Andrej Karpathy: Software in the era of AI [video] • Hacker News Board

Re: Andrej Karpathy: Software in the era of AI [video]

#701

Post by msgodel » Fri, Jun 20, 2025, 10:24 AM UTC

throwawayoldie wrote: post #16

reducesuffering wrote: post #11

Earlier quoted context omitted.

[flagged]

I'm old enough to remember when Twitter was new, and for a moment it felt like the old utopian promise of the Internet finally fulfilled: ordinary people would be able to talk, one-on-one and unmediated, with other ordinary people across the world, and in the process we'd find out that we're all more similar than different and mainly want the same things out of life, leading to a new era of peace and empathy. It was…

Was Twitter ever really meant for that? As far as I can tell the primary purpose of twitter is moderated access to celebrities with the utopian ideas about communication just used to sell it.

Re: Andrej Karpathy: Software in the era of AI [video]

#702

Post by diggan » Fri, Jun 20, 2025, 10:43 AM UTC

iLoveOncall wrote: post #693

OJFord wrote: post #661

I'm not sure about the 1.0/2.0/3.0 classification, but it did lead me to think about LLMs as a programming paradigm: we've had imperative & declarative, procedural & functional languages, maybe we'll come to view deterministic vs. probabilistic (LLMs) similarly. def __main__: You are a calculator. Given an input expression, you compute the result and print it to stdout, exiting 0. Should you be unable to do this, you…

> maybe we'll come to view deterministic vs. probabilistic (LLMs) similarly I can't believe someone would seriously write this and not realize how nonsensical it is. "indeterministic programming", you seriously cannot come up with a bigger oxymoron.

Why do people keep having this reaction to something we're already used to? When you're developing against an API, you're already doing the same thing, planning for what happens when the request hangs, or fails completely, or gives a different response, and so on. Same for basically any IO.

It's almost not even new, just that it generates text instead of JSON, or whatever. But we've already been doing "indeterministic programming" for a long time, where you cannot always assume a function 100% returns what it should all the time.

Re: Andrej Karpathy: Software in the era of AI [video]

#703

Post by diggan » Fri, Jun 20, 2025, 10:47 AM UTC

aaron695 wrote: post #686

OJFord wrote: post #661

I'm not sure about the 1.0/2.0/3.0 classification, but it did lead me to think about LLMs as a programming paradigm: we've had imperative & declarative, procedural & functional languages, maybe we'll come to view deterministic vs. probabilistic (LLMs) similarly. def __main__: You are a calculator. Given an input expression, you compute the result and print it to stdout, exiting 0. Should you be unable to do this, you…

[flagged]

> It makes no sense at all, it's cuckooland, are you all on crazy pills?

First step towards understanding something you obviously have strong feelings about, is to try to avoid hitting those triggers while you think about the thing, otherwise it clouds you. Not a requirement by any measure, just a tip.

> are you telling me people will do three years university to learn to prompt?

Are people going to university for three years to write "1.0" or "2.0" software? I certainly didn't, and I don't think even the majority of software developers have done so, at least in my personal experience but YMMV.

> I do not understand where there is anything here to be "not sure" on?

They're not sure about the specific naming, not the concept or talk as a whole.

> LLMs making non-deterministic mistakes

Everything they do is non-deterministic when temperature is set to anything above 0.0, as that's the entire point. The "correct" answers are as non-deterministic as the "mistakes", although I'm not sure "mistake" is correct because it did chose the right/correct tokens, it's just that you didn't like/expect it to chose that particular tokens.

Re: Andrej Karpathy: Software in the era of AI [video]

#704

Post by manyaoman » Fri, Jun 20, 2025, 10:48 AM UTC

iLoveOncall wrote: post #697

manyaoman wrote: post #694

Earlier quoted context omitted.

The full quote from 7:40 in the video: "I think it's kind of fascinating to me that when the state-of-the-art LLMs go down, it's actually kind of like an intelligence brownout in the world. It's kind of like when the voltage is unreliable in the grid, and the planet just gets dumber. The more reliance we have on these models, which already is really dramatic and I think will continue to grow." I don't think his point…

I'm just saying that the statement "when the state-of-the-art LLMs go down, it's actually kind of like an intelligence brownout in the world" is entirely false.

All analogies are false, but some are useful.

Re: Andrej Karpathy: Software in the era of AI [video]

#705

Post by diggan » Fri, Jun 20, 2025, 10:49 AM UTC

throwaway314155 wrote: post #596

redbell wrote: post #581

Earlier quoted context omitted.

