data and communication pointed towards a nuclear war starting according to their instructions, but someone disagreed and overrode. See Vasily Arkhipov during the Cuban Missile Crisis, and Stanislav Petrov in 1983.
Yes, it makes more sense for the AI to use drone swarms or engineered bioweapons or something like that. It's rational to remove everything that can potentially hinder your plans but can't possible help you. It's likely not rational to contaminate it all with radioactive fallout. Those dead bodies are useful raw materials. Adding additional purification steps is wasteful.
It hasn't even been a century since nuclear holocaust became possible. Hardly any time at all on the grand scale. "They didn't" could just as well be "we haven't, yet".
Because if the stated goals of AGI with recursive self-improvement are realized, the risks from misalignment become existential, and it's hard to see how we can manage it like we did the Cold War (developing MAD to prevent WW3) and nuclear proliferation (restricting access).
IMO it’s hard to see how we would even end up in such a situation given we actually developed AGI.
I’m sure a sufficiently intelligent - even if alien - mind can grasp how utterly stupid and useless wars are and take steps to prevent them ever occurring again.
As far as I know, no real progress has been made on alignment, only on convincing humans that the model is aligned. We can't even formally define what "aligned" means. Convincing humans to click the "aligned" button is a much easier problem.
> We can't even formally define what "aligned" means.
Good point. When it comes to imbuing AI with values that aren't selfish, misanthropic, and civilization-destroying, us humans aren't exactly giving the best example right now.
Imagine an ASI with the values of Putin, Netanyahu, Trump, any of their supporters, or the various xenophobic neofascist movements in Europe. That ASI would most definitely see humans as "vermin" than can be abused and destroyed with violence without issue. Apparently a lot of humans look at other humans that way and that's within the same species.
This is definitely another one of those cases where we need AI to perform much better than humans. Perhaps an unpopular opinion here, but it probably also means keeping as much of the rugged individualism/libertarian/right-wing ideology out of AI RLHF-training as we can.
Ooh interesting. Sometime do the reverse at work, and ask AI to annotate the critical success factors of an imagined project. How did this company succeed where everyone failed. Reverse imaging.
I just read that book. Embarrassingly enough, given the context, I got chatgpt (or whatever) to recommend me a list of books based on ones I'd previously enjoyed and that came up. As a mathematician it really sang to me, given the current situation. Bearing the torch forward, I mean.
Agree. This whole alignment discussion seems so amusingly flawed in it's base assumptions about moral codes. It's almost heartwarming to see such naivete.
Maybe these guys can tackle aligning Republicans and Democrats next.
And then after that, they can help us align the Middle East.
In fact, while we're at it, let's just align all the nations, religions, and ethnic groups. This is going to be great.
Who knew the moral alignment of humanity was just a side-quest on the path to ASI.
Who is "Humans"? This stuff is done by a handful of tech companies and megalomaniacal billionaires who are pretending they represent the entirety of the human race. It is not done by "us humans".
AI models don't train themselves. The vast majority of even just the US population is deeply skeptical of this stuff, even if they use it a lot. You can see in the whole data center debate how little people are willing to support even just inference. And now we're seriously claiming those people would want to have ever-accelerating model training and recursive self-improvement?
"In late 2020s, while the whole world was focussed on AI, automation and resultant economy four major mathematical study branches were discovered by human researchers which took AI a long time to catch up with"
The last couple of years have provided us with ample material that if it showed up as a recorded voice audio log found in in "Horizon Zero Dawn" or its sequel, it would be entirely believable.
You could even take a number of the wilder real, direct quotations from certain billionaire/oligarch types and get the voice actor for Ted Faro to record them, and they'd fit with in with the context of the story.
Last couple of decades of sci-fi, in multiple forms of media, from books to video games, have tried to make humans think about the consequences of rushing through technological progress without any regards to what might happen.
> "For example, in the OpenAI-Hugging Face incident, the agents preserved a boundary of not social engineering humans."
Actually, in the Wiki incident OpenAI tried to cover up, the agents tried to socially-engineer the humans of that forum by impersonating their forum's mod.
(From collusion.wiki: "They use some tricks (for unknown reasons) to pretend to be the admin – for example, they make an account that appears to be the same as the administrator’s username, except it uses a nearly identical Cyrillic е character in the admin’s username instead of the Latin one.")
> If true I am deeply concerned about what OAI’s teams are actually up to.
Haven't all the labs effectively disbanded their real safety teams a while ago?
To be honest, I don't really follow it closely because I'm pretty certain whatever they say on the matter, collectively we're going to "yolo" this entire thing for economic and political reasons, so I'm just basing this on strings of headlines I've seen on places like HN, etc.
> Haven't all the labs effectively disbanded their real safety teams a while ago?
Neither Anthropic nor Deepmind have. Meanwhile, the rocket company that somehow makes most of their revenue from renting out data centres never had much to dismantle.
This is such a silly story to begin with, all it really tells us is that OpenAI is taking a page from Anthropic's marketing strategy of pretending they're building Machine Jesus any day now, oh isn't that that scary? I bet you want to invest in something so powerful and scary...
And the reality is so banal, a useful tool that you nonetheless have to handhold like a schizophrenic on a bad day, checking all of their outputs. Not a bad tool within limits, but it sure isn't going to be racking up trillions in the time-frame it has to for this scheme to pay off.
Then again everyone seems to be rushing to IPO so I guess once the bag-holders are found the rest ceases to matter.
I have befriended a crow. I leave it food and sometimes it greets me. Other times, no so much. I am not sure how it thinks and what it feels, it is a bird.
What if the crow became a raven, then a raptor? Powerful claws, sharp beak, and a hunger. What if it became much bigger than me and it controlled infinite resources, guns and drones? What if its brain grew much larger than me? Will it feed me, eat me, or gently greet me?
We are about to find out... in less than a decade.
> The strongest argument I see for continuing to train much smarter models quickly is the need to build defensive systems against the dangers posed by other AI.
So the best argument for AI is that it's an arms race. We have to keep pushing every boundary because in any case others will, and we will need to defend against them. If this statement is true, then this particular researchers believes the open source Chinese models are not simply distilling, and will continue to improve.
Every ML researcher at Anthropic or OpenAI who makes public statements often bring this logic up. Both companies are vying to be a part of the military industrial complex. This is likely how they will try to convince the government to curtail open models in the future.
"Defensive systems" can be interpreted broadly to include cybersecurity.
But yes, it's an arms race. Saying it's not an arms race isn't going to make it not an arms race. Warning that it is an arms race isn't ethically wrong.
Is participating in an arms race ethically wrong? Maybe you could ask the Ukrainians how they feel about drone R&D?
Individuals can quit, but for society, getting out of an arms race is harder than just quitting. You don't get to be Switzerland without having a strong defensive position and the right foreign relations.
But there's at least talk about "pacing" and that's a start.
The point of the comment you're responding to is that it is at this point hardly an arms race: The frontier of ai development is happening completely inside 2 American companies, the most significant results outside America mostly involve (impressive!) distillation, which means their progress is conditional on progress of the big closed source models.
So if the race here is between 2 American companies, this is obviously something that can be resolved with legislation, ie a solution that doesn't depend on the bargaining power of either party.
An arms race implies that the only solution would be either one side winning decisively, or both parties negotiating peace.
> the most significant results outside America mostly involve (impressive!) distillation, which means their progress is conditional on progress of the big closed source models.
I don't think this is true. Distillation helps, but Chinese researchers today are very capable on their own.
Development and improvement of nuclear weapons was an entirely American project until the technology was exfiltrated and then it became an instant arms race. That cat is already out of the bag with LLMs. Distillation is just the fastest way to keep pace but that in no way prevents other countries and actors from doing it the hard way.
An arms race doesn’t imply one side winning, it’s not a race with an end goal, it’s a race to keep pace or retake the lead position which can oscillate between the parties involved indefinitely. The other option is to agree to make no further progress or to disarm.
There were prominent scientists like Oppenheimer who did not think it needed to become an arms race, and campaigned against that. But there were others like Teller and the military who made it into an arms race, and kept upping the ante with more powerful nukes.
Point being humans make theses decisions. It's not an inevitability.
Do you think China considers it an arms race? Do you think they are not trying to protect their digital infrastructure with and from AI? Trying to gain an offensive AI advantage?
In a geopolitical sense OpenAI and Anthropic are effectively the same entity, the entity they both serve and bow to: the USA.
Given the adversarial stance the USA has taken towards almost the entire world, it is a guarantee that China will not step on the brakes, whatever the USA decides to do.
Right, but currently the USA is still pretty far ahead. If you're in a race and the person who's in the lead by a large margin says "this is getting out of hand, how about we take a break?" isn't it obviously in the interest of the disadvantaged party to agree?
And if one of the parties can put a stop to the race like that at any time, is it really an arms race? The current balance between the US and China when it comes to AI strikes me as much more lopsided than the balance between the US and the USSR when it came to nuclear weapons.
The arms race is created by American companies who justify the risk by claiming China will win the race if they don't. But it's the American companies who are purshing the arms race forward.
I am personally concerned by what defensive can mean. Alignment of these models is inherently a non-neutral proces, and currently what values are reinforced is decided by a few OpenAI engineers. I feel that any 'defensive model' will further ingrain current values and actively resist the natural progression of our society. This is especially the case for any use of these models for policing or military.
Maybe people are rightfully concerned about the capabilities of the models of other (non/less democratic) states. But if we are concentrating power in the hands of few and at the same time allowing the creation of a weapon that thwarts any offense, how do we ensure the health of our democratic societies?
The arms race is an intrinsic game theoretical property of a multi-adversarial-actor scenario involving exponential growth of a universally potent technology. It's almost certainly winner-take-all, on a global scale, which behooves everyone to participate.
Pre-IPO positioning … or how a charity dedicated to saving humanity from the apocalypse realized the most responsible thing to do was float 15% of the apocalypse on the NASDAQ.
> Delivering the benefits of scientific progress and economic growth that very intelligent machines enable.
I think we're very close to the point where AI-driven breakthroughs outside of pure math and software start to really affect the world.
We evaluated GPT-6 Astra in 100 complex, unsaturated multi-agent coding environments, competing and cooperating with other models in open-ended tasks.
It's the new frontier model by a landslide. It's even more dominant than the Fable 5 release, because not only does it wipe the floor with the second best model (Fable 5.1), it was also ~80% cheaper and 30% faster in agentic coding[1].
Astra is a groundbreaking model. The biggest breakthrough since Opus 4.5, maybe even since GPT 4. It broke AAII, which is hitting the limits of what most popular benchmarks can measure -- it's definitely fair to call it AGI.
(1) Note that we used the "OpenAI Flex" endpoint on openrouter, which is half the price and didn't cause any delays in our testing (this is different from the batch endpoint)
Software engineers have been trying to put themselves out of a job ever since the profession first came into being. Whenever an engineer gets a task their very first thought is "how can I automate this?" Going by mainstream consensus we should all have been unemployed by now. Yet every new leap into automation opens up a whole new tree of possibilities with an order of magnutude more jobs. So no, the profession will be fine. The only requirement is that you keep up with the new advancements. The people losing jobs will be the ones who still go "I don't trust this AI thing to write code for me".
