And in 30 years it will be another 30 years away.

LLMs are so incredibly useful and powerful but they will NEVER be AGI. I actually wonder if the success of (and subsequent obsession with) LLMs is putting true AGI further out of reach. All that these AI companies see are the $$$. When the biggest "AI Research Labs" like OpenAI shifted to product-izing their LLM offerings I think the writing was on the wall that they don't actually care about finding AGI.

People over-estimate the short term and under-estimate the long term.

Will LLMs approach something that appears to be AGI? Maybe. Probably. They're already "better" than humans in many use cases.

LLMs/GPTs are essentially "just" statistical models. At this point the argument becomes more about philosophy than science. What is "intelligence?"

If an LLM can do something truly novel with no human prompting, with no directive other than something it has created for itself - then I guess we can call that intelligence.

What the hell is general intelligence anyway? People seem to think it means human-like intelligence, but I can't imagine we have any good reason to believe that our kinds of intelligence constitute all possible kinds of intelligence--which, from the words, must be what "general" intelligence means.

It seems like even if it's possible to achieve GI, artificial or otherwise, you'd never be able to know for sure that thats what you've done. It's not exactly "useful benchmark" material.

They‘ll get cheaper and less hardware demanding but the quality improvements get smaller and smaller, sometimes hardly noticeable outside benchmarks

What was the point of this comment? It's confrontational and doesn't add anything to the conversation. If you disagree, you could have just said that, or not commented at all.

The people who go around saying "LLMs aren't intelligent" while refusing to define exactly what they mean by intelligence (and hence not making a meaningful/testable claim) add nothing to the conversation.

Doesn't even matter. The capabilities of the AI that's out NOW will take a decade or more to digest.

Not even close. Software can now understand human language... this is going to mean computers can be a lot more places than they ever could. Furthermore, software can now understand the content of images... eventually this will have a wild impact on nearly everything.

I am tech founder, who spends most of my day in my own startup deploying LLM-based tools into my own operations, and I'm maybe 1% of the way through the roadmap I'd like to build with what exists and is possible to do today.

100% this. The rearrangement of internal operations has only started and there is just sooo much to do.

To push this metaphor, I'm very curious to see what happens as new organic training material becomes increasingly rare, and AI is fed nothing but its own excrement. What happens as hallucinations become actual training data? Will Google start citing sources for their AI overviews that were in turn AI-generated? Is this already happening?

I figure this problem is why the billionaires are chasing social media dominance, but even on social media I don't know how they'll differentiate organic content from AI content.

maybe silicon valley and the world move at basically different rates

idk AI is just a speck outside of the HN and SV info-bubbles

still early to mass adoption like the smartphone or the internet, mostly nerds playing w it

I really disagree. I had a masseuse tell me how he uses ChatGPT, told it a ton of info about himself, and now he uses it for personalized nutrition recommendations. I was in Atlanta over the weekend recently, at a random brunch spot, and overheard some _very_ not SV/tech folks talk about how they use it everyday. Their user growth rate shows this -- you don't hit hundreds of millions of people and have them all be HN/SV info-bubble folks.

I see ChatGPT as the new Google, not the new Nuclear Power Soruce. maybe im naive

ChatGPT has 400M weekly users. https://backlinko.com/chatgpt-stats

have you wondered how many of these are bots leveraging free chatgpt with proxied vpn IPs?

I'm a ChatGPT paying user but I know no one who's not a developer on my personal circles who also is one.

maybe im an exeception

edit: I guess 400M global users being the US 300M citizens isn't out of scope for such a highly used product amongst a 7B population

But social media like instagram or fb feels like had network effects going for them making their growth faster

and thus maybe why openai is exploring that idea idk

Pretty much everyone in high school or college is using them. Also everyone whose job is to produce some kind of content or data analysis. That's already a lot of people.

That doesn’t match what I hear from teachers, academics, or the librarians complaining that they are regularly getting requests for things which don’t exist. Everyone I know who’s been hiring has mentioned spammy applications with telltale LLM droppings, too.

I can see how students would be first users of this kinda of tech but am not on those spheres, but I believe you.

As per spammy applications, hasn't always been this the case and now made worse due to the cheapness of -generating- plausible data?

I think ghost-applicants where existent already before AI where consultant companies would pool people to try and get a position on a high paying job and just do consultancy/outsourcing things underneath, many such cases before the advent of AI.

AI just accelerates no?

> idk AI is just a speck outside of the HN and SV info-bubbles

> still early to mass adoption like the smartphone or the internet, mostly nerds playing w it

Rather: outside of the HN and SV bubbles, the A"I"s and the fact how one can fall for this kind of hype and dupery is commonly ridiculed.

This is accurate, doubly so for the people who treat it like a religion and fear the coming of their machine god. This, when what we actually have are (admittedly sometimes impressive) next-token predictors that you MUST double-check because they routinely hallucinate.

Then again I remember when people here were convinced that crypto was going to change the world, democratize money, end fiat currency, and that was just the start! Programs of enormous complexity and freedom would run on the blockchain, games and hell even societies would be built on the chain.

A lot of people here are easily blinded by promises of big money coming their way, and there's money in loudly falling for successive hype storms.

Yeah, I'm old enough to remember all the masses who mocked the Internet and smartphones too.

Im not mocking AI, and while the internet and smartphones fundamentally changed how societies operate, and AI will probably do so to, why the Doomerism? Isn't that how tech works? We invent new tech and use it and so on?

What makes AI fundamentally different than smartphones or the internet? Will it change the world? Probably, already has.

Will it end it as we know it? Probably not?

Is AGI even important? I believe the next 10 to 15 years will be Assisted Intelligence. There are things that current LLM are so poor I dont believe a 100x increase in pref / watt is going to make much difference. But it is going to be good enough there wont be an AI Winter. Since current AI has already reached escape velocity and actually increase productivity in many areas.

The most intriguing part is if Humanoid factory worker programming will be made 1000 to 10,000x more cost effective with LLM. Effectively ending all human production. I know this is a sensitive topic but I dont think we are far off. And I often wonder if this is what the current administration has in sight. ( Likely Not )

AI winter is relative, and it's more about outlook and point of view than actual state of the field.

I'll take the "under" on 30 years. Demis Hassabis (who has more credibility than whoever these 3 people are combined) says 5-10 years: https://time.com/7277608/demis-hassabis-interview-time100-20...

1. LLM interactions can feel real. Projections and psychological mirroring is very real.

2. I believe that AI researchers will require some level of embodiment to demonstrate:

a. ability to understand the physical world.

b. make changes to the physical world.

c. predict the outcome to changes in the physical world.

d. learn from the success or failure of those predictions and update their internal model of the external world.

---

I cannot quickly find proposed tests in this discussion.

I just used o3 to design a distributed scheduler that scales to 1M+ sxchedules a day. It was perfect, and did better than two weeks of thought around the best way to build this.

You just asked it to design or implement?

If o3 can design it, that means it’s using open source schedulers as reference. Did you think about opening up a few open source projects to see how they were doing things in those two weeks you were designing?

yeah unless you have very specific requirements I think the baseline here is not building/designing it yourself but setting up an off-the-shelf commercial or OSS solution, which I doubt would take two weeks...

Dunno, in work we wanted to implement a task runner that we could use to periodically queue tasks through a web UI - it would then spin up resources on AWS and track the progress and archive the results.

We looked at the existing solutions, and concluded that customizing them to meet all our requirements would be a giant effort.

Meanwhile I fed the requirement doc into Claude Sonnet, and with about 3 days of prompting and debugging we had a bespoke solution that did exactly what we needed.

While impressive, I'm not convinced that improved performance on tasks of this nature are indicative of progress toward AGI. Building a scheduler is a well studied problem space. Something like the ARC benchmark is much more indicative of progress toward true AGI, but probably still insufficient.

Designing a distributed scheduler is a solved problem, of course an LLM was able to spit out a solution.

Can someone throw some light on this Dwarkesh character? He landed a Zucc podcast pretty early on... how connected is he? Is he an industry plant?

He's awesome.

I listened to Lex Friedman for a long time, and there was a lot of critiques of him (Lex) as an interviewer, but since the guests were amazing, I never really cared.

But after listening to Dwarkesh, my eyes are opened (or maybe my soul). It doesn't matter I've heard of not-many of his guests, because he knows exactly the right questions to ask. He seems to have genuine curiosity for what the guest is saying, and will push back if something doesn't make sense to him. Very much recommend.

AGI is here today... go have a kid.

Not artificial, but yes, it's unclear what advantage an artificial person has over a natural one, or how it's supposed to gain special insights into fusion reactor design and etc. even if it can think very fast.

That would be "GI". The "A" part implies, specifically, NOT having a kid, eh?

Natural intelligence is too expensive. Takes too long for it to grow. If things go wrong then we have to jail it. With computers we just change the software.

”‘AGI is x years away’ is a proposition that is both true and false at the same time. Like all such propositions, it is therefore meaningless.”

The new fusion power

That's 20 years away.

It was also 20 years away 30 years ago.

Would we even recognise it if it arrived? We'd recognise human level intelligence, probably, but that's specialised. What would general intelligence even look like.

If/when we will have AGI, we will likely have something fundamentally superhuman very soon after, and that will be very recognizable.

This is the idea of "hard takeoff" -- because the way we can scale computation, there will only ever be a very short time when the AI will be roughly human-level. Even if there are no fundamental breakthroughs, the very least silicon can be ran much faster than meat, and instead of compensating narrower width execution speed like current AI systems do (no AI datacenter is even close to the width of a human brain), you can just spend the money to make your AI system 2x wider and run it at 2x the speed. What would a good engineer (or, a good team of engineers) be able to accomplish if they could have 10 times the workdays in a week that everyone else has?

