> I wonder if they can even effectively make use of the funds in a reasonable enough timeline.
I read that it cost Google ~$190 million to train Gemini, not even including staff salaries. So feels like a billion gives you about 3 "from scratch" comparable training runs.
Your estimate seems way off given Google already had their own compute hardware and staff. And if this company is going straight for AGI there's no way $1 billion is enough.
I read that it cost Google ~$190 million to train Gemini, not even including staff salaries. So feels like a billion gives you about 3 "from scratch" comparable training runs.