TED: The urgent risks of runaway AI -- and what to do about them | Gary Marcus

The speaker shares their journey in AI, expressing concerns about misinformation and bias in AI systems. They discuss the risks posed by AI, including misinformation, bias, and potential threats like chemical design. The speaker emphasizes the need for a new technical approach and governance to mitigate these risks. They highlight the importance of combining symbolic systems and neural networks to create reliable AI. The talk concludes with a call for global governance and research to manage AI risks effectively.

How does the speaker describe the relationship between symbolic systems and neural networks?

What example does the speaker give to illustrate the need for safety cases in AI?

What does the speaker identify as a significant problem with current AI systems?

What is the speaker's perspective on the future of AI governance?