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

Andrej Karpathy – It will take a decade to work through the issues with agents

dwarkesh.com

Post reply on HN
Print view View original

491–500 of 1001 posts

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#491

Post by __loam »

ctoth wrote: post #289
jjulius wrote: post #247

Earlier quoted context omitted.

>Why is there a presumption that we (as people who have only studied CS) know enough about biology/neuroscience/evolution to make these comparisons? Hubris.

The hubris here isn't CS people making comparisons, it's assuming biological substrate matters. Your brain is doing computation with neurotransmitters instead of transistors. So what? The "chemicals not electricity" distinction is pure carbon chauvinism, like insisting hydraulic computers can't be compared to electronic ones because water isn't electricity. Evolution didn't discover some mystical process that imbues…

There is no evidence that neurons have remotely the same computational mechanism as a transistor.

Memorizing billions of answers from the training set also isn't that impressive.

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#492

Post by __loam »

GoatInGrey wrote: post #331
ctoth wrote: post #289

Earlier quoted context omitted.

The hubris here isn't CS people making comparisons, it's assuming biological substrate matters. Your brain is doing computation with neurotransmitters instead of transistors. So what? The "chemicals not electricity" distinction is pure carbon chauvinism, like insisting hydraulic computers can't be compared to electronic ones because water isn't electricity. Evolution didn't discover some mystical process that imbues…

This seems naively dismissive of arguments around substrates considering that playing "Go at superhuman levels" took 1MW of energy versus the 1-2 (or if you want to assume 100% of the brain was applied to the game, 20) watts consumed by the human brain.

How many examples did each system need to get good at the task too? It's currently a lot less for humans and we don't know why.

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#493

Post by qudat »

joshellington wrote: post #473

To throw two pennies in the ocean of this comment section - I’d argue we still lack schematic-level understanding of what “intelligence” even is or how it works. Not to mention how it interfaces with “consciousness”, and their likely relation to each other. Which kinda invalidates a lot of predictions/discussions of “AGI” or even in general “AI”. How can one identify Artificial Intelligence/AGI without a modicum of u…

The reason why it’s so hard to define intelligence or consciousness is because we are hopelessly biased with a datapoint of 1. We also apply this unjustified amount of mysticism around it.

https://bower.sh/who-will-understand-consciousness

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#494

Post by gabriel-uribe »

leptons wrote: post #154
ciconia wrote: post #36

It's funny how there's such a pervasive cynicism about AI in the developer community, yet everyone is still excited about vibe coding. Strange times...

What developer is excited about "vibe coding"? The only people excited about "vibe coding" are people who can't code.

Deeply excited about vibe coding -- my non-technical cofounder 'codes' now.

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#495

Post by benregenspan »

OtherShrezzing wrote: post #319
louiereederson wrote: post #286

Earlier quoted context omitted.

short oracle

Is anyone _not_ short Oracle? The downside risk for them is that they’ll lose a deal worth 10x their annual revenues. Their potential upside is that OpenAI (a company with lifetime revenues of ~$10bn) have committed to a $300bn lease, if Oracle manages build a fleet of datacenters faster than any company in history. If you’re not short, you definitely shouldn’t be long. They’re the only one of the big tech companies…

With the executive branch now picking "national champion" companies (as in Intel deal), I feel like there's a big new short risk to consider. Would the current administration allow Oracle to go to zero?

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#496

Post by tim333 »

superconduct123 wrote: post #233

I always get a weird feeling when AI researchers and CS people start talking about comparisons between human brains and AI/computers Why is there a presumption that we (as people who have only studied CS) know enough about biology/neuroscience/evolution to make these comparisons/parallels/analogies? I enjoy the discussions but I always get the thought in the back of my head "...remember you're listening to 2 CS major…

AI researchers and CS people and the rest of us are human brain users and so have some familiarity with them even if they haven't studied neuroscience.

You can make some comparisons between how they perform without really understanding how LLMs or brains work, like to me LLMs seem similar to the part human minds where you say stuff without thinking about it. But you never really get an LLM saying I was thinking about that stuff and figured this bit was wrong, because they don't really have that capability.

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#497

Post by nextworddev »

hax0ron3 wrote: post #355

If the transcript is accurate, Karpathy does not actually ever, in this interview, say that AGI is a decade away, or make any concrete claims about how far away AGI is. Patel's title is misleading.

There's a lot of salt here

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#498

Post by Rudybega »

BoorishBears wrote: post #438
scuppernong wrote: post #389

Earlier quoted context omitted.

genuinely curious to hear your reasoning for why this is the case. i'm always somewhere between bemused and annoyed opening the daily HN thread about AGI and seeing everyone's totally unfounded confidence in their predictions. my position is I have no idea what is going to happen.

what about the fact frontier labs are spending more compute on viral AI video slop and soon-to-be-obsoleted workplace usecases than research? Even if you don't understand the technicals, surely you understand if any party was on the verge of AGI they wouldn't behave as these companies behave?

> what about the fact frontier labs are spending more compute on viral AI video slop and soon-to-be-obsoleted workplace usecases than research?

That's a bold claim, please cite your sources.

It's hard to find super precise sources on this for 2025, but epochAI has a pretty good summary for 2024. (with core estimates drawn from the Information and NYT

https://epoch.ai/data-insights/openai-compute-spend

The most relevant quote: "These reports indicate that OpenAI spent $3 billion on training compute, $1.8 billion on inference compute, and $1 billion on research compute amortized over “multiple years”. For the purpose of this visualization, we estimate that the amortization schedule for research compute was two years, for $2 billion in research compute expenses incurred in 2024."

Unless you think that this rough breakdown has completely changed, I find it implausible that Sora and workplace usecases constitute ~42% of total training and inference spend (and I think you could probably argue a fair bit of that training spend is still "research" of a sort, which makes your statement even more implausible).

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#499

Post by echoangle »

BoorishBears wrote: post #438
scuppernong wrote: post #389

Earlier quoted context omitted.

genuinely curious to hear your reasoning for why this is the case. i'm always somewhere between bemused and annoyed opening the daily HN thread about AGI and seeing everyone's totally unfounded confidence in their predictions. my position is I have no idea what is going to happen.

what about the fact frontier labs are spending more compute on viral AI video slop and soon-to-be-obsoleted workplace usecases than research? Even if you don't understand the technicals, surely you understand if any party was on the verge of AGI they wouldn't behave as these companies behave?

What does that tell you about AI in 100 years though? We could have another AI winter and then a breakthrough and maybe the same cycle a few times more and could still somehow get AGI at the end. I’m not saying it’s likely but you can’t predict the far future from current companies.

Re: Andrej Karpathy – It will take a decade to work through the issues with agents

#500

Post by balder1991 »

overfeed wrote: post #431
tyre wrote: post #226

Earlier quoted context omitted.

There is some evidence from Anthropic that LLMs do model the world. This paper[0] tracing their "thought" is fascinating. Basically an LLM translating across languages will "light up" (to use a rough fMRI equivalent) for the same concepts (e.g. bigness) across languages. It does have clusters of parameters that correlate with concepts, not just randomly "after X word tends to have Y word." Otherwise you would expect…

> Basically an LLM translating across languages will "light up" for the same concepts across languages Which is exactly what they are trained to do. Translation models wouldn't be functional if they are unable to correlate an input to specific outputs. That some hiddel-layer neurons fire for the same concept shouldn't come as a surprise, and is a basic feature required for the core functionality.

And if it is true that the language is just the last step after the answer is already conceptualized, why do models perform differently in different languages? If it was just a matter of language, they’d have the same answer but just with a broken grammar, no?
Post reply on HN