Security Writer :verified: :donor: on Nostr: nprofile1q…rcfcs Just regarding the diminishing returns of scaling LLMs, and that ...
nprofile1qy2hwumn8ghj7un9d3shjtnddaehgu3wwp6kyqpq66uezaptup9yn3xlym6ynw3gyymajvhrme68j9krncghdqw5jeaqarcfcs (nprofile…cfcs) Just regarding the diminishing returns of scaling LLMs, and that small language models are more practical.
We’re also reaching the limit of the amount we can ‘tune’ out of LLM generation before they become as dumb as rocks and standard big data processing becomes more effective again.
We’re also reaching the limit of the amount we can ‘tune’ out of LLM generation before they become as dumb as rocks and standard big data processing becomes more effective again.