How Afraid of the AI Apocalypse Should We Be? | The Ezra Klein Show

How Afraid of the AI Apocalypse Should We Be? | The Ezra Klein Show
Yudkowsky is a pioneer of A.I. safety research, who started warning about the existential risks of the technology decades ago, – influencing a lot of leading figures in the field. But over the last couple of years, talk of an A.I. apocalypse has become a little passé. Many of the people Yudkowsky influenced have gone on to work for A.I. companies, and those companies are racing ahead to build the superintelligent systems Yudkowsky thought humans should never create. But Yudkowsky is still out there sounding the alarm. He has a new book out, co-written with Nate Soares, “If Anyone Builds It, Everyone Dies,” trying to warn the world before it’s too late.

So what does Yudkowsky see that most of us don’t? What makes him so certain? And why does he think he hasn’t been able to persuade more people?

0:00 Intro
2:45 A.I. is more alien than it seems
10:00 Danger signs are here already
16:00 “Alignment faking”
19:45 A.I. can jump out of the system
25:27 A.I. language
27:30 Does A.I. “want” things?
30:00 A.I. is like natural selection
42:00 Why would A.I. go totally rogue?
46:30 His epiphany of doom
52:00 The A.I. industry racing ahead
59:59 Build the off switch