zlacker

For a long time there has been back chatter on how to turn programming into a more professional field, more like actual engineering where when something goes wrong actual people and companies start to take security seriously, and get held accountable for their mistakes, and start to actually earn their high salaries.

Getting AI to hallucinate its way into secure and better quality code seems like the antithesis of this. Why don't we have AI and robots working for humanity with the boring menial tasks - mowing laws, filing taxes, washing dishes, driving cars - instead of attempting to take on our more critical and creative outputs - image generation, movie generation, book writing and even website building.

>>hostyl+(OP)
The problem with this argument is that it's not what's going to happen. In the trajectory I see of LLM code generation, security quality between best-practices well-prompted (ie: not creatively well prompted, just people with a decent set of Instructions.md or whatever) and well trained human coders is going to be a wash. Maybe in 5 years SOTA models will clearly exceed human coders on this, but my premise is all progress stops and we just stick with what we have today.

But the analysis doesn't stop there, because after the raw quality wash, we have to consider things LLMs can do profoundly better than human coders can. Codebase instrumentation, static analysis, type system tuning, formal analysis: all things humans can do, spottily, on a good day but that empirically across most codebases they do not do. An LLM can just be told to spend an afternoon doing them.

I'm a security professional before I am anything else (vulnerability research, software security consulting) and my take on LLM codegen is that they're likely to be a profound win for security.

>>hostyl+(OP)
> Why don't we have AI and robots working for humanity with the boring menial tasks - mowing laws, filing taxes, washing dishes, driving cars

I mean, we do have automation for literally all of those things, to varying degrees of effectiveness.

There's an increasing number of little "roomba" style mowers around my neighborhood. I file taxes every year with FreeTaxUSA and while it's still annoying, a lot of menial "form-filling" labor has been taken away from me there. My dishwasher does a better job cleaning my dishes than I would by hand. And though there's been a huge amount of hype-driven BS around 'self-driving', we've undeniably made advances in that direction over the last decade.

>>tptace+0I
Isn't formal analysis exactly the kind of thing LLMs can't do at all? Or do you mean an LLM invoking a proof assistant or something like that?

>>pxc+8n1
Yes, I mean LLMs generating proof specs and invoking assistants, not that they themselves do any formal modeling.