This is often conflated with the idea that AGI is very imminent. I don't think we are particularly close to that yet. But I do think that if we ever get there, things will get very weird very quickly.

There's a test for this: https://arcprize.org/arc-agi

Basically a captcha. If there's something that humans can easily do that a machine cannot, full AGI has not been achieved.

We sort of are able to recognize Nobel-worthy breakthroughs

One of the many definitions I have for AGI is being able to create the proofs for the 2030, 2050, 2100, etc Nobel Prizes, today

A sillier one I like is that AGI would output a correct proof that P ≠ NP on day 1

Isn't AGI just "general" intelligence as in -like a regular human- turing test kinda deal?

aren't you thinking about ASI/ Superintelligence way capable of outdoing humans?

Yes, a general consensus is AGI should be able to perform any task an average human is able to perform. Definitely nothing of Nobel prize level.

A bit poorly named; not really very general. AHI would be a better name.

Another general consensus is that humans possess general intelligence.

Yes, we do seem to have a very high opinion of ourselves.

> Yes, a general consensus is AGI should be able to perform any task an average human is able to perform.

The goalposts are regularly moved so that AI companies and their investors can claim/hype that AGI will be around in a few years. :-)

I learned the definition I provided back in mid 90s, and it hasn't really changed since then.

AI will face the same limitations we face: availability of information and the non deterministic nature of the world.

you'd be able to give them a novel problem and have them generalize from known concepts to solve it. here's an example:

1 write a specification for a language in natural language

2 write an example program

can you feed 1 into a model and have it produce a compiler for 2 that works as reliably as a classically built one?

I think that's a low bar that hasn't been approached yet. until then I don't see evidence of language models' ability to reason.

I'd accept that as a human kind of intelligence, but I'm really hoping that AGI would be a bit more general. That clever human thinking would be a subset of what it could do.

You could ask Gemini 2.5 to do that today and it's well within its capabilities, just as long as you also let it write and run unit tests, as a human developer would.

AGI isn't ASI; it's not supposed to be smarter than humans. The people who say AGI is far away are unscientific woo-mongers, because they never give a concrete, empirically measurable definition of AGI. The closest we have is Humanity's Last Exam, which LLMs are already well on the path to acing.

I'd expect it to be generalised, where we (and everything else we've ever met) are specialised. Our intelligence is shaped by our biology and our environment; the limitations on our thinking are themselves concepts the best of us can barely glimpse. Some kind of intelligence that inherently transcends its substrate.

What that would look like, how it would think, the kind of mental considerations it would have, I do not know. I do suspect that declaring something that thinks like us would have "general intelligence" to be a symptom of our limited thinking.

Thirty years. Just enough time to call it quits and head to Costa Rica.

Two more weeks

I guess I am agnostic then.

LLMs are basically a library that can talk.

That’s not artificial intelligence.

There’s increasing evidence that LLMs are more than that. Especially work by Anthropic has been showing how to trace the internal logic of an LLM as it answers a question. They can in fact reason over facts contained in the model, not just repeat already seen information.

A simple example is how LLMs do math. They are not calculators and have not memorized every sum in existence. Instead they deploy a whole set of mental math techniques that were discovered at training time. For example, Claude uses a special trick for adding 2 digit numbers ending in 6 and 9.

Many more examples in this recent reach report, including evidence of future planning while writing rhyming poetry.

https://www.anthropic.com/research/tracing-thoughts-language...

Grammar engines. Or value matrix engines.

Everytime I try to work with them I lose more time than I gain. Net loss every time. Immensely frustrating. If i focus it on a small subtask I can gain some time (rough draft of a test). Anything more advanced and its a monumental waste of time.

They are not even good librarians. They fail miserably at cross referencing and contextualizing without constant leading.

I feel the opposite.

LLMs are unbelievably useful for me - never have I had a tool more powerful to assist my brain work. I useLLMs for work and play constantly every day.

It pretends to sound like a person and can mimic speech and write and is all around perhaps the greatest wonder created by humanity.

It’s still not artificial intelligence though, it’s a talking library.

Fair. For engineering work they have been a terrible drain on me save for the most minor autocomplete. Its recommendations are often deeply flawed or almost totally hallucinated no matter the model. Maybe I am a better software engineer than a “prompt engineer”.

Ive tried to use them as a research assistant in a history project and they have been also quite bad in that respect because of the immense naivety in its approaches.

I couldn’t call them a librarian because librarians are studied and trained in cross referencing material.

They have helped me in some searches but not better than a search engine at a monumentally higher investment cost to the industry.

Then again, I am also speaking as someone who doesn’t like to offload all of my communications to those things. Use it or lose it, eh

"Literally who" and "literally who" put out statements while others out there ship out products.

Many such cases.