An AI uses blackmail to save itself, and threats make AIs work better

In episode 53 of The AI Fix, our hosts suspect the CEO of Duolingo has been kidnapped by an AI, Sergey Brin says AIs work better if you threaten them with physical violence, Graham wonders how you put a collar on a headless robot dog, Mark asks why kickboxing robots wear head guards, and the CEO of Anthropic says AI could wipe out entry-level jobs.

Graham asks your favourite AI how it feels about being kidnapped, and Mark explains how an AI tried to save itself by blackmailing the engineer responsible for turning it off.

Episode links:

• Duolingo CEO walks back AI memo.
• Large Language Models, Small Labor Market Effects.
• Humanoid robots fight in Chinese kick-boxing competition.
• Anthropic CEO warns AI could eliminate half of all entry-level white-collar jobs.
• ‘Dystopian’ video shows robot dog scaring real dog in NYC: ‘That thing is possessed’.
• Sam Altman: Tens of millions of dollars well spent--you never know.
• ChatGPT burns tens of millions of Softbank dollars listening to you thanking it.
• Sergey Brin: All AI models tend to do better if you threaten them.
• System Card: Claude Opus 4 & Claude Sonnet 4.

The AI Fix

The AI Fix podcast is presented by Graham Cluley and Mark Stockley.

Grab T-shirts, hoodies, mugs and other goodies in our online store.

Learn more about the podcast at theaifix.show, and follow us on Bluesky at @theaifix.show.

Never miss another episode by following us in your favourite podcast app. It's free!

Like to give us some feedback or sponsor the podcast? Get in touch.

Support the show and gain access to ad-free episodes by becoming a supporter: Join The AI Fix Plus!

Our Sponsors:
* Check out Anthropic: https://claude.ai/aifix


Advertising Inquiries: https://redcircle.com/brands

Privacy & Opt-Out: https://redcircle.com/privacy