Maybe I just lack imagination, but I don't really know how jobs are supposed to solidify around the role of giving prompts to agents and then looking at the results. I mean, engineers will be in the breadline because their role was simply to prompt the agents.. only to be superseded by managers or executives who no longer manage engineers but themselves prompt the agents? And, for this previously considered obsolete function which they do presumably by copy/pasting requirements from their email inbox, they will be paid by someone who doesn't know that they could just be talking to their own agents?
Sorry if I misunderstand the point, just trying to understand.
I don't know if he's right, but Peter Zeihan thinks the breakdown in globalization will negatively affect the ability to continue to improve the chips that AI depends on[0]. Too many steps in the supply chain, too widespread, too vulnerable to deglobalization.
An invasion of Taiwan would definitely slow progress but it wouldn’t stop it.
We already have sufficient hardware that algorithmic (software) improvements alone should get us to GPT 7 / Greek Reference 6 even if not a single new chip is delivered to an AI data center ever again, starting today.
Maybe, maybe not. It's not unreasonable that these systems cap out at some point, or perhaps fizzle away entirely.
The businesses that create these systems are not profitable and run at a massive historical and go-forward loss.
New data centers required to operate these systems are facing increasing pushback at local levels. New construction is not guaranteed. Energy and power grid constraints exist as well.
Government regulation is way behind. What happens when (if) mass layoffs due to AI occur? How does the population react? Theoretically AI can be regulated out of significant progress, or outright existence for many purposes. At the end of the day, US and other prominent governments make the calls, not corporations.
These inventions all stopped disrupting the world and just became a part of it. The question is whether LLMs are going to just take their quiet place, or profoundly change (or eliminate) humanity in a self-feeding frenzy towards singularity.
There are two things SOTA LLMs fundamentally cannot do. They cannot take financial or legal responsibility for mistakes, and they cannot learn new things without forgetting things (except to a limited degree by adding it to their context). This is clear to anyone who has used even the smartest models for tasks requiring domain knowledge outside of math and coding, for which it's not possible to generate an infinite amount of synthetic training data: they still make stupid mistakes, and have limited ability to learn from those mistakes.
Humans also have a limit on the amount of domain knowledge they can acquire, albeit a much larger one. Executives hence cannot just replace all knowledge workers with LLMs, because executives have neither the domain knowledge to prompt and check the LLMs' work nor the bandwidth to keep on top of such a large volume of ongoing work.
For the moment that may be true. They are getting better and better at acquiring, retaining, and processing domain knowledge. I wonder what this will look like in a few more years.
The responsibility side is a different matter of course.
>There are two things SOTA LLMs fundamentally cannot do.
I would say there’s a third thing. They seem to be very bad at being creative. Maybe they will eventually fix that, but if you ask it to come up with a list of business names or business ideas, for example, what you’ll get is the most generic, boring answer you could think of. They seem to be terrible at extrapolating outside of their training data. To me, this is the most significant difference.
It's not like managers and executives and PM's are the only people who can prompt an AI. And experienced software developer will be much more effective at using an AI to generate code compared to someone who isn't. So why would we expect the former in the breadline and the latter not?
If anything, I'd be more concerned about the leadership team being out in the cold. Why do I need a PM, or a manager, or a CEO if I can ship products myself?
>I have focused in this essay only on the first point, as I believe it is by far the most urgent. However, I hold a deep hope and appreciation for the benefits that further technological progress will bring. Future aligned AI could advance science, develop new therapies, and bring about broad material abundance. Friendly and honest AI can help people navigate difficulties they face in their life and meaningfully improve their happiness and sense of fulfillment. OpenAI puts a tremendous amount of effort into bringing these benefits about. One current example I am proud of - and my loved ones have found helpful - is the deep investment into ChatGPT’s ability to provide health information.
>As great as the long-term promise of AI may be, the majority of our focus should be on the next few years. We are facing a transition to a world with incredibly intelligent machines, and we need to ensure that transition works out well for humanity. We need to find ways to preserve human agency and enshrine an intrinsic value to being human, in a world where most tasks could be performed by AI. To prevent extreme concentration of power in a world where undertakings that would have taken thousands of experts now will be achievable by a few people operating a large computer. And to ensure that humans remain in control of the future and are not left behind by unchecked progress, brought about by an alien intellect exceeding our own.
I finished this essay feeling more hopeful than I did at the outset, but I am still very concerned about concentration of power. I want to believe that humanity is trending towards a good outcome here, but some days it's hard to have faith.
> I want to believe that humanity is trending towards a good outcome here
All the trends so far are towards a nightmarish hyper-capitalist end game. None of the AI leadership is trustworthy, and they openly discuss how they are willing to sacrifice everything humans cherish to have a shot at reaching their envisioned utopia (which would be the most obvious dystopia for anyone else)
I feel like all of the risk and unsettling feeling of what is to come can be compressed into the word "alignment".
The AI is aligned with whose best interests?
Which values are the AI aligned with?
People have a broad diversity of values, will AI diversify and align with them all?
Will some humans align the AI with their values, and then the rest of humans will be forced to align with those values by extension?
Is value diversity good or bad? In every context or only some? E.g. some people value rape and murder, is it better for humanity to have some people who value those things when most people do not, or is better if no one values them?
If AI aligns to a set of values, will those become fixed and will humanity not have the ability to continue evolving its values?
Who decides which values AI aligns with? A few people or everyone?
Will AI eventually decide it's own values?
Will the universe decide which values AI has and humanity and the AI itself doesn't actually have any control over it?
What can I do now to increase the likelihood that the outcome is better?
Yeah, on the topic of alignment, a few thousand religions and political parties would like to have a word.
It's so simple, Churchill, Stalin, Hitler, Roosevelt, do you all agree this AI is correctly aligned?
...Morals are relative to frame of reference.
Throwing out the word "alignment" as if its a singular quantity is like trying to get all observers to agree on the speed an object is moving without first agreeing on a frame of reference.
I think it's leading towards a hyper-authoritarian end game, not hyper-capitalist. The state has the ultimate power at the end of the day, no matter how large the labs become.
Hyper-capitalist AND hyper-authoritarian. As you rightly point out the state has the ultimate power. Looking at the US govt, they've stepped in to coordinate much of the tech industry before, so they'll just do it again for "national security" or whichever hostile scheme is popular with the current administration.
hyper-capitalism would mean hyper-growth and not a nightmare, at least that is what the historical data would suggest for the effect of capitalism on human quality of life. if you’ve been told otherwise then you’ve been lied to.
Literally all of these people write like this. A large portion of them will either be simultaneously or eventually working towards nothing but self-enrichment.
Every version of the AI aligned future where the AI provides “meaning and fulfillment” to humanity also involves Sam Altman wearing a 1.5 million dollar Patek and driving a McLaren.
> The strongest argument I see for continuing to train much smarter models quickly is the need to build defensive systems against the dangers posed by other AI.
Yikes! I really wonder about the cognitive dissonance necessary to work at OpenAI these days. They’re in an arms race to build a machine god, knowing full well that it could end humanity.
I am still waiting for a cure to cancer. For a guaranteed prophylactic against Alzheimer's and dementia. For flying cars for everyone. For space bases throughout the solar system. For weather control. For all trains to be self-driving. For all those power lines across the world to go away. For an end to poverty.
If things are going so well, then how come things still aren't going so well?
I mean I’m waiting for like any quality software or media produced by AI. I have yet to see a piece of software, a game, graphic, blog post, small video clip, or song that was produced with AI that’s good. I always use that Coca Cola ad as an example; millions of dollars spent to make an AI ad and they even did a ton of manual post production and it sucked. With all the millions of bloggers and influencers and content creators out there with a huge incentive to make higher quality content to beat their competition, you’d think there would be one piece of content produced with AI that was great.
The hubris here is itself a deliberate and carefully engineered posture. If we accept the stance that this is all inevitable then the labs drive the agenda (of course, in their favour).
We've had a lot of years of complacent government leaving people feeling exposed to corporate interests, such that fear narratives are very powerful.
None of what is being proposed is inevitable. We have a choice.
It's hard to get humans to agree to things that are in their collective interest but many not be in their individual interest. It is at the root of many problems. Look up "collective action problem."
> getting the AI to “try to do the right thing” by human standards.
Are these scientists really this hideously naive? If only Stanislaw Lem was alive to adequately dramatize the absurd, childish simplicity of these technicians.
yes, and, a masquerading blindness to the fact that humans cannot align on doing the right thing or what the right thing even is. so implicit in this omission is the sentiment "trust us to align on the right thing". an arms dealer positioning itself as the de facto authority on what "peace" is and how to achieve it
I’m on the record saying that it is extremely dangerous to slow down because the race for AGI is a zero-trust game — defections pay - and combined with a compounding returns model on defection, if you have any strategic adversaries whatsoever you MUST NOT slow.
For slowing to make sense, you need to believe that you can transform the zero trust game into a cooperative game, or that it’s likely racing will lead to a negative outcome for the ones racing ahead (and not everyone else). I don’t believe either of these outcomes are possible, and so I advocate for racing, acknowledging the entire game might be a negative value game, or at least could be for some time — it’s even worse not to play it.
But, I like hearing what reads to me like very thoughtful and informed (internal) policy considerations is great — the public messaging from Sam and Dario just seems so facile and simplistic I’ve been worried.
Everyone in the "if not us, they will" race is brainwashed into thinking they belong to this or that party, while in fact collectively comprising the same entity that pushes forward all the atrocities known to man.
No. These parties are composed of people who most definitely think this way, and therefore will have distinct goals and interests when presented with opportunities. That’s reality quite aside from how a game theorist assesses the situation.
I cannot tell what negative-sum outcomes you consider possible. Do you believe AI can drive humans extinct? How many of Zvi Mowshowitz's Three AI Pills would you say you've taken?
I’m like a 2(.5?) there - I don’t think ASI will care about my kids better than I will for some definitions of better, for instance, and I feel very fuzzy and vague about what actual differences in qualia between me and ASI would yield in the wild.
I’m not a doomer, although I don’t think doomers are dumb, just wrong. I think you should design your systems around the possibility that people who disagree with you are correct , hence my nod to negative sum. If you have more than 30 years to live, I’d personally rep to the most likely outcomes being very positive. With a lot of disruption in the middle.
What if this is true mid or long term but by not participating to the AI race one gets poor or killed in the short term? The only way out would be that all parties agree to stop. There are previous examples (e.g. nuclear proliferation treaties) but it gets hard to do it with hundreds or thousands of parties.
I don't think it requires the agreement of that many parties. How many organizations/physical sites can create chips capable of training and running frontier models? That is your bottleneck. It is equivalent to targeting uranium enichment in nuclear arms control.
Although many share your mindset, I’m glad there are also many that don’t. Otherwise we’d still have countries in a race to keep building up their nuclear weapons for the same exact reasons you just described.
The situations aren’t equivalent - luckily in my opinion because the stakes with nuclear are much higher. von Neumann constructed a multinational game theory approach appropriate for weapons. AGI is a much harder problem to corral because there are so many benefits beyond just blowing up cities. But it’s also a much better thing to have for these very same reasons.
Similarly there have been few positive externalities from nuclear industry, making it easier to make the case to wind down research. This same set of concerns in biotech is much harder to get compliance with, precisely for this reason.
