Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI 2027

ai-2027.com

Post reply on HN
Print view View original

451–460 of 641 posts

Re: AI 2027

#451

Post by nikisil80 »

visarga wrote: post #431

The story is entertaining, but it has a big fallacy - progress is not a function of compute or model size alone. This kind of mistake is almost magical thinking. What matters most is the training set. During the GPT-3 era there was plenty of organic text to scale into, and compute seemed to be the bottleneck. But we quickly exhausted it, and now we try other ideas - synthetic reasoning chains, or just plain synthetic…

Best reply in this entire thread, and I align with your thinking entirely. I also absolutely hate this idea amongst tech-oriented communities that because an AI can do some algebra and program an 8-bit video game quickly and without any mistakes, it's already overtaking humanity. Extrapolating from that idea to some future version of these models, they may be capable of solving grad school level physics problems and programming entire AAA video games, but again - that's not what _humanity_ is about. There is so much more to being human than fucking programming and science (and I'm saying this as an actual nuclear physicist). And so, just like you said, the AI arm's race is about getting it good at _known_ science/engineering, fields in which 'correctness' is very easy to validate. But most of human interaction exists in a grey zone.

Thanks for this.

Re: AI 2027

#452

Post by loganmhb »

vonneumannstan wrote: post #428
fmap wrote: post #366

Earlier quoted context omitted.

> even if this doesn’t lead to AGI, at the very least it’s likely the final “warning shot” we’ll get before it’s suddenly and irreversibly here. I agree that it's good science fiction, but this is still taking it too seriously. All of these "projections" are generalizing from fictional evidence - to borrow a term that's popular in communities that push these ideas. Long before we had deep learning there were people l…

>All of these "projections" are generalizing from fictional evidence - to borrow a term that's popular in communities that push these ideas. This just isn't correct. Daniel and others on the team are experienced world class forecasters. Daniel wrote another version of this in 2021 predicting the AI world in 2026 and was astonishingly accurate. This deserves credence. https://www.lesswrong.com/posts/6Xgy6CAf2jqHhynHL/…

I respect the forecasting abilities of the people involved, but I have seen that report described as "astonishingly accurate" a few times and I'm not sure that's true. The narrative format lends itself somewhat to generous interpretation and it's directionally correct in a way that is reasonably impressive from 2021 (e.g. the diplomacy prediction, the prediction that compute costs could be dramatically reduced, some things gesturing towards reasoning/chain of thought) but many of the concrete predictions don't seem correct to me at all, and in general I'm not sure it captured the spiky nature of LLM competence.

I'm also struck by the extent to which the first series from 2021-2026 feels like a linear extrapolation while the second one feels like an exponential one, and I don't see an obvious justification for this.

Re: AI 2027

#453

Post by nthingtohide »

lo_zamoyski wrote: post #421
zdragnar wrote: post #293

Earlier quoted context omitted.

> What's the point of our existence if we have no way to meaningfully contribute to our own world? You may find this to be insightful: https://meltingasphalt.com/a-nihilists-guide-to-meaning/ In short, "meaning" is a contextual perception, not a discrete quality, though the author suggests it can be quantified based on the number of contextual connections to other things with meaning. The more densely connected somet…

People talk about meaning, but they rarely define it. Ultimately, "meaning" is a matter of "purpose", and purpose is a matter of having an end, or telos . The end of a thing is dependent on the nature of a thing. Thus, the telos of an oak tree is different from the telos of a squirrel which is different from that of a human being. The telos or end of a thing is a marker of the thing's fulfillment or actualization as…

Meaning is a matter of context. Most of the context resides in the past and future. Ludwig's claim that word's meaning is dependent on how it is used. This applies generally.

Daniel Dennett - Information & Artificial Intelligence

https://www.youtube.com/watch?v=arEvPIhOLyQ

Daniel Dennett bridges the gap between everyday information and Shannon-Weaver information theory by rejecting propositions as idealized meaning units. This fixation on propositions has trapped philosophers in unresolved debates for decades. Instead, Dennett proposes starting with simple biological cases—bacteria responding to gradients—and recognizing that meaning emerges from differences that affect well-being. Human linguistic meaning, while powerful, is merely a specialized case. Neural states can have elaborate meanings without being expressible in sentences. This connects to AI evolution: "good old-fashioned AI" relied on propositional logic but hit limitations, while newer approaches like deep learning extract patterns without explicit meaning representation. Information exists as "differences that make a difference"—physical variations that create correlations and further differences. This framework unifies information from biological responses to human consciousness without requiring translation into canonical propositions.

Re: AI 2027

#454

Post by maerF0x0 »

> OpenBrain reassures the government that the model has been “aligned” so that it will refuse to comply with malicious requests

Of course the real issue being that Governments have routinely demanded that 1) Those capabilities be developed for government monopolistic use, and 2) The ones who do not lose the capability (geo political power) to defend themselves from those who do.

Using a US-Centric mindset... I'm not sure what to think about the US not developing AI hackers, AI bioweapons development, or AI powered weapons (like maybe drone swarms or something), if one presumes that China is, or Iran is, etc then whats the US to do in response?

I'm just musing here and very much open to political science informed folks who might know (or know of leads) as to what kinds of actual solutions exist to arms races. My (admittedly poor), understanding of the cold war wasn't so much that the US won, but that the Soviets ran out of steam.

Re: AI 2027

#455

Post by maerF0x0 »

tim333 wrote: post #395
fmap wrote: post #366

Earlier quoted context omitted.

