How to Accelerate ML Research

How far does it get us? We're talking so far here about we have large language models which are trained on, you know, giant transformers with billions of parameters. And then we're adjusting the data going into them or adjusting small things about the way the layers are connected together. But how far does that gets us without any big fundamental shift or big new ideas on how to generate intelligence? Yeah, this is unclear. Even with like our current models, I think this could accelerate people's research,. Like my ability to do research as a Google researcher, maybe five times more than before.