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

321–330 of 827 posts

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

#321

Post by abdullin »

3dsnano wrote: post #277
diggan wrote: post #266

Earlier quoted context omitted.

> A truly terrible and demotivating way to work and produce anything of real quality You clearly have strong feelings about it, which is fine, but it would be much more interesting to know exactly why it would terrible and demotivating, and why it cannot produce anything of quality? And what is "real quality" and does that mean "fake quality" exists? > million monkey interns banging on one million keyboards and submi…

> And what is "real quality" and does that mean "fake quality" exists? I think there is no real quality or fake quality, just quality. I am referencing the quality that Persig and C. Alexander have written about. It’s… qualitative, so it’s hard to measure but easy to feel. Humans are really good at perceiving it then making objective decisions. LLMs don’t know what it is (they’ve heard about it and think they know).

It is actually funny that current AI+Coding tools benefit a lot from domain context and other information along the lines of Domain-Driven Design (which was inspired by the pattern language of C. Alexander).

A few teams have started incorporating `CONTEXT.MD` into module descriptions to leverage this.

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

#322

Post by 9rx »

elif wrote: post #308
osigurdson wrote: post #296

Earlier quoted context omitted.

I'm not sure that AI code has to be sloppy. I've had some success with hand coding some examples and then asking codex to rigorously adhere to prior conventions. This can end up with very self consistent code. Agree though on the "pick the best PR" workflow. This is pure model training work and you should be compensated for it.

Yep this is what Andrej talks about around 20 minutes into this talk. You have to be extremely verbose in describing all of your requirements. There is seemingly no such thing as too much detail. The second you start being vague, even if it WOULD be clear to a person with common sense, the LLM views that vagueness as a potential aspect of it's own creative liberty.

> You have to be extremely verbose in describing all of your requirements. There is seemingly no such thing as too much detail.

If only there was a language one could use that enables describing all of your requirements in a unambiguous manner, ensuring that you have provided all the necessary detail.

Oh wait.

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

#323

Post by tamersalama »

blobbers wrote: post #114

Software 3.0 is the code generated by the machine, not the prompts that generated it. The prompts don't even yield the same output; there is randomness. The new software world is the massive amount of code that will be burped out by these agents, and it should quickly dwarf the human output.

How I understood it is that natural language will form relatively large portions of stacks (endpoint descriptions, instructions, prompts, documentations, etc…). In addition to code generated by agents (which would fall under 1.0)

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

#324

Post by NoOn3 »

infecto wrote: post #318
NoOn3 wrote: post #310

Earlier quoted context omitted.

It's cheap now. But if you take into account all the training costs, then at such prices they cannot make a profit in any way. This is called dumping to capture the market.

No doubt the complete cost of training and to getting where we are today has been significant and I don’t know how the accounting will look years from now but you are just making up the rest based on feelings. We know operationally OpenAI is profitable on purely the runtime side, nobody knows how that will look when accounting for R&D but you have no qualification to say they cannot make a profit in any way.

Yes, if you do not take into account the cost of training, I think it is very likely profitable. The cost of working models is not so high. This is just my opinion based on open models and I admit that I have not carried out accurate calculations.

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

#325

Post by shakna »

azan_ wrote: post #232
shakna wrote: post #143

Earlier quoted context omitted.

Human brain's aren't generalist! We have multiple parts of the brain that interact in vastly different ways! Your cerebellum won't be running the role of the pons. Most parts of the brain cannot take over for others. Self-healing is the exception, not the rule. Yes, we have a degree of neuroplasticity, but there are many limits. (Sidenote: Driver's license here is 240 hours.)

> We have multiple parts of the brain that interact in vastly different ways! Yes, and thanks to that human brains are generalist

Only if that was a singular system, however, it is not. [0]

For example... The nerve cells in your gut may speak to the brain, and interact with it in complex ways we are only just beginning to understand, but they are separate systems that both have control over the nervous system, and other systems. [1]

General Intelligence, the psychological theory, and General Modelling, whilst sharing words, share little else.

[0] https://doi.org/10.1016/j.neuroimage.2022.119673

[1] https://doi.org/10.1126/science.aau9973

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

#326

Post by crmi »

diggan wrote: post #269
crmi wrote: post #245

Earlier quoted context omitted.

I've got a working theory that models perform differently when used in different timezones... As in during US working hours they dont work as well due to high load. When used at 'offpeak' hours not only are they (obviously) snappier but the outputs appear to be a higher standard. Thought this for a while but now noticing with Claude4 [thinking] recently. Textbook case of anecdata of course though.

Interesting thought, if nothing less. Unless I misunderstand, it would be easy to run a study to see if this is true; use the API to send the same but slightly different prompt (as to avoid the caches) which has a definite answer, then run that once per hour for a week and see if the accuracy oscillates or not.

Yes good idea - although it appears we would also have to account for the possibility of providers nerfing their models. I've read others also think models are being quantized after a while to cut costs.

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

#328

Post by shakna »

Zanfa wrote: post #188
shakna wrote: post #143

Earlier quoted context omitted.

Human brain's aren't generalist! We have multiple parts of the brain that interact in vastly different ways! Your cerebellum won't be running the role of the pons. Most parts of the brain cannot take over for others. Self-healing is the exception, not the rule. Yes, we have a degree of neuroplasticity, but there are many limits. (Sidenote: Driver's license here is 240 hours.)

> Human brain's aren't generalist! What? Human intelligence is literally how AGI is defined. Brain’s physical configuration is irrelevant.

A human brain is not a general model. We have multiple overlapping systems. The physical configuration is extremely relevant to that.

AGI is defined in terms of "General Intelligence", a theory that general modelling is irrelevant to.

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

#329

Post by crmi »

sensanaty wrote: post #187
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…

There's also those instances where Microsoft unleashed Copilot on the .NET repo, and it resulted in the most hilariously terrible PRs that required the maintainers to basically tell Copilot every single step it should take to fix the issue. They were basically writing the PRs themselves at that point, except doing it through an intermediary that was much dumber, slower and less practical than them. And don't get me s…

To add to this, I ran into a lot of issues too. And similar when using cursor... Until I started creating a mega list of rules for it to follow that attaches to the prompts. Then outputs improved (but fell off after the context window got too large). At that stage I then used a prompt to summarize, to continue with a new context.

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

#330

Post by 9rx »

OvbiousError wrote: post #160
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…

I don't think the human is the problem here, but the time it takes to run the full testing suite.

Unless you are doing something crazy like letting the fuzzer run on every change (cache that shit), the full test suite taking a long time suggests that either your isolation points are way too large or you are letting the LLM cross isolated boundaries and "full testing suite" here actually means "multiple full testing suites". The latter is an easy fix: Don't let it. Force it stay within a single isolation zone just like you'd expect of a human. The former is a lot harder to fix, but I suppose ending up there is a strong indicator that you can't trust the human picking the best LLM result in the first place and that maybe this whole thing isn't a good idea for the people in your organization.
Post reply on HN