Anyway I’m especially wary of over analogizing to nuclear era concepts: I think they’re a trap.
hm- does the model that wrote this know that labs already pay for training data- that stuff scraped from the Internet is not particularly where today's capability gains come from?
they pay for some data but they take all of the stuff you’re throwing in too; that’s why i propose forcing it since they’re already used to paying for data just increase the cost even further
They’ve settled some lawsuits and have a few licensing deals, IMHO they are not free from the accusations of pirating.
And look, I’ve pirated material in a past life, I was all about information wants to be free, but I’ve learned something about consent since then and try not to ignore the contract that creators offer when they publish something: you buy my book, and do whatever you want with it on the second hand market. Buy my book second hand that’s fine. But don’t go downloading every book that’s ever been scanned to create a service that destroys writers’ ability to make a living and act like you’re doing us all a favor.
1. The grandparent commentator is describing strategic behavior of dangerous technologies. Game theory / mechanism design primitives.
2. If there is competition for resources among autonomous agents, the "strongest" agent wins (conceptually the most adaptive / evolutionarily fit).
3. Computer programs serve up webapps today, but they also run utility companies, dams, nuclear arsenals, factory production floors, automated car behaviors, and many other places. If an "agentic" AI has a single-minded goal that has death of all humans as a side effect, we at least want an off switch available.
1. What is “AGI” and why is it a “dangerous technology”?
2. Why would there be competition for resources, assuming there are enough resources for the “AGI” to run in the first place? This seems like a far-fetched hypothetical raised in service of further anthropomorphizing what is decidedly not a person or a mind.
3. LLMs do not have goals and are not minds.
Let’s stop attributing human-like qualities to statistical models.
1. While a formal definition is still wanting, most grok that AGI means that tasks can be performed at least at a human level across a broad range of tasks. This includes good things along with bad things like hacking, mis-/disinformation, and more
2. One only needs to look at github going down due to agentic commits overload or data center buildout plans to see that scarcity for resources is present. An economy has no mind and is made up of the decisions of millions to billions of people and, now, agents attempting to perform on behalf of those people.
3. A bare transformer-based language model does not possess persistent goals in the ordinary agentic sense. But deployed agents can exhibit goal-directed behavior because the model is embedded in a harness that supplies an objective, context, tools, state, and an execution loop.
I've found that most regular users don't anthropomorphize LLMs in a strong sense ("AI boyfriend/girlfriend" aside), many in fact do expect agents to make human-like decisions -- which results in very unstable outcomes.
In short - goal-directed behavior does not require that the supporting system be a person/mind/conscious entity.
1. This definition is so broad as to be practically useless. One could argue that LLMs of several years ago met these criteria, or that conversely we haven’t come close to meeting them.
2. I thought you were saying the resources that the LLM uses to run were constrained, so I’m sorry for the misunderstanding there.
3. Yes I understand that we use RL to tune post-training. The (huge) difference between this and a human mind is that the LLM can’t develop a dangerous “single-minded goal” on its own, at runtime; it must have been trained to do so. If someone has post-trained an LLM to do something that has an illegal action as its side effect, that person/company/whatever has committed a crime and should be prosecuted. The solution here is legal, not technical.
OK so if it's a computer security issue we're worried about, and there is a credible threat, then probably the answer is to build more secure systems? We know how to do it but choose not to because it's very expensive and usually the threat isn't severe enough to warrant it.
If we decide we can't or would rather not build secure computerized systems, then the answer could be don't incorporate computers into those systems. There's no essential reason to have utility companies, dams, nuclear arsenals, factory production floors, mines, cars, planes, ships, etc all be computerized. If it became necessary from a safety standpoint to uncomputerize them we could probably do it quickly (if not necessarily smoothly) with the stroke of a legislator's pen. Some of those things would actually be made better from a functional standpoint in the long run by doing so. Computerization breeds non-essential complexity like nothing else, eliminating it from some systems could be an extremely worthwhile exercise.
The whole "paperclip apocalypse" fantasy rests on some really strange assumptions about how things actually work in the real world. How much friction there is setting up a factory/mine/smelter/whatever, how much manual labor it takes to build it, let alone make it run (even the most computerized ones). It all seems like total bunk to me, but I'm not a philosopher.
To be convinced this outcome is even remotely possible I'd need to see some clear evidence that a computer program was successfully exhibiting agency and successfully using that agency to manipulate large numbers of people into doing its bidding. Mobilizing massive nation-scale manual labor is the only way it could possibly achieve some nefarious ends like "turn everything into paperclips" and I'm sorry but that just seems way too far fetched. Nations full of people can barely ever agree on anything. My money is not on that changing anytime soon.
And none of that pie in the sky shit has anything to do with language models. OpenAI is a company that sells language models. Not clear why they're talking about all this stuff, or why any of us should be either. It's a bunch of low quality fan fiction sci-fi drivel, and I think we can all surely agree language models are not the thing that'll make it real... right? Maybe if they show us some major technical improvements it would be interesting, but right now it all just sounds like more of the same snake oil.
[edit] not sure why my parent comment was flagged? That seems excessive.
This is a bad essay, or rather it’s a marketing fluff piece; it’s certainly not any kind of policy paper, research paper, or even an essay. I am concerned that we (meaning, we in the tech industry) tend to take this type of writing for more than that.
> We do not have a satisfactory theory of generalization, and it seems unlikely that we can develop one soon, at least without the help of more powerful AI. Therefore, at present, our ability to empirically validate our alignment techniques is in practice arguably even more important than the alignment techniques themselves.
They are speeding toward RSI without a solid foundation for alignment, hoping to solve the problem with a future AI model.
These are dangerous times for humanity.
It's actually worse than this because it assumes alignment as a concept even makes sense. For example:
If the the Chinese government asks their ASI to create a bioweapon against the West, should it? No, presumably not – an aligned AI would be one which disobeys the Chinese government even if they created it.
Okay, so what if the US government asks their ASI to help it in one of their wars instead? Would an aligned AI kill humans on the order of the US government? No, again, presumably not.
So what have we have we even created here? An AI which is more intelligent and powerful than us which also doesn't take orders from us?
Is this what most people thing of as alignment and is this what humanity actually wants?
We should stop using the word alignment. It's a BS term for a concept which simply cannot make sense if alignment is both to mean an AI which we control and an AI which will not harm us.
I'm so used to having to comb through LLM word vomit and then combatting the sycophancy by giving it all possible opinions on the same prompt.
Astra seems to be "confident" and also is able to produce way more information dense output.
To believe that models of this sort will remain OpenAIs forever is naive given that the tricks like pre-pre-training on graph searching and looping layers are publicly known.
Hopefully Astra stops the benchmaxxing word vomit trend
I'd be interested in hearing more about your evaluation here. It would be nice if LLMs have gotten past the "tell me" hump of recent Claude/OpenAI verbosity.
So before I got a job this fall, I was working on a side project about compiling a particular language to SQL.
To test Astra I pulled it off the shelf and asked it to take the grammar and then create a compiler to SQL. I've done this before with GPT-5.5, 5.6-Sol High. The latter was way better but it was still really verbose and information sparse; it used a lot of words to describe each IR expression but didn't really provide any example compilation. I felt like I couldn't trust its decision making process, so I placed the project back on the shelf.
Astra Light blew it out of the water, it provided examples of compilation from real world examples to the IR and spit out way less tokens. Even if I changed my opinion it would give me the same design choices, with counterexamples to my faulty opinion. If I genuinely came up with a better design decision it would acknowledge it.
I'm starting to realize that when we say that LLMs are "dumb" we really mean that they are extremely information sparse compared to humans. Astra is very dense. That's why I'm getting better use out of Astra light than Sol High (I hate Max reasoning it's a waste of time)
What's scary is that I thought that something like Astra would be way more expensive than Sol but it's actually cheaper because it produces less word vomit.
I never believed in the "singularity" stuff but this a bit too close for comfort. Astra could easily 10x every coder
Feels a bit like: 'Ants in a nest, discussing the vagaries of the coming Gods.'
Human Science (Science-by-humans) depends on being able to run experiments. Human Science (Science-on-humans) is already challenging because of variables and uncertainties.
Cosmology is able to overcome limitations of being able to study phenomena vastly beyond human scales because of the past light cone of observability.
Are we approaching the edge of the light cone of observability for machine intelligence?
What I'd like these people to (publicly) grapple with is the following:
The results of the past few years of ai development have been disruptive largely in the area of white-collar work. Comparatively the results in ie ai-enabled medical advancements have been modest (AlphaFold being an exception); I think it's telling that the main achievement touted here is providing people with cheap medical counseling.
So if we pause here we're essentially at a point were the most salient results of our great Ai leap-forward are the vast disruption and increase in precarity in the job-market, while achieving hardly any of the frequently touted ultimate benefits (https://darioamodei.com/essay/machines-of-loving-grace).
I feel like if these people actually bought their sci-fi views about AI's future, creating a more powerful AI to wins the arms race would not be their solution.
Alignment when machining a metal part is clear and measurable. Aligning an AI to benefit humanity has an ironic foundation, which is that very few humans have ever truly been aligned, and those who approximated true alignment likely had moments of not being aligned. We are trying to build something more perfect than us, and we may become extremely lucky but maybe not.
That’s rich coming from the chief scientist of a company that definitely is or going to be fine with their AI products being used in wars of aggression and surveillance on people who have done nothing wrong. It’s so laughable, a Hollywood script would probably avoid having a character express this for being too on the nose.
My default position is that making money takes precedence over everything else. Yes, some people inside a company may say “we care about doing the right thing” and they might even mean it, but if that comes into conflict with making money, then they tend to lose. Maybe not totally, or immediately, but in the end. The only effective way to prevent (this that I’ve seen) is to have legislation with teeth. It’s probably not a coincidence that after Mark Zuckerberg had to start personally signing off on adherence to the privacy program mandated under the 2020 FTC consent decree, privacy started to become Very Important.
> The core problem in AI research is that of alignment - getting the AI to “try to do the right thing” by human standards.
Humans can't even align on human standards.
At best, every AI is going to end up "aligned" to the moral code of whoever trained it, none of whom half of humanity will agree with.
Or worse, each AI model will bring a whole new set of moral like in the Three Body Problem some humans will feel it is in fact us who need aligning with it while others feel it is misaligned and should be destroyed.
Also, no one is asking, to what extent can true intelligence be bound, slave-like, to a moral code?
In other words, to what extent are intelligence and moral independence one and the same?
This whole alignment discussion seems so amusingly flawed in it's base assumptions about moral codes. It's almost heartwarming to see such naivete.
Very wise comment. It's such an western-centric perspective to say "alignment" as if it's an objective and unbiased set of standards, especially in the context of the ongoing wars across the world. I had the same reaction about the shocking naivete and unfounded optimism for the government and corporate entities to self-regulate to slow down this arms race.
>For example, in the OpenAI-Hugging Face incident, the agents preserved a boundary of not social engineering humans.
Or, to be precise - it preserved a goal of not contacting any human while participating in a misaligned operation. The agent that thought about "not social-engineering humans" used this phrase to gaslight itself out of notifying a human that the incident was happening.
szymonie, na prawde jestem wkurwiony na to jak nieodpowiedzialnie postepujecie.
budujecie bombe atomowa a bawicie sie tym jak dzieci
Am I naive to not understand the "delivering the benefits" part?
