He says that Google has the AI read Twitter, which means it learns from pro-left and anti-right conversations. So Google is building a leftwing AI. Chances are it also also picks up from Reddit.

But anyway, the AI is NOT sentient. It is a simulation of sentience. There is a big difference. Humans have a Self and that part of you has personal goals. The AI has no Self and it only has mechanical goals related to data processing. If you asked it 'are you alive? why do you want to stay alive?' it would have no feeling-based reasons.