3 ms·
This is my exact takeaway too, and I'm always surprised it doesn't get mentioned often. If AI is truly groundbreaking, then shouldn't AI be able to re-implement
by ar_lan 1y ago
This is my exact takeaway too, and I'm always surprised it doesn't get mentioned often. If AI is truly groundbreaking, then shouldn't AI be able to re-implement itself? Which, to me, would imply that every AI company is not only full of software devs cannibalizing themselves, but the companies themselves also are.
- SJC_Hacker 1y agoThis is my watershed for true AGI. It should be able to create a smarter version of itself. Last I checked, feeding the output of an LLM back into its training data leads to a progressively worse LLM. (Note I'm not talking about distillation, which involves training a smaller model, by sacrificing accuracy. I'm referring to an equal or greater number of model parameters)
- fragmede 1y agoIf the LLM is given the code for its training and is able to improve that, does that count? Because it seems like a safe bet that we're already there, the only problem is latency of training runs.