This is a list of my top recommended links to learn more about the potential species-ending risks of artificial intelligence.

This isn’t a standalone post, but something I link to from other AI-related posts (such as my 33-slide carousel-slide post on LinkedIn explaining the existential-level risk of super-intelligent AI) (audio version of that is here on Spotify and Apple Podcasts).

Intro to AI Safety:

• Intro to AI Safety, Remastered, by Robert Miles
• Why Would AI Want to do Bad Things? Instrumental Convergence, by Robert Miles
• Why Asimov’s Laws of Robotics Don’t Work – Computerphile
• Superintelligence FAQ, by Scott Alexander on LessWrong

AI Safety expert Eliezer Yudkowsky:

• Time Magazine: Pausing AI Developments Isn’t Enough. We Need to Shut it All Down
• Lex Fridman podcast interview: Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization
• Bankless podcast interview: We’re All Gonna Die with Eliezer Yudkowsky
• Discussion of Bankless episode in LessWrong community
• The Hanson-Yudkowsky AI-Foom Debate
• Response to Eliezer Yudkowsky’s Letter in Time Magazine, by Zvi on LessWrong
• Artificial Intelligence as a Positive and Negative Factor in Global Risk (a more detailed technical paper on some of the concepts I discussed, e.g., how exactly an AI might build nano-robots—check out pages 25-28)
• Three Worlds Collide (short fiction about an alien encounter that exemplifies what it might look like for a non-human intelligence to be highly sophisticated and not evil, but with very different morality from us)
• Yudkowsky is also well known for his “AI Boxing” challenges where he plays the role of an AI contained in a secured system and convinces humans to release him into the wild. This can work even if the human knows that the AI will be dangerous. The challenges are meant to show how effectively AI’s might be able to use social engineering to convince us to do things we don’t want to do. Here’s one simple example of how it might work (though Yudkowsky’s challenges have often been much more complex, lasting hours).

AI Safety expert Paul Christiano:

• Bankless podcast interview: How We Prevent the AI’s from Killing us with Paul Christiano
• What failure looks like

Examples of Midjourney (realistic image generation):

• The same prompts one year apart
• What Midjourney thinks professors look like, based on their department
• What if wars were fought with Guitars instead of Guns?

Movies with nuanced depictions of AI:

• Her
• Ex Machina
• A.I. Artificial Intelligence
• Blade Runner 2049
• Ghost in the Shell (the anime original)
• The Matrix
• After Yang
• Westworld (TV show)

Other perspectives I found valuable:

• 2022 Expert Survey on Progress in AI
• Why A.I. Might Not Take Your Job or Supercharge the Economy, Ezra Klein podcast
• What Biden’s Top A.I. Thinker Concluded We Should Do, Ezra Klein podcast with policy expert Alondra Nelson
• A Skeptical Take on the A.I. Revolution, Ezra Klein podcast with professor Gary Marcus
• Who Wins — and Who Loses — in the A.I. Revolution?, Ezra Klein podcast with OpenAI CEO Sam Altman
• Is A.I. the Problem? Or Are We?, Ezra Klein podcast with author Brian Christian
• Why AI Won’t Cause Unemployment, by Marc Andreessen
• AI machines aren’t ‘hallucinating’. But their makers are
• Why is consciousness something worth talking about?, Philosophize This podcast
• AI Takeoff posts on LessWrong
• AI Risk posts on LessWrong