> even if this doesn’t lead to AGI, at the very least it’s likely the final “warning shot” we’ll get before it’s suddenly and irreversibly here. I agree that it's good science fiction, but this is still taking it too seriously. All of these "projections" are generalizing from fictional evidence - to borrow a term that's popular in communities that push these ideas. Long before we had deep learning there were people l…

>There is no evidence or argument for exponential growth I think the growth you are thinking of, self improving AI, needs the AI to be as smart as a human developer/researcher to get going and we haven't got there yet. But we quite likely will at some point.

and the article specifically mentions the fictional company (clearly designed to generalize the Google/OpenAI's of the world) are supposedly (according to the article) working on building that capability. First by augmenting human researchers, later by augmenting itself.

Re: AI 2027

#456

Post by lblume »

throwup238 wrote: post #262
ivraatiems wrote: post #230

Earlier quoted context omitted.

I think there is a good chance you are roughly right. I also think that the "secret sauce" of sapience is probably not something that can be replicated easily with the technology we have now, like LLMs. They're missing contextual awareness and processing which is absolutely necessary for real reasoning. But even so, solving that problem feels much more attainable than it used to be.

I think the missing secret sauce is an equivalent to neuroplasticity. Human brains are constantly being rewired and optimized at every level: synapses and their channels undergo long term potentiation and depression, new connections are formed and useless ones pruned, and the whole system can sometimes remap functions to different parts of the brain when another suffers catastrophic damage. I don’t know enough about…

Transformers already are very flexible. We know that we can basically strip blocks at will, reorder modules, transform their input in predictable ways, obstruct some features and they will after a very short period of re-training get back to basically the same capabilities they had before. Fascinating stuff.

Re: AI 2027

#457

Post by somebodythere »

andrepd wrote: post #399
stego-tech wrote: post #279

It’s good science fiction, I’ll give it that. I think getting lost in the weeds over technicalities ignores the crux of the narrative: even if this doesn’t lead to AGI, at the very least it’s likely the final “warning shot” we’ll get before it’s suddenly and irreversibly here. The problems it raises - alignment, geopolitics, lack of societal safeguards - are all real, and happening now (just replace “AGI” with “corpo…

You said it right, science fiction. Honestly is exactly the tenor I would expect from the AI hype: this text is completely bereft of any rigour while being dressed up in scientific language. There's no evidence, nothing to support their conclusions, no explanation based on data or facts or supporting evidence. It's purely vibes based. Their promise is unironically "the CEOs of AI companies say AGI is 3 years away" !…

Did you see the supplemental material that explains how they arrived at their timelines/capabilities forecasts? https://ai-2027.com/research

Re: AI 2027

#458

Post by Aldipower »

No one can predict the future. Really, no one. Sometimes there is a hit, sure, but mostly it is a miss.

The other thing is in their introduction: "superhuman AI" _artificial_ intelligence is always, by definition, different from _natural_ intelligence. That they've chosen the word "superhuman" shows me that they are mixing the things up.

Re: AI 2027

#459

Post by bko »

boh wrote: post #423
bko wrote: post #404

Earlier quoted context omitted.

> The problems it raises - alignment, geopolitics, lack of societal safeguards - are all real, and happening now (just replace “AGI” with “corporations”, and voila, you have a story about the climate crisis and regulatory capture). Can you point to the data that suggests these evil corporations are ruining the planet? Carbon emissions are down in every western country since 1990s. Not down per-capita, but down in abs…

Thanks for letting us know everything is fine, just in case we get confused and think the opposite.

You're welcome. I know too many upper middle class educated people that don't want to have kids because they believe the earth will cease to be inhabitable in the next 10 years. It's really bizarre to see and they'll almost certainly regret it when they wake up one day alone in a nursing home, look around and realize that the world still exists.

And I think the neuroticism around this topic has led young people into some really dark places (anti-depressants, neurotic anti social behavior, general nihilism). So I think it's important to fight misinformation about end of world doomsday scenarios with both facts and common sense.

Re: AI 2027

#460

Post by tomp »

visarga wrote: post #431

The story is entertaining, but it has a big fallacy - progress is not a function of compute or model size alone. This kind of mistake is almost magical thinking. What matters most is the training set. During the GPT-3 era there was plenty of organic text to scale into, and compute seemed to be the bottleneck. But we quickly exhausted it, and now we try other ideas - synthetic reasoning chains, or just plain synthetic…

Did we read the same article?

They clearly mention, take into account and extrapolate this; LLM have first scaled via data, now it's test time compute, but recent developments (R1) clearly show this is not exhausted yet (i.e. RL on synthetically (in-silico) generated CoT) which implies scaling with compute. The authors then outline further potential (research) developments that could continue this dynamic, literally things that have already been discovered just not yet incorporated into edge models.

Real-world data confirms their thesis - there have been a lot of sceptics about AI scaling, somewhat justified ("whoom" a.k.a. fast take-off hasn't happened - yet) but their fundamental thesis has been wrong - "real-world data has been exhausted, next algorithmic breakthroughs will be hard and unpredictable". The reality is, while data has been exhausted, incremental research efforts have resulted in better and better models (o1, r1, o3, and now Gemini 2.5 which is a huge jump! [1]). This is similar to how Moore's Law works - it's not given that CPUs get better exponentially, it still requires effort, maybe with diminishing returns, but nevertheless the law works...

If we ever get to models be able to usefully contribute to research, either on the implementation side, or on research ideas side (which they CANNOT yet, at least Gemini 2.5 Pro (public SOTA), unless my prompting is REALLY bad), it's about to get super-exponential.

Edit: then once you get to actual general intelligence (let alone super-intelligence) the real-world impact will quickly follow.

Post reply on HN