Comment by jennyholzer3
17 hours ago
I don't know about all this AI stuff.
How are LLMs going to stay on top of new design concepts, new languages, really anything new?
Can LLMs be trained to operate "fluently" with regards to a genuinely new concept?
I think LLMs are good for writing certain types of "bad code", i.e. if you're learning a new language or trying to quickly create a prototype.
However to me it seems like a security risk to try to write "good code" with an LLM.
I've used LLMs to find bugs in and write code for a language almost nobody uses, that has terrible documentation. My assumption is that it did so in the same manner a human would: "ok this looks kind of like Algol or C" and "after reading a bunch of this code I think I get what's going on".
I suspect it will still fall on humans (with machine assistance?) to move the field forward and innovate, but in terms of training an LLM on genuinely new concepts, they tend to be pretty nimble on that front (in my experience).
Especially with the massive context windows modern LLMs have. The core idea that the GPT-3 paper introduced was (summarizing):
You do realise they can search the web? They can read documentation and api specs?
They can't think though. They can't be creative.
They are retrained every 12-24 months and constantly getting new/updated reinforcement learning layers. New concepts are not the problem. The problem is outdated information in the training data, like only crappy old Postgres syntax in most of the Stackoverflow body.
> They are retrained every 12-24 months and constantly getting new/updated reinforcement learning layers
This is true now, but it can't stay true, given the enormous costs of training. Inference is expensive enough as is, the training runs are 100% venture capital "startup" funding and pretty much everyone expects them to go away sooner or later
Can't plan a business around something that volatile
You don't need to retrain the whole thing from scratch every time.
GPT-5.1 was based on over 15 months old data IIRC, and it wasn’t that bad. Adding new layers isn’t that expensive.
Google's training runs aren't funded by VC. The Chinese models probably aren't either.