Hacker News Reader

Comment by MattPalmer1086 | Hacker News Reader

How will AIs train on new tech without new data to train on?

On the bright side, maybe that means the end of new javascript frameworks every 6 months :)

If needed it can synthetically create the datasets. There isn't a need for software engineers to make a dataset when AI could even more easily put one together.

Do you understand reinforcement learning environments? This already a solved problem

I think they’ll be writing that new tech, and will be able to read the source code, and reason from first principles. I doubt they will need human generated training data.

I thought training new models on AI generated output leads to model collapse?

My point is more that they won’t need new training data to extrapolate to new problems, especially if that new problem is just a new syntax or API. Put the whole shebang into the context. Done. Yes, you need super long context for this, or just pretty good search over it, but I think this is likely to become a solved problem.

Agreed if its just a new API using a current language using common patterns, that is mostly just having a big context.

However, the ability to reason generally on novel problems is AGI and we aren't there yet. Eventually in the absence of AGI, we will have to train models on them, and that will require data.

I think it case of coding it may not be as bad, because new training data (AI generated code) is always empirically validated by tooling and by consumer. It may not be good but it mostly works, otherwise it is discarded or patched, so it has a bottom bar of "it works".

Reply on news.ycombinator.com