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

591–600 of 1001 posts

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

#591

Post by bigtones »

Andrej Karpathy seems to me like a national (world) treasure.

He has the ability to explain concepts and thoughts with analogies and generalizations and interesting sayings that allow you to keep interest in what he is talking about for literally hours - in a subject that I don't know that much about. Clearly he is very smart, as is the interviewer, but he is also a fantastic communicator and does not come across as arrogant or pretentious, but really just helpful and friendly. Its quite a remarkable and amazing skillset. I'm in awe.

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

#592

Post by ajross »

pastel8739 wrote: post #571
ajross wrote: post #293

Earlier quoted context omitted.

> Knowing which parts-of-speech about sunrises appear together and where is not the same as understanding a sunrise What does "understanding a sunrise" mean though? Arguments like this end up resting on semantics or tautology, 100% of the time. Arguments of the form "what AI is really doing" likewise fail because we don't know what real brains are "really" doing either . I mean, if we knew how to model human language…

Is it really so rare? I feel like I know of tons of fields where we have methods that work empirically but don’t understand all the theory. I’d actually argue that we don’t know what’s “actually” happening _ever_, but only have built enough understanding to do useful things.

I mean, most big changes in the tech base don't have that characteristic. Semiconductors require only 1920's physics to describe (and a ton of experimentation to figure out how to manufacture). The motor revolution of the early 1900's was all built on well-settled thermodynamics (chemistry lagged a bit, but you don't need a lot of chemical theory to burn stuff). Maxwell's electrodynamics explained all of industrial electrification but predated it by 50 years, etc...

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

#593

Post by skydhash »

jhanschoo wrote: post #520
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…

Let's make this more concrete than talking about "understanding knowledge". Oftentimes I want to know something that cannot feasibly be arrived at by reasoning, only empirically. Remaining within the language domain, LLMs get so much more useful when they can search the web for news, or your codebase to know how it is organized. Similarly, you need a robot that can interact with the world and reason from newly collec…

> LLMs get so much more useful when they can search the web for news, or your codebase to know how it is organized

But their usefulness is only surface-deep. The news that matters to you is always deeply contextual, it's not only things labelled as breaking news or happening near you. Same thing happens with code organization. The reason is more human nature (how we think and learn) than machine optimization (the compiler usually don't care).

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

#594

Post by hax0ron3 »

tim333 wrote: post #508
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.

He says re agents: >They don't have enough intelligence, they're not multimodal enough, they can't do computer use and all this stuff. They don't do a lot of the things you've alluded to earlier. They don't have continual learning. You can't just tell them something and they'll remember it. They're cognitively lacking and it's just not working. >It will take about a decade to work through all of those issues. (2:20)

Him saying that it will take a decade to work through agents' issues isn't the same as him saying that there will be AGI in a decade, though

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

#595

Post by jonas21 »

CamperBob2 wrote: post #397
kart23 wrote: post #162

Earlier quoted context omitted.

I don't know a single one of the "Social Science" items, and I'm pretty sure 90% of college educated people wouldn't know a single one either.

Not only that, but the notion that GPT-5 will answer those questions with only 2% accuracy seems suspect. Those are exactly the kinds of questions that current models are great at. Nothing about that page makes much sense.

The percentages are added, not averaged. Each category sums to 10%, and the General Knowledge category has 5 equally-weighted subcategories, so 2% is the best possible score you can get in the social science subcategory.

I don't know why they decided to do it this way. It's very confusing.

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

#596

Post by hax0ron3 »

dang wrote: post #518
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.

Hmm good point. I skimmed the transcript looking for an accurate, representative quote that we could use in the title above. I couldn't exactly find one (within HN's 80 char limit), so I cobbled together "It will take a decade to get agents to work", which is at least closer to what Karpathy actually said. If anyone can suggest a more accurate and representative title, we can change it again. Edit: I thought of using…

To be fair to the OP of the thread, he's just using Patel's title word-for-word. It's Patel who is being inaccurate.

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

#597

Post by hax0ron3 »

meindnoch wrote: post #320

People keep talking about AGI as if it's some mystical leap beyond human capability. But let's be honest; software development at a modern startup is already the upper bound of applied intelligence. You're juggling shifting product specs, ambiguous user feedback, legacy code written by interns, and five competing JS frameworks, all while shipping on a Friday. Models can now do that. They can reason about asynchronous…

Models can't do that now, though. If they could, pretty much every human software engineer would be unemployed right now.

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

#598

Post by scrubs »

awesome_dude wrote: post #527
jhanschoo wrote: post #520

Earlier quoted context omitted.

Let's make this more concrete than talking about "understanding knowledge". Oftentimes I want to know something that cannot feasibly be arrived at by reasoning, only empirically. Remaining within the language domain, LLMs get so much more useful when they can search the web for news, or your codebase to know how it is organized. Similarly, you need a robot that can interact with the world and reason from newly collec…

I know the attributes of an Apple, i know the attributes of a Pear. As does a computer. But only i can bite into one and know without any doubt what it is and how it feels emotionally.

You have half a point. "Without any doubt" is merely the apex of a huge undefined iceberg.

I write half .. eating is multi modal and consequential. The llm can read the menu, but it didn't eat the meal. Even humans are bounded. Feeling, licking, smelling, or eating the menu still is not eating the meal.

There is an insuperable gap in the analogy ... a gap in the concept and of sensory data doing it.

Back to first point: what one knows through that sensory data ... is not clear at present or even possible with llms.

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

#599

Post by chipsrafferty »

hashiyakshmi wrote: post #578
klipt wrote: post #563

Earlier quoted context omitted.

What about a blind human? Are they just like an LLM? What about a multimodal model trained on video? Is that like a human?

This is actually a great point but for the opposite reason - if you ask a blind person if the night sky is beautiful, they would say they don't know because they've never seen it (they might add that they've heard other people describe it as such). Meanwhile, I just asked ChatGPT "Do you think the night sky is beautiful?" And it responded "Yes, I do..." and went on to explain why while describing senses its incapable…

I just asked Gemini and it said "I don't have eyes or the capacity to feel emotions like "beauty""
Post reply on HN