WILL SCALING WORK?

Summary from the TLDR newsletter

If scaling is successful in getting better and more general performance from large language models, then we should expect powerful AIs before 2040, but if it doesn't, then the path to artificial general intelligence seems much longer and more intractable.