Hacker News Reader

Comment by photochemsyn | Hacker News Reader

“Fable was the strongest model on the market” - explain why anyone should believe that claim.

I’ve been trying to track LLM code generation adoption in the critical infrastructure world - as far as I can tell, it’s nill. Zero. Nada. Nobody is relying on these models to write secure code for anything where failure is catastrophic. Planes falling out of the sky. Nuclear reactors going into meltdown. Electrical grids loosing synchronicity. Lots of these BS claims from the marketing and investment crowd, but - it’s just a useful tool for non-critical areas. That’s all it is.

View on HN

I don't understand the point you're making.

It can be both the most powerful LLM on the market, and have no adoption in critical infrastructure.

> it’s just a useful tool for non-critical areas. That’s all it is.

Okay. Let's say I agreed with you.

If you look at all technology and break down the total market for Critical Workloads vs non-critical workloads, what do you think that works out too, percentage wise? 12% critical? 18%? What if it was 30%! That would still mean 70% of the world's software could possibly be handled by an LLM. If that happens, the 30% of the Critical Workloads stuff is gonna get very, very competitive.

Not if the government bans them.

What government? Models are available all over the world. They can be run locally. Are we banning the Internet?

Genies and bottles.

i know someone who works on nuclear power plants that uses codex

obviously you need to review it

That's terrifying.

It is. Not per sé because the code might be of poor quality, but because someone sent that source code to a public API under the promise that oh noooo we won't use your code for training. Probably.

If they ever used Fable, it's sitting in Anthropic's servers for a month

enterprise agreements with model providers let you opt out of training

That's what I mean with the promise. Nobody knows what they're doing on their end. The data goes over the wire, and then you need to assume they are true to their word and the data is not intercepted by anyone.

Reply on news.ycombinator.com