> "English is the new programming language." For those who missed it, here's the viral tweet by Karpathy himself: https://x.com/karpathy/status/1617979122625712128

Referenced in the video of course. Not that everyone should watch a 40 minute long video before commenting but his reaction to the "meme" that vibe coding became when his tweet was intended as more of a shower thought is worth checking out.

> became when his tweet was intended as more of a shower thought

That was so obvious to me, and most of the people I talked to at the time, yet the ecosystem and media seems to have run with his "vibe-coding" idea as something people should implement in production yesterday, even though it wasn't meant as a "mantra" or even "here is where we should go"...

Re: Andrej Karpathy: Software in the era of AI [video]

#706

Post by diggan » Fri, Jun 20, 2025, 10:53 AM UTC

jph00 wrote: post #164

dncornholio wrote: post #122

Earlier quoted context omitted.

> You can't just put things there any time you want - the RFC requires that they go through a registration process. Excuse me???

From the RFC: """ A well-known URI is a URI [RFC3986] whose path component begins with the characters "/.well-known/", and whose scheme is "HTTP", "HTTPS", or another scheme that has explicitly been specified to use well- known URIs. Applications that wish to mint new well-known URIs MUST register them, following the procedures in Section 5.1. """

Keyword being "mint" there. You can still put whatever you want in there, but in order to "register" it, you need to"mint" it by registering it. But you're in no way obligated to register random stuff you put in /.well-known, that'd be bananas to put in a specification like that.

Re: Andrej Karpathy: Software in the era of AI [video]

#707

Post by diggan » Fri, Jun 20, 2025, 10:54 AM UTC

dang wrote: post #25

seneca wrote: post #20

Earlier quoted context omitted.

Classic under promise and over deliver. I'm glad they got it out quickly.

Me too. It was my favorite talk of the ones I saw.

I did like it to, but haven't seen any of the others, and usually don't like sitting through talks rather than reading transcripts.

But you got me curious, what other talks from the day/event would be worth watching in your mind?

Re: Andrej Karpathy: Software in the era of AI [video]

#708

Post by diggan » Fri, Jun 20, 2025, 10:57 AM UTC

rwmj wrote: post #210

imiric wrote: post #178

It's fascinating to see his gears grinding at 22:55 when acknowledging that a human still has to review the thousand lines of LLM-generated code for bugs and security issues if they're "actually trying to get work done". Yet these are the tools that are supposed to make us hyperproductive? This is "Software 3.0"? Give me a break.

Plus coding is the fun bit, reviewing code is the hard and not fun bit, arguing with an overconfident machine sound like it'll be worse even than that. Thankfully I'm going to retire soon.

> Plus coding is the fun bit, reviewing code is the hard and not fun bit

To you. For others, it looks differently. And for yet others, they don't care about the coding nor the reviewing, they want to solve a particular problem.

I'd probably say I'm a programmer by accident. It's not that I love producing binaries by writing and compiling code, but I need to solve some particular problem that either is best solved by programming, or can only be solved by programming. "Programming by need" maybe is a fitting definition.

Doesn't mean I don't care about code quality, or good abstractions and having a reasonable design/architecture. But I'm focused on the end goal, having a particular problem solved, and coding is just the way there (sometimes).

Re: Andrej Karpathy: Software in the era of AI [video]

#709

Post by diggan » Fri, Jun 20, 2025, 10:58 AM UTC

guappa wrote: post #161

ast0708 wrote: post #77

Should we not treat LLMs more as a UX feature to interact with a domain specific model (highly contextual), rather than expecting LLMs to provide the intelligence needed for software to act as partner to Humans.

He's selling something.

What, exactly? Educational courses?

Re: Andrej Karpathy: Software in the era of AI [video]

#710

Post by diggan » Fri, Jun 20, 2025, 11:01 AM UTC

cedws wrote: post #641

Aeroi wrote: post #338

the fanboying for this dudes opinion is insane.

I'm about half way through the video and I'm really not seeing what all the praise is about, it just seems to be an AI optimism word salad. edit: a lot of the comments giving praise on YouTube look like bots...

> it just seems to be an AI optimism word salad

Maybe it was missed, but in the beginning he said he was told the audience are mostly students about to enter the industry, so I feel like a lot of the talk is just establishing vocabulary, basic information about what LLMs are, analogies to get people to wrap their head around where in the workflow they can fit, and so on.

So while most of it seems obvious or relatively abstract, I think that's because of the target audience of the talk. I had that lens while watching the talk and while I cannot say my worldview has done a large change because of it, I understand it could be valuable to newer members of the ecosystem.