Google released a new open-source language model, Flan-T5, "dramatically improving prompting and multi-step reasoning abilities".
Forecasting their language model evolution to getting close to average human expert level in 2024 in its ability to reason.
Source:
Twitter @quocleix
Principal Scientist, Google Brain Team
New open-source language model from Google AI: Flan-T5 🍮
— Quoc Le (@quocleix) October 21, 2022
Flan-T5 is instruction-finetuned on 1,800+ language tasks, leading to dramatically improved prompting and multi-step reasoning abilities.
Public models: https://t.co/bnYVnocJW2
Paper: https://t.co/3KPGJ3tgMw pic.twitter.com/PQpUQM0net