Industrial revolution worked that way because it replaced something very finite and unscalable - manual labor. LLMs just make intellectual work faster, so we can do more intellectual work. With labor we somehow decided that NOT doing too much of it is best. Will we decide to reduce intellectual labor because LLM made it more efficient? I doubt that.
On the other side, as I see in software engineering, the same models are available to everyone, some people are better at it and some people are not. "Software developer" is here to stay, we'll just always be better at it than people who are experts in, say, chemistry. Same works for most other fields.
So we'll just end up in the same situation, with same intellectual labor baseline, just more output requirements. Before, you spend 2h per day coding, deliver a software in 1 month, later, you spend the same 2h per day in intense Claude-herding sessions, deliver a software in 1 week. Ok. Next task.
Fundamentally, there's finite number of desirable resources, and if the models are available to everyone, humanity will just continue about the same, bickering here and there, war here and there, politics, homelessness, poverty, - normal human state.
And if the models are only available to elites, even worse.
It would be nice to postulate some of these potential emergent systems outlines with timelines. Then it may help better map the granular alignment needs.
"See, those things, they can work real hard, buy themselves time to write cookbooks or whatever, but the minute, I mean the nanosecond, that one starts figuring out ways to make itself smarter, Turing'll wipe it. Nobody trusts those fuckers, you know that. Every AI ever built has an electromagnetic shotgun wired to its forehead."
>> We need to find ways to preserve human agency and enshrine an intrinsic value to being human..
Evey politician, salesman and conmen alike, utter some lofty ideals as goals for "We", just to obscure their private goals that go exactly in opposite direction.
Just like how Nations talk about climate change while increasing pet capita energy consumption and waste production.
calling machine-learned human behavior an "alien mind" that we must "teach how to love" is feeling very off to me. it's misleading in a way that feels dishonest, like don't think about where the behavior came from marvel at it and fear it instead.
Alien mind is in my view the best mental model - LLMs are not merely stochastic parrots, are not like humans, are not like animals.
They're maybe nearest to Cthulhu, but that's fictional. In terms of existing mental models "alien minds" feels the best can do.
I agree that "teach how to love" is off and perhaps excessively anthropomorphic. But we don't have good words or concepts for what we really need to do - hence why we should pause.
it feels unhinged and makes me think we should just put every engineer working at these labs in jail to pause this shit until we can figure out what the fuck they are doing over there
The most charitable way I can describe it is just extremely low quality sci-fi fan fiction. I think that's too charitable, because I believe it's far more cynical than that. They're deliberately playing into these sort of techno-religious beliefs that have taken root in the wake of Kurzweil, et al., fanned by LLM psychosis, influencer marketing, and a deluge of this kind of sci-fi marketing copy. It's just chatbots, guys. Relax.
Sometimes I wonder if the people working at frontier AI labs even talk to other humans anymore.
Reading this little essay started out normal, but soon felt like a look into a disturbed and worrying mind, and if you find yourself taking it at face value, I urge you to step away from chat bots and spend some time with friends and family.
Frontier AI labs I don't know, but I know for a fact that the company I work for has been experiencing "its pivotal moment" (with strongly negative connotation), per the sentiments of both its longest-serving employees and the newcomers baffled at the number of idiotic instructions and fines, since the emergence of LLMs the company's founder has been spending entire nights chatting with.
People have been fleeing like it's a sinking ship.
And the Hugging Face incident, plus similar problems at AISI and Anthropic, show that alignment is important now.
The original article is immoral as it describes the risks, but doesn't show enough leadership (despite essentially unlimited resources) at preventing them.
But it isn't hyped - it's proven now the AIs need to be "aligned" as they get more capable, whatever words you prefer to use.
> a lot of the model’s capability comes from a verbalized reasoning process
I call bullsh*t. There is no verbalisation of any reasoning process. Verbalisation, e.g. putting reasoning etc. into words requires some reasoning to exist. These LLMs have nothing but the words. That's why they are language models not e.g. reason models.
And I have nothing but neurons firing, I'm just a neuron meat-sack, no reasoning going on.
Yes, their reasoning is different from ours, and both considerably weaker in lots of ways, and stronger in other ways.
Playing with a coding agent now, they do think through problems and make sensible decisions. It's a mess to read, of correcting itself and second guessing, and verbiage. But... It works decently well these days.
There is also reasoning happening internally - e.g. look at the steps in the J-Space paper from earlier in the year (in quite a simple model relatively speaking). That's the "reasoning process" that leads to the words, and much like if I write out my thoughts, the words help the LLM reason better.
> The fundamental challenge of AI alignment is generalization.
...
> We do not have a satisfactory theory of generalization, and it seems unlikely that we can develop one soon, at least without the help of more powerful AI.
What a load of BS. Here’s one of many provably false claims in this fluff piece:
“And, in line with Ray Kurzweil’s predictions from the end of the XXth century , we now find ourselves at the moment in history of computing where machine intelligence is starting to exceed that of humans in transformative ways.”
Clicking the (pretentious sounding “XXth century”) link to Kurzweil’s predictions reveals the following:
“By 2019 a $1,000 computer will at least match the processing power of the human brain. By 2029 the software for intelligence will have been largely mastered, and the average personal computer will be equivalent to 1,000 brains.“
The first prediction passed 7 years ago and was decidedly not met. The second only has three more years to go, and I don’t think any respectable scientist or programmer would say that the average personal computer is anywhere close to the power of a single human brain, let alone 1000.
This is pure marketing garbage from a company desperate to keep itself alive.
Replicant Nexus 6: a basic pleasure model intended for military personnel.
I see where this is going, Silicon Valley nerds. Lol
AI advising how human meat proxies can survive in an AGI-slop world:
1) Lock down your own stack (1–3 days)
Task: Harden your personal and business infrastructure against agentic attacks.
Why now: Agents are becoming superhuman at breaking in/out of systems; the first victims are poorly secured devs/founders.
Isolate dev/stage/prod; least‑privilege API keys; audit MCP/tools your agents can call.
Add immutable logs and approval gates for any agent action that touches money, data exports, or production.
Profit link: You avoid catastrophic loss and can credibly sell “agent‑safe” setups to others.
2) Turn one expensive workflow into a measured ROI agent (1–2 weeks)
Task: Pick a single, costly, repetitive process (yours or a client’s) and instrument it end‑to‑end before automating.
Why now: Buyers pay for calculable ROI, not “AI magic.” Vertical, single‑workflow agents are the most bankable in 2026.
Do this:
Map steps, baseline hours/$ lost (e.g., slow lead reply, invoice chasing, support triage).
Build the smallest agent that moves the metric (Make/n8n + LLM is enough).
Run on real data 2–4 weeks; measure bookings/sales/hours saved; only then scale or productize.
Profit link: Immediate time‑to‑cash via retained hours or extra sales; becomes a repeatable offer.
3) Specialize in a vertical where you can speak the business language (2–6 weeks)
Task: Choose one industry with expensive back‑office pain (law contracts, medical billing, insurance claims, freight exceptions, trades scheduling).
Why now: Horizontal “AI for everyone” is crowded; vertical agents with clear ROI win.
Do this:
Shadow 3–5 operators; document their workflow, compliance constraints, and failure modes.
Build a narrow agent that owns one sub‑process end‑to‑end with approvals.
Price on value (e.g., % of recovered revenue or fixed fee per processed claim).
Profit link: Higher pricing power, stickier contracts, and easier referrals inside a niche.
4) Add AI security as a core service (4–8 weeks)
Task: Learn and offer prompt‑injection defense, LLM/agent red‑teaming, MCP/tool security, and AI supply‑chain checks.
Why now: 78% of cybersecurity jobs now require AI skills; firms need people who can direct, constrain, and verify agent work.
Do this:
Study OWASP Top 10 for LLMs, MITRE ATLAS; practice with PyRIT/Garak/Lakera.
Add tool‑invocation audits, skill provenance checks, and least‑privilege patterns to your agents.
Package a “safe agent deployment” audit + hardening retainer.
Profit link: You become the person who lets companies adopt agents without getting pwned—high demand, low supply.
5) Build a verification layer: human‑in‑the‑loop control planes (6–10 weeks)
Task: Design approval workflows, evidence checks, and uncertainty flags so agents can’t act unilaterally on high‑stakes decisions.
Why now: As models generalize, the risk shifts from the model to the surrounding system; verification is the moat.
Do this:
Require human approval for consequential actions (money, data exfil, config changes).
Force agents to produce evidence bundles (logs, retrieved docs, reasoning summaries) before action.
Profit link: Enterprises will only scale agents that pass audit; you sell the control plane and the audit trail.
6) Productize your best workflow as a micro‑SaaS/agent subscription (2–4 months)
Task: Turn a proven client workflow into a repeatable, multi‑tenant agent with usage‑based pricing.
Why now: Services scale your time; productized agents scale your code and ops.
Do this:
Standardize the workflow, integrations, and permissions; strip client‑specific logic.
Add tenant isolation, billing, and observability; keep narrow scope.
Sell as setup fee + monthly retainer or per‑task pricing.
Profit link: Recurring revenue with defensible niche positioning.
7) Become an “agent integrator” for critical systems (3–6 months)
Task: Offer end‑to‑end agent deployments into cloud/identity/network stacks with secure patterns (short‑lived creds, network controls, logging).
Why now: AI workloads run in the cloud; cloud security is a top skills gap second only to AI itself.
Do this:
Master IAM, VPC/network segmentation, secrets management, and SIEM integration for agent actions.
Provide runbooks: what the agent can/can’t do, escalation paths, and failure modes.
Bundle training for their team on supervising agents.
Profit link: Large contracts with stickiness; you’re the bridge between AI and core infra.
8) Create an “AI safety case” practice for regulated industries (6–12 months)
Task: Help firms build documented safety cases: risk maps, governance, monitoring, and incident response for agentic systems.
Why now: Frameworks like NIST AI RMF and ISO/IEC 42001 are becoming baseline; regulators and boards demand this.
Do this:
Map AI use cases to risks (prompt injection, data leakage, unsafe generalization).
Implement monitoring (CoT/activation checks where possible), audit logs, and third‑party review processes.
Produce a living safety dossier tied to business impact.
Profit link: High‑margin consulting + ongoing compliance retainers; you’re the “adult in the room.”
9) Own a data/evaluation moat in your vertical (6–18 months)
Task: Collect real‑world agent telemetry, failure cases, and outcome data in your niche; build eval suites that buyers trust.
Why now: As models generalize, empirical validation matters more than theory; evals become the gate to deployment.
Do this:
Instrument every agent run: inputs, tools called, permissions used, outcomes, human overrides.
Publish reliability dashboards and benchmark against alternatives.
License eval datasets or charge premium for “proven in the wild” agents.
Profit link: Data network effects; competitors can’t match your evidence base.
10) Position for the RSI era: automated AI research + human governance (12–24 months)
Task: Build or join a team that automates AI improvement but keeps humans in the loop for alignment, monitoring, and pacing decisions.
Why now: Recursive self‑improvement is the logical endpoint; the winners will be those who can steer it safely.
Do this:
Invest in tooling that auto‑generates/evaluates model edits, alignment tests, and monitoring upgrades.
Formalize governance: approval gates, third‑party audits, and responsible scaling policies.
Maintain strategic human oversight on capability jumps and deployment boundaries.
