The gap will close between the quality of open source language models and Google's internal language models in 5 years

"Large models aren’t more capable in the long run if we can iterate faster on small models" within five years

Between 2030 and 2035 there will be evidence of different models of neural network collaborating better with eachother

The gap will close between the quality of open source language models and Google's internal language models in 2 years

Will it be possible to disentangle most of the features learned by a model comparable to GPT-2 this decade?

We are going to start running out of data to train large language models in [YEAR]

"Large models aren’t more capable in the long run if we can iterate faster on small models" within two years

Neural Nets will be able to robustly pursue a plan over multiple days better than the best human by the end of 2025

Will mechanistic interpretability be essentially solved for GPT-4 before 2030?

Neural Nets will be better at typical manual labour tasks than humans by the end of 2025