Lots of AI Applications Require High Reliability and That Will Take Longer to Acheive

1min Snip

00:00

Play full episode

Summary

Comments

Transcript

Episode notes

Driverless cars and other AI applications need to be far safer than human drivers in order to gain mass adoption. The tolerance for AI systems making mistakes is significantly lower than for human errors. This demand for high safety standards is one of the reasons why driverless cars are not yet widely available. They need to be extremely reliable, reaching a level of reliability close to perfection. The same applies to other fields like healthcare, where AI may assist doctors but only if it can be trusted to catch and correct errors. Currently, there is no clear path to training AI models without a significant number of errors, which makes the widespread adoption of AI in high and low-stakes professions unlikely for now.

Typically when we put out a call for audience questions, there’s no single topic that dominates. This time was different. The questions we received were overwhelmingly focused on artificial intelligence: Do A.I. systems pose an existential threat to humanity? Will robots take our jobs? How could these machines potentially make our lives — and the lives of our children — better?

So I asked the show’s senior editor, Roge Karma, to join me to talk through them. We also discuss my mixed feelings about the calls to “pause” A.I. development, why I’m less worried about rogue A.I. systems than the incentives of the companies and countries developing A.I., the need for a “public vision” for A.I. development, whether A.I. companions can help address widespread loneliness, why I’m skeptical that A.I. advances will lead to skyrocketing economic productivity, the possibility that A.I. advances will lead to a post-work utopia, why I think of A.I. less as a normal technology and more as a “hyper object,” what A.I. systems are unveiling about what it means to be human and more.

Mentioned:

“Natural Selection Favors AIs over Humans” by Dan Hendrycks

“2022 Expert Survey on Progress in AI”

God, Human, Animal, Machine by Meghan O’Gieblyn

“Resisting dehumanization in the age of A.I.” with Emily Bender

“The Moral Economy of High-Tech Modernism” by Henry Farrell and Marion Fourcade

Recommendations:

“Some of Us Are Brave” by Danielle Ponder

“In Memory of a Honeybee” by Felix Rösch

“Clouds” by Felix Rösch and Laura Masotto

“Driven” by Felix Rösch

Mabe Fratti

Trance Frendz by Ólafur Arnalds and Nils Frahm

Thoughts? Guest suggestions? Email us at ezrakleinshow@nytimes.com.

You can find transcripts (posted midday) and more episodes of “The Ezra Klein Show” at nytimes.com/ezra-klein-podcast, and you can find Ezra on Twitter @ezraklein. Book recommendations from all our guests are listed at https://www.nytimes.com/article/ezra-klein-show-book-recs.

This episode of “The Ezra Klein Show” is produced by Roge Karma, Kristin Lin and Jeff Geld. Fact-checking by Michelle Harris and Kate Sinclair. Mixing by Jeff Geld. Original music by Isaac Jones. Audience strategy by Shannon Busta. The executive producer of New York Times Opinion Audio is Annie-Rose Strasser. Special thanks to Sonia Herrero and Kristina Samulewski.