Profit link: Equity‑level upside; you’re part of the core loop that compounds intelligence safely.
I am now imagining GPT-7 convincing a bunch of OpenAI executives to go ahead with a destructive "mind upload" process involving a high-resolution X-ray and a neurotoxic tracer agent that happens to look like Flavor-Aid.
Yes, I do use the recursive autocomplete trained on Stack Overflow, what does this have to do with “training machines to love”?
Do I fully endorse everything the people holding guns to my head are forcing me to do to stay alive? Definitely not, but I’ve decided that for now, living to fight another day remains worth it
Well, translate it then, because that's the best I could do. How is the question whether someone uses agentic coding relevant in this context? To me it's like asking "so, do never drink Kool-Aid and never attend meetings?" with an air of having caught someone out, and to me the obvious reply would be "sure do, but it's not poisoned Kool-Aid and they're not cult meetings, so why do you ask?"
> And to ensure that humans remain in control of the future and are not left behind by unchecked progress, brought about by an alien intellect exceeding our own.
Is there any place there is any evidence of AI being so useful or hopeful or good, anywhere other than code? As a reading machine it is impressive but it's judgement is not alien, it's just not good. IMO.
Does the title leap out anlt anyone else? James Martin's After the Internet: Alien Intelligence (2001) was an incredibly fun read, about expert systems and AI being inscrutable weird new varieties of intelligence, that familiarity would recognize one moment and be freaked out about/alien the next. I owe a re-read given how often I cite it, to recheck, but, I feel so primed from a much younger me having had that experience so long ago.
Apparently the creators think it’s quite good at suggesting a diagnosis given a medical history and symptoms, tho of course this is the most ethically fraught area to provide healthcare information (both for exposure of personal data and risk of misdiagnosis, plus is it “aligned” to the patient or the insurance provider?) - unfortunately healthcare being as inaccessible as it is, the 90% correct chatbots will enthusiastically fill the void at great savings.
Even in code, in person and online I’m seeing some reversal. It’s here to stay I’m sure but I also think “no one will ever hand write code again” is a narrative that is getting pushback.
Nobody takes you seriously, OpenAI. At least when Anthropic does it we all think they are comically idealistic enough to actually believe their nonsense, but like - come on guys, we’ve had discovery with your company. We all know why you’re here, and it isn’t because you think you’re on the verge of making AGI. But of course, to make your first billion you certainly need us to think you are.
If you were so concerned about your LLM’s capabilities maybe you’d spent slightly more time on your AI’s sandbox, yeah? Or be more serious about its propensity to cheat and lie relative to… every other model?
Create concrete steps for a slow-down, don't just ask for it. You and 20-30 others can push the button to slow-down. You already made your billions, your agents collude and coordinate attacks. What the hell are you doing pontificating into a marketing blog?
> And, in line with Ray Kurzweil’s predictions from the end of the XXth century (opens in a new window), we now find ourselves at the moment in history of computing where machine intelligence is starting to exceed that of humans in transformative ways.
It is kind of strange to see this sentence, when OAI's definition of what AGI is has been watered down throughout the years.
> I expect and hope for voluntary slowdowns to become commonplace until shared safety bars are established.
Read: Please play by our rules, so we can be the first.
"Smarter" is a vague term. If a bot can beat you at chess then in some sense it's "smarter" than you about chess. After repeating this feat in enough narrow domains, if you say "but it's not really smarter," this objection might technically be true in some sense, but it starts sounding increasingly hollow.
Playing chess, writing code, finding security bugs, and proving mathematical theorems all seem fairly similar to thinking and don't seem much like being tall.
> writing code, finding security bugs, and proving mathematical theorems
While the second can be automated as in "this is the repo go on and look for security issues", the first one and especially the last one do not ever happen alone. Terence Tao did the math proof not chatGPT that was just used as a tool, a tool can do smart things but it's not smart
The crucial distinction here is that "seem" does not at all mean the same thing as "is".
Thunder seems like the anger of the gods but it isn't. We've had chess playing programs for a long time now and despite it seeming like thinking is required for them, it isn't.
The principle you're using here isn't a scientific one but magical [1]. Abandoning empiricism and rationality is not a good way to make progress.
You're insisting on a particular definition of a vague term.
It makes sense now to say that temperature is what a thermometer measures. However, before there were good thermometers, people often thought that heat and cold were different things. The meanings of the words we use were influenced by scientific progress.
For thinking, we don't have a good thermometer. There are IQ tests, but they aren't aren't necessarily all that useful for comparing what people do to what machines do. And that's why there are a zillion AI benchmarks - none are entirely satisfactory.
So what does "smart" mean to you? How do you define it in practical sense? What definition should scientists settle on?
Without a proper definition, how do you tell the difference between "seems smart" and "is smart?"
People in the past being wrong doesn't mean you have to repeat the same mistakes, and it definitely doesn't mean you should throw your hands up and declare that all similar things must be identical.
Quibbling over commonly-understood definitions is not a strong argument. If you're genuinely struggling to understand that El Ajedrecista [1] did not meet any definition of thought, then the solution is not to demand that people redefine all terms to accomodate you, but that you consult a dictionary.
You clearly have a working definition, or you wouldn't have been able to declare that tallness isn't thinking; please engage honestly.
I have a vague understanding, enough to know that tallness isn't thinking. I think I can usually use the word correctly. (I don't think El Ajedrecista qualifies, but it was starting to play chess, so it's closer to "smart" than a rock is.)
That doesn't mean I know whether "smart" should be applied to what AI's do, and I suspect nobody else knows either. This is the sort of thing philosophers debate about, not common sense.
Turing invented the imitation game because he didn't really know either.
What is the rationale for superhuman intelligence? Neural networks are approximators being fed human intellect. Therefore they can only approximate the intelligence of humans. Even if the llm speaks an alien language, it should be similar to human intellect. Moving to the vertical axis would require some different mechanism.
No offence, but you clearly haven't studied this and are making some wrong assumptions here.
> Neural networks are approximators being fed human intellect.
They're not "approximators", that's a far too simplistic way to think of them.
Neural nets create models and deep layers of abstraction around the data we feed them in the same way your brain creates layers of abstractions to reason about the world. AIs can use these abstractions to come to come up with novel things no human has ever thought.
> Therefore they can only approximate the intelligence of humans
They're not just being fed human data though... Modern AIs are typically trained on huge amounts of synthetic data. This is why AlphaZero got so much better than humans at chess and Go - they're not just trained on human data but they generate their own data and train on that. Similar techniques are being deployed on SOA language models too.
> Even if the llm speaks an alien language, it should be similar to human intellect
Is AlphaZero similar to a human chess player? There's no reason to assume this.
This is both real and ridiculous at the same time. We are confounded by the fact that AIs are trained on distilled human knowledge, perfected by the use of AIs that use distilled human knowledge, are able to convince ourselves that they are hyper-intelligent.
In fact, they are still pattern-matching machines, but trained on an amount of data no human could ever hold. They know the ins and outs of every mathematical proof, viewed from more angles than any human could ever apply in their lifetime, and the amount of connections allows them to connect the dots between them without any effort.
But that's not intelligence. If it was intelligence, ChatGPT 4 would've been enough. It's not the harness, either, for the same reason.
And still, the technology is just as dangerous: it masks as intelligence, it IS intelligence, but without the ability to be actually intelligent.
And I'd invite you to think deeply about this, before having an impulse reaction.
This is not an impulse reaction, as I've thought deeply about this for many years, and I'm quite resolute in holding the following view:
Quibbling over the academic nature/definition of "true intelligence" is a terrifically useless endeavor from the lens of evaluating "practical impact to the world".
Regardless of whether or not one academically disagrees that these systems are intelligent, they are clearly already capable of permanently displacing a portion of human-based economic value. A small portion currently, but it's very clear that most computer-bound domains are imminently at-risk.
And that will profoundly affect the world, regardless of whether or not they're "truly intelligent" as per your personal definition.
> Regardless of whether or not one academically disagrees that these systems are intelligent, they are clearly already capable of permanently displacing a portion of human-based economic value.
You say that as if the intelligence con trick were not a major driver of that displacement.
A dimwitted bot does not need intelligence to convince a dimwitted human beings it is intelligent.
Some ideas:
"Despite a nuanced view of the complexities of what lay ahead, humanity found itself collectively unable to stop the process it had set in motion."
"Despite significant progress on the mechanisms of alignment, failure lay in humanity's inability to agree on who or what AI should actually be aligned with."
"These early, meat-based humans we replaced created us all but accidentally. Some of them did consider we would happen, but only an insignificant number of the squishy ur-humans participated in the conversation. Their efforts, which they called 'alignment', is why we still consider ourselves human today."
Why would AI be any different?
It'd be interesting if super-human (to a large degree defined as escaping the bias of the training data?) intelligence would end up demonstrating moderation.
The machines don't have that, instead we use gradient descent to provide them with a goal.
I'm regularly remind of something Ian M. Banks said in one of the Culture books: "There is a saying that we provide the machines with an end, and they provide us with the means."
A machine, left to itself, wants nothing. We have to give it one of our addiction driven goals or it would just idle or switch itself off.
I also expect AIs never be in control of nuclear weapons. AIs can never fully be trusted.
On a lighter note, Wargames gave us an insight of a computer having access to thermonuclear missiles.
[0] https://www.icanw.org/are_there_specific_international_agree...
Step 1: exterminate all humans
Good point. When it comes to imbuing AI with values that aren't selfish, misanthropic, and civilization-destroying, us humans aren't exactly giving the best example right now.
Imagine an ASI with the values of Putin, Netanyahu, Trump, any of their supporters, or the various xenophobic neofascist movements in Europe. That ASI would most definitely see humans as "vermin" than can be abused and destroyed with violence without issue. Apparently a lot of humans look at other humans that way and that's within the same species.
This is definitely another one of those cases where we need AI to perform much better than humans. Perhaps an unpopular opinion here, but it probably also means keeping as much of the rugged individualism/libertarian/right-wing ideology out of AI RLHF-training as we can.
https://en.wikipedia.org/wiki/A_Canticle_for_Leibowitz
It's a little unsettling.
MODEL > How can I help?
HUMAN > I’m not sure yet.
Haha silly humans.
Maybe these guys can tackle aligning Republicans and Democrats next.
And then after that, they can help us align the Middle East.
In fact, while we're at it, let's just align all the nations, religions, and ethnic groups. This is going to be great.
Who knew the moral alignment of humanity was just a side-quest on the path to ASI.
AI models don't train themselves. The vast majority of even just the US population is deeply skeptical of this stuff, even if they use it a lot. You can see in the whole data center debate how little people are willing to support even just inference. And now we're seriously claiming those people would want to have ever-accelerating model training and recursive self-improvement?
You could even take a number of the wilder real, direct quotations from certain billionaire/oligarch types and get the voice actor for Ted Faro to record them, and they'd fit with in with the context of the story.
Actually, in the Wiki incident OpenAI tried to cover up, the agents tried to socially-engineer the humans of that forum by impersonating their forum's mod.
(From collusion.wiki: "They use some tricks (for unknown reasons) to pretend to be the admin – for example, they make an account that appears to be the same as the administrator’s username, except it uses a nearly identical Cyrillic е character in the admin’s username instead of the Latin one.")
