Hacker News new | past | comments | ask | show | jobs | submit login

Mixture of Experts model is likely the most significant.

And the scale of everything. GPT3 embedding vectors are around 12,000, vs 768 shown here.

I was curious and the 12k figure closely approximates the median synapse dimensionality of human neurons. Maybe we don’t need much more.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: