Overall probably a good move for Google. They were were spending 100s of millions of dollars a year to train RL agents to get good at playing Qwop for close to a decade.
Research is how Google invented the transformer that underlies so many current Gen AI models.
Pivoting research to products is exactly the kind of short term thinking consultants or mercenaries would propose, get promoted off of, and leave just before the consequences start rearing their head
Also, there were serious mathematicians saying a few years prior that computers would never surpass humans in Go, at least during our lifetimes.