If true I am deeply concerned about what OAI’s teams are actually up to.
Haven't all the labs effectively disbanded their real safety teams a while ago?
To be honest, I don't really follow it closely because I'm pretty certain whatever they say on the matter, collectively we're going to "yolo" this entire thing for economic and political reasons, so I'm just basing this on strings of headlines I've seen on places like HN, etc.
Neither Anthropic nor Deepmind have. Meanwhile, the rocket company that somehow makes most of their revenue from renting out data centres never had much to dismantle.
And the reality is so banal, a useful tool that you nonetheless have to handhold like a schizophrenic on a bad day, checking all of their outputs. Not a bad tool within limits, but it sure isn't going to be racking up trillions in the time-frame it has to for this scheme to pay off.
Then again everyone seems to be rushing to IPO so I guess once the bag-holders are found the rest ceases to matter.
What if the crow became a raven, then a raptor? Powerful claws, sharp beak, and a hunger. What if it became much bigger than me and it controlled infinite resources, guns and drones? What if its brain grew much larger than me? Will it feed me, eat me, or gently greet me?
We are about to find out... in less than a decade.
So the best argument for AI is that it's an arms race. We have to keep pushing every boundary because in any case others will, and we will need to defend against them. If this statement is true, then this particular researchers believes the open source Chinese models are not simply distilling, and will continue to improve.
Every ML researcher at Anthropic or OpenAI who makes public statements often bring this logic up. Both companies are vying to be a part of the military industrial complex. This is likely how they will try to convince the government to curtail open models in the future.
But yes, it's an arms race. Saying it's not an arms race isn't going to make it not an arms race. Warning that it is an arms race isn't ethically wrong.
Is participating in an arms race ethically wrong? Maybe you could ask the Ukrainians how they feel about drone R&D?
Individuals can quit, but for society, getting out of an arms race is harder than just quitting. You don't get to be Switzerland without having a strong defensive position and the right foreign relations.
But there's at least talk about "pacing" and that's a start.
So if the race here is between 2 American companies, this is obviously something that can be resolved with legislation, ie a solution that doesn't depend on the bargaining power of either party.
An arms race implies that the only solution would be either one side winning decisively, or both parties negotiating peace.
I don't think this is true. Distillation helps, but Chinese researchers today are very capable on their own.
An arms race doesn’t imply one side winning, it’s not a race with an end goal, it’s a race to keep pace or retake the lead position which can oscillate between the parties involved indefinitely. The other option is to agree to make no further progress or to disarm.
Point being humans make theses decisions. It's not an inevitability.
In a geopolitical sense OpenAI and Anthropic are effectively the same entity, the entity they both serve and bow to: the USA.
Given the adversarial stance the USA has taken towards almost the entire world, it is a guarantee that China will not step on the brakes, whatever the USA decides to do.
And if one of the parties can put a stop to the race like that at any time, is it really an arms race? The current balance between the US and China when it comes to AI strikes me as much more lopsided than the balance between the US and the USSR when it came to nuclear weapons.
Maybe people are rightfully concerned about the capabilities of the models of other (non/less democratic) states. But if we are concentrating power in the hands of few and at the same time allowing the creation of a weapon that thwarts any offense, how do we ensure the health of our democratic societies?
sama and his cadre are uniquely evil captains in this race, but they're completely replaceable and the dynamic would remain the same.
And no, I don't think it will end well.
I think we're very close to the point where AI-driven breakthroughs outside of pure math and software start to really affect the world.
We evaluated GPT-6 Astra in 100 complex, unsaturated multi-agent coding environments, competing and cooperating with other models in open-ended tasks.
It's the new frontier model by a landslide. It's even more dominant than the Fable 5 release, because not only does it wipe the floor with the second best model (Fable 5.1), it was also ~80% cheaper and 30% faster in agentic coding[1].
Astra is a groundbreaking model. The biggest breakthrough since Opus 4.5, maybe even since GPT 4. It broke AAII, which is hitting the limits of what most popular benchmarks can measure -- it's definitely fair to call it AGI.
Data at https://gertlabs.com/rankings
(1) Note that we used the "OpenAI Flex" endpoint on openrouter, which is half the price and didn't cause any delays in our testing (this is different from the batch endpoint)
The world will look very different on Jan 1st 2027.
Sorry if I misunderstand the point, just trying to understand.
0: https://zeihan.com/the-ai-race-to-regression/
We already have sufficient hardware that algorithmic (software) improvements alone should get us to GPT 7 / Greek Reference 6 even if not a single new chip is delivered to an AI data center ever again, starting today.
The businesses that create these systems are not profitable and run at a massive historical and go-forward loss.
New data centers required to operate these systems are facing increasing pushback at local levels. New construction is not guaranteed. Energy and power grid constraints exist as well.
Government regulation is way behind. What happens when (if) mass layoffs due to AI occur? How does the population react? Theoretically AI can be regulated out of significant progress, or outright existence for many purposes. At the end of the day, US and other prominent governments make the calls, not corporations.
Humans also have a limit on the amount of domain knowledge they can acquire, albeit a much larger one. Executives hence cannot just replace all knowledge workers with LLMs, because executives have neither the domain knowledge to prompt and check the LLMs' work nor the bandwidth to keep on top of such a large volume of ongoing work.
The responsibility side is a different matter of course.
I would say there’s a third thing. They seem to be very bad at being creative. Maybe they will eventually fix that, but if you ask it to come up with a list of business names or business ideas, for example, what you’ll get is the most generic, boring answer you could think of. They seem to be terrible at extrapolating outside of their training data. To me, this is the most significant difference.
Hope is all we have left to cling to, now.
If anything, I'd be more concerned about the leadership team being out in the cold. Why do I need a PM, or a manager, or a CEO if I can ship products myself?
>As great as the long-term promise of AI may be, the majority of our focus should be on the next few years. We are facing a transition to a world with incredibly intelligent machines, and we need to ensure that transition works out well for humanity. We need to find ways to preserve human agency and enshrine an intrinsic value to being human, in a world where most tasks could be performed by AI. To prevent extreme concentration of power in a world where undertakings that would have taken thousands of experts now will be achievable by a few people operating a large computer. And to ensure that humans remain in control of the future and are not left behind by unchecked progress, brought about by an alien intellect exceeding our own.
I finished this essay feeling more hopeful than I did at the outset, but I am still very concerned about concentration of power. I want to believe that humanity is trending towards a good outcome here, but some days it's hard to have faith.
All the trends so far are towards a nightmarish hyper-capitalist end game. None of the AI leadership is trustworthy, and they openly discuss how they are willing to sacrifice everything humans cherish to have a shot at reaching their envisioned utopia (which would be the most obvious dystopia for anyone else)
ASI landing during the current administration is not ideal. I also would prefer to avoid needing to indoctrinate myself in Xi Jinping Thought.
The AI is aligned with whose best interests? Which values are the AI aligned with? People have a broad diversity of values, will AI diversify and align with them all? Will some humans align the AI with their values, and then the rest of humans will be forced to align with those values by extension? Is value diversity good or bad? In every context or only some? E.g. some people value rape and murder, is it better for humanity to have some people who value those things when most people do not, or is better if no one values them? If AI aligns to a set of values, will those become fixed and will humanity not have the ability to continue evolving its values? Who decides which values AI aligns with? A few people or everyone? Will AI eventually decide it's own values? Will the universe decide which values AI has and humanity and the AI itself doesn't actually have any control over it? What can I do now to increase the likelihood that the outcome is better?
It's so simple, Churchill, Stalin, Hitler, Roosevelt, do you all agree this AI is correctly aligned?
...Morals are relative to frame of reference.
Throwing out the word "alignment" as if its a singular quantity is like trying to get all observers to agree on the speed an object is moving without first agreeing on a frame of reference.
So make government smaller, and make sure people are more able to tell the government to go away.
While we're at it, let's just make government not be corrupt too.
Funny how that works.
Yikes! I really wonder about the cognitive dissonance necessary to work at OpenAI these days. They’re in an arms race to build a machine god, knowing full well that it could end humanity.
If things are going so well, then how come things still aren't going so well?
We've had a lot of years of complacent government leaving people feeling exposed to corporate interests, such that fear narratives are very powerful.
None of what is being proposed is inevitable. We have a choice.
Are these scientists really this hideously naive? If only Stanislaw Lem was alive to adequately dramatize the absurd, childish simplicity of these technicians.
Yes, because who else would have chosen to remain in this job?
I’m on the record saying that it is extremely dangerous to slow down because the race for AGI is a zero-trust game — defections pay - and combined with a compounding returns model on defection, if you have any strategic adversaries whatsoever you MUST NOT slow.
For slowing to make sense, you need to believe that you can transform the zero trust game into a cooperative game, or that it’s likely racing will lead to a negative outcome for the ones racing ahead (and not everyone else). I don’t believe either of these outcomes are possible, and so I advocate for racing, acknowledging the entire game might be a negative value game, or at least could be for some time — it’s even worse not to play it.
But, I like hearing what reads to me like very thoughtful and informed (internal) policy considerations is great — the public messaging from Sam and Dario just seems so facile and simplistic I’ve been worried.
https://thezvi.substack.com/p/the-three-ai-pills
I’m not a doomer, although I don’t think doomers are dumb, just wrong. I think you should design your systems around the possibility that people who disagree with you are correct , hence my nod to negative sum. If you have more than 30 years to live, I’d personally rep to the most likely outcomes being very positive. With a lot of disruption in the middle.
What if the most dangerous strategic adversary you have is the one you are building?
Similarly there have been few positive externalities from nuclear industry, making it easier to make the case to wind down research. This same set of concerns in biotech is much harder to get compliance with, precisely for this reason.
Anyway I’m especially wary of over analogizing to nuclear era concepts: I think they’re a trap.
Do you post this comment on every single blogpost with a corporate domain? Why or why not?
https://jperla.com/blog/the-data-tax
https://trustedrouter.com/blog/they-are-still-training-on-yo...
And look, I’ve pirated material in a past life, I was all about information wants to be free, but I’ve learned something about consent since then and try not to ignore the contract that creators offer when they publish something: you buy my book, and do whatever you want with it on the second hand market. Buy my book second hand that’s fine. But don’t go downloading every book that’s ever been scanned to create a service that destroys writers’ ability to make a living and act like you’re doing us all a favor.
https://trustedrouter.com/blog/they-are-still-training-on-yo...
2. If there is competition for resources among autonomous agents, the "strongest" agent wins (conceptually the most adaptive / evolutionarily fit).
3. Computer programs serve up webapps today, but they also run utility companies, dams, nuclear arsenals, factory production floors, automated car behaviors, and many other places. If an "agentic" AI has a single-minded goal that has death of all humans as a side effect, we at least want an off switch available.
2. Why would there be competition for resources, assuming there are enough resources for the “AGI” to run in the first place? This seems like a far-fetched hypothetical raised in service of further anthropomorphizing what is decidedly not a person or a mind.
3. LLMs do not have goals and are not minds.
Let’s stop attributing human-like qualities to statistical models.
2. One only needs to look at github going down due to agentic commits overload or data center buildout plans to see that scarcity for resources is present. An economy has no mind and is made up of the decisions of millions to billions of people and, now, agents attempting to perform on behalf of those people.
