AI that can match humans at any task will be here in five to 10 years, Google DeepMind CEO says

At the moment it's more like not ever. There's a logarithmic decrease in gains from throwing more processing at LLMs, and we don't have a model for AGI that wouldn't require more knowledge than humans possess.

Because no investors are interested in putting up money for something if they’ll be too old to enjoy the profits. So, you tell them five years. When you fail to deliver, it’s OK. You tell them just five more and ask for more money.