I=4
Zuckerberg thinks AI bubble is possible, based on past precedence
In a recent interview with Access Podcast, Zuckerberg said he thinks it’s quite possible to have an AI bubble, based on past precedence (min 1:08:43).
He pointed out that if they misspend a couple of hundred billion dollars, it could be very unfortunate. But the risk for not being aggressive enough is higher (min 1:11:00).
On starting to see first glimpses of AI improving itself (as he mentioned some months ago), he gave an example of a Facebook team using Llama 4 to make autonomous agent that improved Facebook algorithm (min 1:15:25)