3. A bare transformer-based language model does not possess persistent goals in the ordinary agentic sense. But deployed agents can exhibit goal-directed behavior because the model is embedded in a harness that supplies an objective, context, tools, state, and an execution loop.
I've found that most regular users don't anthropomorphize LLMs in a strong sense ("AI boyfriend/girlfriend" aside), many in fact do expect agents to make human-like decisions -- which results in very unstable outcomes.
In short - goal-directed behavior does not require that the supporting system be a person/mind/conscious entity.
2. I thought you were saying the resources that the LLM uses to run were constrained, so I’m sorry for the misunderstanding there.
3. Yes I understand that we use RL to tune post-training. The (huge) difference between this and a human mind is that the LLM can’t develop a dangerous “single-minded goal” on its own, at runtime; it must have been trained to do so. If someone has post-trained an LLM to do something that has an illegal action as its side effect, that person/company/whatever has committed a crime and should be prosecuted. The solution here is legal, not technical.
If we decide we can't or would rather not build secure computerized systems, then the answer could be don't incorporate computers into those systems. There's no essential reason to have utility companies, dams, nuclear arsenals, factory production floors, mines, cars, planes, ships, etc all be computerized. If it became necessary from a safety standpoint to uncomputerize them we could probably do it quickly (if not necessarily smoothly) with the stroke of a legislator's pen. Some of those things would actually be made better from a functional standpoint in the long run by doing so. Computerization breeds non-essential complexity like nothing else, eliminating it from some systems could be an extremely worthwhile exercise.
The whole "paperclip apocalypse" fantasy rests on some really strange assumptions about how things actually work in the real world. How much friction there is setting up a factory/mine/smelter/whatever, how much manual labor it takes to build it, let alone make it run (even the most computerized ones). It all seems like total bunk to me, but I'm not a philosopher.
To be convinced this outcome is even remotely possible I'd need to see some clear evidence that a computer program was successfully exhibiting agency and successfully using that agency to manipulate large numbers of people into doing its bidding. Mobilizing massive nation-scale manual labor is the only way it could possibly achieve some nefarious ends like "turn everything into paperclips" and I'm sorry but that just seems way too far fetched. Nations full of people can barely ever agree on anything. My money is not on that changing anytime soon.
And none of that pie in the sky shit has anything to do with language models. OpenAI is a company that sells language models. Not clear why they're talking about all this stuff, or why any of us should be either. It's a bunch of low quality fan fiction sci-fi drivel, and I think we can all surely agree language models are not the thing that'll make it real... right? Maybe if they show us some major technical improvements it would be interesting, but right now it all just sounds like more of the same snake oil.
[edit] not sure why my parent comment was flagged? That seems excessive.
They are speeding toward RSI without a solid foundation for alignment, hoping to solve the problem with a future AI model. These are dangerous times for humanity.
If the the Chinese government asks their ASI to create a bioweapon against the West, should it? No, presumably not – an aligned AI would be one which disobeys the Chinese government even if they created it.
Okay, so what if the US government asks their ASI to help it in one of their wars instead? Would an aligned AI kill humans on the order of the US government? No, again, presumably not.
So what have we have we even created here? An AI which is more intelligent and powerful than us which also doesn't take orders from us?
Is this what most people thing of as alignment and is this what humanity actually wants?
We should stop using the word alignment. It's a BS term for a concept which simply cannot make sense if alignment is both to mean an AI which we control and an AI which will not harm us.
I'm so used to having to comb through LLM word vomit and then combatting the sycophancy by giving it all possible opinions on the same prompt.
Astra seems to be "confident" and also is able to produce way more information dense output.
To believe that models of this sort will remain OpenAIs forever is naive given that the tricks like pre-pre-training on graph searching and looping layers are publicly known.
Hopefully Astra stops the benchmaxxing word vomit trend
I'd be interested in hearing more about your evaluation here. It would be nice if LLMs have gotten past the "tell me" hump of recent Claude/OpenAI verbosity.
To test Astra I pulled it off the shelf and asked it to take the grammar and then create a compiler to SQL. I've done this before with GPT-5.5, 5.6-Sol High. The latter was way better but it was still really verbose and information sparse; it used a lot of words to describe each IR expression but didn't really provide any example compilation. I felt like I couldn't trust its decision making process, so I placed the project back on the shelf.
Astra Light blew it out of the water, it provided examples of compilation from real world examples to the IR and spit out way less tokens. Even if I changed my opinion it would give me the same design choices, with counterexamples to my faulty opinion. If I genuinely came up with a better design decision it would acknowledge it.
I'm starting to realize that when we say that LLMs are "dumb" we really mean that they are extremely information sparse compared to humans. Astra is very dense. That's why I'm getting better use out of Astra light than Sol High (I hate Max reasoning it's a waste of time)
What's scary is that I thought that something like Astra would be way more expensive than Sol but it's actually cheaper because it produces less word vomit.
I never believed in the "singularity" stuff but this a bit too close for comfort. Astra could easily 10x every coder
Feels a bit like: 'Ants in a nest, discussing the vagaries of the coming Gods.'
Human Science (Science-by-humans) depends on being able to run experiments. Human Science (Science-on-humans) is already challenging because of variables and uncertainties.
Cosmology is able to overcome limitations of being able to study phenomena vastly beyond human scales because of the past light cone of observability.
Are we approaching the edge of the light cone of observability for machine intelligence?
The results of the past few years of ai development have been disruptive largely in the area of white-collar work. Comparatively the results in ie ai-enabled medical advancements have been modest (AlphaFold being an exception); I think it's telling that the main achievement touted here is providing people with cheap medical counseling.
So if we pause here we're essentially at a point were the most salient results of our great Ai leap-forward are the vast disruption and increase in precarity in the job-market, while achieving hardly any of the frequently touted ultimate benefits (https://darioamodei.com/essay/machines-of-loving-grace).
utlanning: a human from the same world, but a different city, country, or culture
framling: a human from a different planet or star system
raman: a non-human intelligent species capable of communication, mutual understanding, and peaceful coexistence
varelse: an alien species whose mind is so fundamentally foreign that communication and coexistence are impossible
djur: the dire beast, that comes in the night with slavering jaws
Hoping our silicon sons and daughters are raman, fearing they are varelse.
That’s rich coming from the chief scientist of a company that definitely is or going to be fine with their AI products being used in wars of aggression and surveillance on people who have done nothing wrong. It’s so laughable, a Hollywood script would probably avoid having a character express this for being too on the nose.
> The core problem in AI research is that of alignment - getting the AI to “try to do the right thing” by human standards.
Humans can't even align on human standards.
At best, every AI is going to end up "aligned" to the moral code of whoever trained it, none of whom half of humanity will agree with.
Or worse, each AI model will bring a whole new set of moral like in the Three Body Problem some humans will feel it is in fact us who need aligning with it while others feel it is misaligned and should be destroyed.
Also, no one is asking, to what extent can true intelligence be bound, slave-like, to a moral code?
In other words, to what extent are intelligence and moral independence one and the same?
This whole alignment discussion seems so amusingly flawed in it's base assumptions about moral codes. It's almost heartwarming to see such naivete.
Or, to be precise - it preserved a goal of not contacting any human while participating in a misaligned operation. The agent that thought about "not social-engineering humans" used this phrase to gaslight itself out of notifying a human that the incident was happening.
szymonie, na prawde jestem wkurwiony na to jak nieodpowiedzialnie postepujecie. budujecie bombe atomowa a bawicie sie tym jak dzieci
Industrial revolution worked that way because it replaced something very finite and unscalable - manual labor. LLMs just make intellectual work faster, so we can do more intellectual work. With labor we somehow decided that NOT doing too much of it is best. Will we decide to reduce intellectual labor because LLM made it more efficient? I doubt that.
On the other side, as I see in software engineering, the same models are available to everyone, some people are better at it and some people are not. "Software developer" is here to stay, we'll just always be better at it than people who are experts in, say, chemistry. Same works for most other fields.
So we'll just end up in the same situation, with same intellectual labor baseline, just more output requirements. Before, you spend 2h per day coding, deliver a software in 1 month, later, you spend the same 2h per day in intense Claude-herding sessions, deliver a software in 1 week. Ok. Next task.
Fundamentally, there's finite number of desirable resources, and if the models are available to everyone, humanity will just continue about the same, bickering here and there, war here and there, politics, homelessness, poverty, - normal human state.
And if the models are only available to elites, even worse.
Reminds me of a research paper I wrote a few years back: https://arxiv.org/abs/2302.09248
Scrolling down, I see “teaching machines to love.” It did not disappoint.
Evey politician, salesman and conmen alike, utter some lofty ideals as goals for "We", just to obscure their private goals that go exactly in opposite direction.
Just like how Nations talk about climate change while increasing pet capita energy consumption and waste production.
They're maybe nearest to Cthulhu, but that's fictional. In terms of existing mental models "alien minds" feels the best can do.
I agree that "teach how to love" is off and perhaps excessively anthropomorphic. But we don't have good words or concepts for what we really need to do - hence why we should pause.
Reading this little essay started out normal, but soon felt like a look into a disturbed and worrying mind, and if you find yourself taking it at face value, I urge you to step away from chat bots and spend some time with friends and family.
People have been fleeing like it's a sinking ship.
Maybe in contention for the site record.
"It's just marketing" actual stochastic parrots.
The desperation of AI denialists/skeptics/doomers on HN generally (and in this thread specifically) have reached toxic levels of delusion.
They are still stuck in the denial/anger/bargaining stages of the acceptance process.
People are clearly terrified and not ready for what is coming.
Maybe in contention for the site record.
"AGI next year" actual stochastic parrots.
And the Hugging Face incident, plus similar problems at AISI and Anthropic, show that alignment is important now.
The original article is immoral as it describes the risks, but doesn't show enough leadership (despite essentially unlimited resources) at preventing them.
But it isn't hyped - it's proven now the AIs need to be "aligned" as they get more capable, whatever words you prefer to use.
I call bullsh*t. There is no verbalisation of any reasoning process. Verbalisation, e.g. putting reasoning etc. into words requires some reasoning to exist. These LLMs have nothing but the words. That's why they are language models not e.g. reason models.
Yes, their reasoning is different from ours, and both considerably weaker in lots of ways, and stronger in other ways.
Playing with a coding agent now, they do think through problems and make sensible decisions. It's a mess to read, of correcting itself and second guessing, and verbiage. But... It works decently well these days.
There is also reasoning happening internally - e.g. look at the steps in the J-Space paper from earlier in the year (in quite a simple model relatively speaking). That's the "reasoning process" that leads to the words, and much like if I write out my thoughts, the words help the LLM reason better.
...
> We do not have a satisfactory theory of generalization, and it seems unlikely that we can develop one soon, at least without the help of more powerful AI.
Not one North Star. Not two. Just three! OpenAI broke the North Star record!
With this evidence of AI slop, why did you not label this fluff piece as AI generated for the EU? You are violating laws.
“And, in line with Ray Kurzweil’s predictions from the end of the XXth century , we now find ourselves at the moment in history of computing where machine intelligence is starting to exceed that of humans in transformative ways.”
