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

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

youtube.com

Post reply on HN
Print view View original

211–220 of 827 posts

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

#211

Post by yahoozoo »

abdullin wrote: post #102

Tight feedback loops are the key in working productively with software. I see that in codebases up to 700k lines of code (legacy 30yo 4GL ERP systems). The best part is that AI-driven systems are fine with running even more tight loops than what a sane human would tolerate. Eg. running full linting, testing and E2E/simulation suite after any minor change. Or generating 4 versions of PR for the same task so that the h…

The problem is that every time you run your full automation with linting and tests, you’re filling up the context window more and more. I don’t know how people using Claude do it with its <300k context window. I get the “your message will exceed the length of this chat” message so many times.

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

#212

Post by belter »

coreyh14444 wrote: post #170
mentalgear wrote: post #157

Meanwhile, I asked this morning Claude 4 to write a simple EXIF normalizer. After two rounds of prompting it to double-check its code, I still had to point out that it makes no sense to load the entire image for re-orientating if the EXIF orientation is fine in the first place. Vibe vs reality, and anyone actually working in the space daily can attest how brittle these systems are. Maybe this changes in SWE with more…

https://theeducationist.info/everything-amazing-nobody-happy...

AI Snake Oil: https://press.princeton.edu/books/hardcover/9780691249131/ai...

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

#213

Post by bgwalter »

mellosouls wrote: post #199
bgwalter wrote: post #183

I'd like to hear from Linux kernel developers. There is no significant software that has been written (plagiarized) by "AI". Why not ask the actual experts who deliver instead of talk? This whole thing is a religion.

There is no significant software that has been written (plagiarized) by "AI". How do you know? As you haven't evidenced your claim, you could start by providing explicit examples of what is significant. Even if you are correct, the amount of llm-assisted code is increasing all the time, and we are still only a couple of years in - give it time. Why not ask the actual experts Many would regard Karpathy in the expert c…

I think you should not turn things around here. Up to 2021 we had a vibrant software environment that obviously had zero "AI" input. It has made companies and some developers filthy rich.

Since "AI" became a religion, it is used as an excuse for layoffs while no serious software is written by "AI". The "AI" people are making the claims. Since they invading a functioning software environment, it is their responsibility to back up their claims.

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

#214

Post by kubb »

diggan wrote: post #189
kubb wrote: post #174

Earlier quoted context omitted.

Or will they ever be reliable. Your question is already making an assumption.

They're reliable already if you change the way you approach them. These probabilistic token generators probably never will be "reliable" if you expect them to 100% always output exactly what you had in mind, without iterating in user-space (the prompts).

I also think they might never become reliable.

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

#215

Post by voidhorse »

practal wrote: post #124
singularity2001 wrote: post #86

Earlier quoted context omitted.

lean 4/5 will be a rising star!

You would definitely think so, Lean is in a great position here! I am betting though that type theory is not the right logic for this, and that Lean can be leapfrogged.

Why? By the completeness theorem, shouldn't first order logic already be sufficient?

The calculus of constructions and other approaches are already available and proven. I'm not sure why we'd need a special logic for LLMs unless said logic somehow accounts for their inherently stochastic tendencies.

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

#216

Post by diggan »

rwmj wrote: post #202
diggan wrote: post #190

Earlier quoted context omitted.

What counts as "significant software"? Only kernels I guess?

Can you point to any significant open source software that has any kind of significant AI contributions? As an actual open source developer I'm not seeing anything. I am getting bogus pull requests full of AI slop that are causing problems though.

> Can you point to any significant open source software that has any kind of significant AI contributions?

No, but I haven't looked. Can you?

As an actual open source developer too, I do get some value from replacing search engine usage with LLMs that can do the searching and collation for me, as long as they have references I can use for diving deeper, they certainly accelerate my own workflow. But I don't do "vibe-coding" or use any LLM-connected editors, just my own written software that is mostly various CLIs and chat-like UIs.

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

#217

Post by agile-gift0262 »

yahoozoo wrote: post #204
ramon156 wrote: post #172

Earlier quoted context omitted.

The real question is how long it'll take until they're not brittle

“Treat it like a junior developer” … 5 years later … “Treat it like a junior developer”

while True:

  print("This model that just came out changes everything. It's flawless. It doesn't have any of the issues the model from 6 months ago had. We are 1 year away from AGI and becoming jobless")
  sleep(timedelta(days=180).total_seconds)

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

#218

Post by diggan »

xvilka wrote: post #201
diggan wrote: post #190

Earlier quoted context omitted.

What counts as "significant software"? Only kernels I guess?

Office software, CAD systems, Web Browsers, the list is long.

Microsoft (famously developing somewhat popular office-like software) seems to be going in the direction of almost forcing developers to use LLMs to assist with coding, at least going by what people are willing to admit publicly and seeing some GitHub activity.

Google (made a small browser or something) also develops their own models, I don't think it's far fetched to imagine there is at least one developer on the Chrome/Chromium team that is trying to dogfood that stuff.

As for Autodesk, I have no idea what they're up to, but corporate IT seems hellbent on killing themselves, not sure Autodesk would do anything differently so they're probably also trying to jam LLMs down their employees throats.

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

#219

Post by demaga »

diggan wrote: post #169
mentalgear wrote: post #157

Meanwhile, I asked this morning Claude 4 to write a simple EXIF normalizer. After two rounds of prompting it to double-check its code, I still had to point out that it makes no sense to load the entire image for re-orientating if the EXIF orientation is fine in the first place. Vibe vs reality, and anyone actually working in the space daily can attest how brittle these systems are. Maybe this changes in SWE with more…

> Meanwhile What do you mean "meanwhile", that's exactly (among other things) the kind of stuff he's talking about? The various frictions and how you need to approach it > anyone actually working in the space Is this trying to say that Karpathy doesn't "actually work" with LLMs or in the ML space? I feel like your whole comment is just reacting to the title of the YouTube video, rather than actually thinking and refl…

I'm pretty sure "actually work" part refers to SWE space rather than LLM/ML space
Post reply on HN