3 ms·
1/20 is less than my guess would have been. It's interesting that Wikipedia is heavily weighted in LLM training corpuses. Eventually there will be a feedback lo
by princearthur 2y ago
1/20 is less than my guess would have been. It's interesting that Wikipedia is heavily weighted in LLM training corpuses. Eventually there will be a feedback loop not unlike the one among meat popsicles.
How many covers are better than the original? Would we expect that to be possible with AI?