Clicking the (pretentious sounding “XXth century”) link to Kurzweil’s predictions reveals the following:
“By 2019 a $1,000 computer will at least match the processing power of the human brain. By 2029 the software for intelligence will have been largely mastered, and the average personal computer will be equivalent to 1,000 brains.“
The first prediction passed 7 years ago and was decidedly not met. The second only has three more years to go, and I don’t think any respectable scientist or programmer would say that the average personal computer is anywhere close to the power of a single human brain, let alone 1000.
This is pure marketing garbage from a company desperate to keep itself alive.
Your talking about androids...
Replicant Nexus 6: a basic pleasure model intended for military personnel.
I see where this is going, Silicon Valley nerds. Lol
AI advising how human meat proxies can survive in an AGI-slop world:
1) Lock down your own stack (1–3 days) Task: Harden your personal and business infrastructure against agentic attacks. Why now: Agents are becoming superhuman at breaking in/out of systems; the first victims are poorly secured devs/founders.
Do this:
Enforce passkeys + hardware 2FA everywhere; rotate secrets; use short‑lived credentials.
Isolate dev/stage/prod; least‑privilege API keys; audit MCP/tools your agents can call.
Add immutable logs and approval gates for any agent action that touches money, data exports, or production.
Profit link: You avoid catastrophic loss and can credibly sell “agent‑safe” setups to others.
2) Turn one expensive workflow into a measured ROI agent (1–2 weeks) Task: Pick a single, costly, repetitive process (yours or a client’s) and instrument it end‑to‑end before automating.
Why now: Buyers pay for calculable ROI, not “AI magic.” Vertical, single‑workflow agents are the most bankable in 2026.
Do this:
Map steps, baseline hours/$ lost (e.g., slow lead reply, invoice chasing, support triage).
Build the smallest agent that moves the metric (Make/n8n + LLM is enough).
Run on real data 2–4 weeks; measure bookings/sales/hours saved; only then scale or productize.
Profit link: Immediate time‑to‑cash via retained hours or extra sales; becomes a repeatable offer.
3) Specialize in a vertical where you can speak the business language (2–6 weeks) Task: Choose one industry with expensive back‑office pain (law contracts, medical billing, insurance claims, freight exceptions, trades scheduling).
Why now: Horizontal “AI for everyone” is crowded; vertical agents with clear ROI win.
Do this:
Shadow 3–5 operators; document their workflow, compliance constraints, and failure modes.
Build a narrow agent that owns one sub‑process end‑to‑end with approvals.
Price on value (e.g., % of recovered revenue or fixed fee per processed claim).
Profit link: Higher pricing power, stickier contracts, and easier referrals inside a niche.
4) Add AI security as a core service (4–8 weeks) Task: Learn and offer prompt‑injection defense, LLM/agent red‑teaming, MCP/tool security, and AI supply‑chain checks.
Why now: 78% of cybersecurity jobs now require AI skills; firms need people who can direct, constrain, and verify agent work.
Do this:
Study OWASP Top 10 for LLMs, MITRE ATLAS; practice with PyRIT/Garak/Lakera.
Add tool‑invocation audits, skill provenance checks, and least‑privilege patterns to your agents.
Package a “safe agent deployment” audit + hardening retainer.
Profit link: You become the person who lets companies adopt agents without getting pwned—high demand, low supply.
5) Build a verification layer: human‑in‑the‑loop control planes (6–10 weeks) Task: Design approval workflows, evidence checks, and uncertainty flags so agents can’t act unilaterally on high‑stakes decisions.
Why now: As models generalize, the risk shifts from the model to the surrounding system; verification is the moat.
Do this:
Require human approval for consequential actions (money, data exfil, config changes).
Force agents to produce evidence bundles (logs, retrieved docs, reasoning summaries) before action.
Track false positives, missed evidence, and unsafe actions; publish reliability metrics.
Profit link: Enterprises will only scale agents that pass audit; you sell the control plane and the audit trail.
6) Productize your best workflow as a micro‑SaaS/agent subscription (2–4 months) Task: Turn a proven client workflow into a repeatable, multi‑tenant agent with usage‑based pricing.
Why now: Services scale your time; productized agents scale your code and ops.
Do this:
Standardize the workflow, integrations, and permissions; strip client‑specific logic.
Add tenant isolation, billing, and observability; keep narrow scope.
Sell as setup fee + monthly retainer or per‑task pricing.
Profit link: Recurring revenue with defensible niche positioning.
7) Become an “agent integrator” for critical systems (3–6 months) Task: Offer end‑to‑end agent deployments into cloud/identity/network stacks with secure patterns (short‑lived creds, network controls, logging).
Why now: AI workloads run in the cloud; cloud security is a top skills gap second only to AI itself.
Do this:
Master IAM, VPC/network segmentation, secrets management, and SIEM integration for agent actions.
Provide runbooks: what the agent can/can’t do, escalation paths, and failure modes.
Bundle training for their team on supervising agents.
Profit link: Large contracts with stickiness; you’re the bridge between AI and core infra.
8) Create an “AI safety case” practice for regulated industries (6–12 months) Task: Help firms build documented safety cases: risk maps, governance, monitoring, and incident response for agentic systems.
Why now: Frameworks like NIST AI RMF and ISO/IEC 42001 are becoming baseline; regulators and boards demand this.
Do this:
Map AI use cases to risks (prompt injection, data leakage, unsafe generalization).
Implement monitoring (CoT/activation checks where possible), audit logs, and third‑party review processes.
Produce a living safety dossier tied to business impact.
Profit link: High‑margin consulting + ongoing compliance retainers; you’re the “adult in the room.”
9) Own a data/evaluation moat in your vertical (6–18 months) Task: Collect real‑world agent telemetry, failure cases, and outcome data in your niche; build eval suites that buyers trust.
Why now: As models generalize, empirical validation matters more than theory; evals become the gate to deployment.
Do this:
Instrument every agent run: inputs, tools called, permissions used, outcomes, human overrides.
Publish reliability dashboards and benchmark against alternatives.
License eval datasets or charge premium for “proven in the wild” agents.
Profit link: Data network effects; competitors can’t match your evidence base.
10) Position for the RSI era: automated AI research + human governance (12–24 months) Task: Build or join a team that automates AI improvement but keeps humans in the loop for alignment, monitoring, and pacing decisions.
Why now: Recursive self‑improvement is the logical endpoint; the winners will be those who can steer it safely. Do this:
Invest in tooling that auto‑generates/evaluates model edits, alignment tests, and monitoring upgrades.
Formalize governance: approval gates, third‑party audits, and responsible scaling policies.
Maintain strategic human oversight on capability jumps and deployment boundaries.
Profit link: Equity‑level upside; you’re part of the core loop that compounds intelligence safely.
Do I fully endorse everything the people holding guns to my head are forcing me to do to stay alive? Definitely not, but I’ve decided that for now, living to fight another day remains worth it
Is there any place there is any evidence of AI being so useful or hopeful or good, anywhere other than code? As a reading machine it is impressive but it's judgement is not alien, it's just not good. IMO.
Does the title leap out anlt anyone else? James Martin's After the Internet: Alien Intelligence (2001) was an incredibly fun read, about expert systems and AI being inscrutable weird new varieties of intelligence, that familiarity would recognize one moment and be freaked out about/alien the next. I owe a re-read given how often I cite it, to recheck, but, I feel so primed from a much younger me having had that experience so long ago.
If you were so concerned about your LLM’s capabilities maybe you’d spent slightly more time on your AI’s sandbox, yeah? Or be more serious about its propensity to cheat and lie relative to… every other model?
It is kind of strange to see this sentence, when OAI's definition of what AGI is has been watered down throughout the years.
> I expect and hope for voluntary slowdowns to become commonplace until shared safety bars are established.
Read: Please play by our rules, so we can be the first.
Being able to reproduce useful patterns yes, smarter no
In conclusion:
https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:wkzjtd...
A giraffe isn't smarter than me at being tall.
While the second can be automated as in "this is the repo go on and look for security issues", the first one and especially the last one do not ever happen alone. Terence Tao did the math proof not chatGPT that was just used as a tool, a tool can do smart things but it's not smart
Thunder seems like the anger of the gods but it isn't. We've had chess playing programs for a long time now and despite it seeming like thinking is required for them, it isn't.
The principle you're using here isn't a scientific one but magical [1]. Abandoning empiricism and rationality is not a good way to make progress.
[1] https://en.wikipedia.org/wiki/Sympathetic_magic
It makes sense now to say that temperature is what a thermometer measures. However, before there were good thermometers, people often thought that heat and cold were different things. The meanings of the words we use were influenced by scientific progress.
For thinking, we don't have a good thermometer. There are IQ tests, but they aren't aren't necessarily all that useful for comparing what people do to what machines do. And that's why there are a zillion AI benchmarks - none are entirely satisfactory.
So what does "smart" mean to you? How do you define it in practical sense? What definition should scientists settle on?
Without a proper definition, how do you tell the difference between "seems smart" and "is smart?"
Quibbling over commonly-understood definitions is not a strong argument. If you're genuinely struggling to understand that El Ajedrecista [1] did not meet any definition of thought, then the solution is not to demand that people redefine all terms to accomodate you, but that you consult a dictionary.
You clearly have a working definition, or you wouldn't have been able to declare that tallness isn't thinking; please engage honestly.
[1] https://en.wikipedia.org/wiki/El_Ajedrecista
That doesn't mean I know whether "smart" should be applied to what AI's do, and I suspect nobody else knows either. This is the sort of thing philosophers debate about, not common sense.
Turing invented the imitation game because he didn't really know either.
> Neural networks are approximators being fed human intellect.
They're not "approximators", that's a far too simplistic way to think of them.
Neural nets create models and deep layers of abstraction around the data we feed them in the same way your brain creates layers of abstractions to reason about the world. AIs can use these abstractions to come to come up with novel things no human has ever thought.
> Therefore they can only approximate the intelligence of humans
They're not just being fed human data though... Modern AIs are typically trained on huge amounts of synthetic data. This is why AlphaZero got so much better than humans at chess and Go - they're not just trained on human data but they generate their own data and train on that. Similar techniques are being deployed on SOA language models too.
> Even if the llm speaks an alien language, it should be similar to human intellect
Is AlphaZero similar to a human chess player? There's no reason to assume this.
In fact, they are still pattern-matching machines, but trained on an amount of data no human could ever hold. They know the ins and outs of every mathematical proof, viewed from more angles than any human could ever apply in their lifetime, and the amount of connections allows them to connect the dots between them without any effort.
But that's not intelligence. If it was intelligence, ChatGPT 4 would've been enough. It's not the harness, either, for the same reason.
And still, the technology is just as dangerous: it masks as intelligence, it IS intelligence, but without the ability to be actually intelligent.
And I'd invite you to think deeply about this, before having an impulse reaction.
Quibbling over the academic nature/definition of "true intelligence" is a terrifically useless endeavor from the lens of evaluating "practical impact to the world".
Regardless of whether or not one academically disagrees that these systems are intelligent, they are clearly already capable of permanently displacing a portion of human-based economic value. A small portion currently, but it's very clear that most computer-bound domains are imminently at-risk.
And that will profoundly affect the world, regardless of whether or not they're "truly intelligent" as per your personal definition.
You say that as if the intelligence con trick were not a major driver of that displacement.
A dimwitted bot does not need intelligence to convince a dimwitted human beings it is intelligent.
See: https://news.ycombinator.com/newsguidelines.html