5 ms·
We just need to train an LLM how to make an LLM that can make even better LLMs then rinse and repeat. Nothing beats working yourself out of a job.
by StimDeck 3y ago
We just need to train an LLM how to make an LLM that can make even better LLMs then rinse and repeat. Nothing beats working yourself out of a job.
- starbugs 3y agoWhat training data are you gonna use for that? That which comes out of an LLM? OK. Now you have a degrading loop of self-reinforcing falsehood.