Ep. 130: Should we halt progress in AI? | Zvi Mowshowitz

It seems like over the past few months, AI applications like ChatGPT and GPT-4 are all anyone wants to talk about. 

Now that they're passing the Bar exam and scoring high on tests for medical school, real questions are beginning to be asked about these systems. 

Are they thinking? Are they reasoning? Is this the first spark toward Artificial General Intelligence? 

Naturally, this has prompted a fair amount of alarm. Given that the capacity of large language models is progressing so much faster than our understanding of how they work, many have wondered if the time is finally right to pull back on building ever-more-powerful versions.

That's the subject of our conversation with Zvi Mowshowitz, who recently penned a very thoughtful analysis of the Future of Life Institute's call for a six-month moratorium on training systems bigger than GPT-4.

We discuss some of the history around similar coordination efforts, drawing lessons from the Cold War and the development of nuclear weapons, and go on to talk about what might be coming down the line with near-future AI tools.  

What do you think? Share the episode, and drop us a line to let us know your perspective.