Hacker Newsnew | past | comments | ask | show | jobs | submit | dist-epoch's commentslogin

A company which doesn't want the big LLM providers to see it's prompts or data - military, health, finance, research

Your hometown also has public lightning, water pumps, and probably some other stuff.

Yes, it can generate Crysis with diffusion models at 60 fps.

In 1998 or 2005 two persons could single-handedly start a Google or a Facebook. Not possible anymore today in Internet.

But in AI a single person created OpenClaw.

It's called low-hanging fruits.


> But in AI a single person created OpenClaw.

Do you think no one can create anything alone ever again? Or can they only do it by adopting the bleeding edge?

> It's called low-hanging fruits.

1 in a million ideas are 1 in a million, and they don't require being a bleeding edge adopter of anything. Do you think no one can create a better version of a first-try service? Is the agentic world now closed because someone built a mediocre version of it?

For a start-up based board, this point-of-view just feels so sad and myopic.


NVIDIA Jensen predicted today that soon engineers will spend $20k in tokens every month.

General idea is true, except for this particular technology.

When AI will be easy to pick up and guide, guess what, there will be no need for a programmer to pick it up. AI will be using itself, Claude Manager driving Claude programmers.

So leverage AI while you still can provide value doing so.

It's literally a "use it or lose it situation".


If you want people to self-host, this is a gateway to that.

And those who are actually curious will look into what it's doing under the hood.

Everyone has to start somewhere.


Capability and alignment are orthogonal.

Stalin was AGI-level.


"Stalin was AGI-level" perfectly catches the core of my concerns. Thanks!

You do have a credit card, right?

I remember when "real programmers" were supposed to look at the assembly code generated by compilers because it was bloated, inefficient, and totally unsuitable to use in a real system.

Cue in "non-determinism" retort.


Hardware restrictions might have contributed to that. Anyway, analogs and metaphors do not prove what they sneakily try to imply. They might help thinking about a problem, but they leave out the actual argument, and in this case, the jump is substantial.

I think the problem is less determinism than predictability. Hashing algorithms are deterministic.

Will people start .gitignore-ing their src directories and only save prompts?


This article [1] would argue ”no”, because then you would be ridding yourself of a “repository of determinism”, which the prompts cannot replace.

You can build a system with non-deterministic properties but you need some sort of deterministic foundation to build working, usable systems. Non determinism from top to bottom is building on quicksand in a swamp.

[1] https://www.oreilly.com/radar/can-language-models-replace-co...


There are multiple reasons why binary repository managers like Artifactory exist.

And, arguably, the primary reason that perforce is still popular in some domains is that it works well with both large and opaque objects.

Once you've decided to version your src directory, whether a priori or because once bitten, twice shy, the next question is:

Is git the correct version control for this? Or are all the changes so large that git's advantages in merging and manually diffing things become irrelevant?



That you anticipated a retort isn’t enough. You also have to refute it.

Yeah compilers are deterministic and LLMs are not. The response to that?

The answer could very well be something like what’s in TFA namely formal verification. But an answer here is needed.


Human programmers are not deterministic either - give the same spec/task to 3 programmers and you'll get three different implementations.

Yet somehow this didn't stopped a giant software industry existing.


I have millions of years of training data on the non-determinism or whatever of humans. That’s how, uh, humans can interoperate with humans.

So you agree hard formal non-determinism is not required, just a sort of convergence of code and spec. Turns out LLMs can do this non-formal-deterministic human-deterministic programming too.

In my very first comment I made an openening for some kind of counter-argument, explicitly singling out the submitted piece here about formal verification. That you now seem to think that you are dribbling me into making some kind of concession that there might be use-case for LLMs here is beyond me.

I was asking you for a refutal, after all, not asking myself. Sarcastically pointing to compilers and meatbags in order to triumphantly conclude, aha so there’s a chance, isn’t quite it.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: