Comment on Emergent introspective awareness in large language models
rah@hilariouschaos.com 3 days agoLLM limitations like “they only predict the next token” and other things that have already been falsified
What do LLMs do beyond predicting the next token?
kromem@lemmy.world 2 days ago
A few months back it was found that when writing rhyming couplets the model has already selected the second rhyming word when it was predicting the first word of the second line, meaning the model was planning the final rhyme tokens at least one full line ahead and not just predicting that final rhyme when it arrived at that token.
It’s probably wise to consider this finding in concert with the streetlight effect.
rah@hilariouschaos.com 2 days ago
What do you mean by that? What does it mean to “select” something in the context of a neural net with input nodes and output nodes?
How have you come to that conclusion?
kromem@lemmy.world 2 days ago
Read it for yourself here.
See the “Planning in Poems” section.
rah@hilariouschaos.com 2 days ago
Are you able to explain succinctly what you mean by “selected” so that we can communicate?