Sycophancy in AI: Risks and Realities

Welcome back to The FAIK Files!

In this week's episode:

• We explore the One Million Chessboards project, a massive collaborative web experiment where users can move pieces across a million shared chessboards in real-time
• Anthropic's model welfare research program, AI ethics, and the need for interpretability
• OpenAI's recent struggle with ChatGPT's personality crisis as they roll back an update that made the AI too sycophantic
• Meta's troubling chatbot sex problem: Social Media, LLMS, sex, and Zuckerberg -- what could go wrong?

Check out The Deception Project to learn about our upcoming Offensive Cyber Deception Masterclass and more.

Also check out Perry's new newsletter, Deceptive Minds: a newsletter about how we are fooled, how we fool ourselves, and what we can do about it. Subscribe on LinkedIn https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7319922626200510464

Want to leave us a voicemail? Here's the magic link to do just that: https://sayhi.chat/FAIK

You can also join our Discord server here: ⁠⁠https://faik.to/discord⁠

***** NOTES AND REFERENCES *****

ONE MILLION CHESS BOARDS:

• One Million Chessboards website: https://onemillionchessboards.com/
• Creator's blog post explaining the project: https://eieio.games/blog/one-million-chessboards/
• Nolen Royalty's previous viral project, One Million Checkboxes: https://corecursive.com/one-million-checkboxes-with-nolen-royalty/
• AI Cheating at Chess - Time Magazine report: https://time.com/7259395/ai-chess-cheating-palisade-research/

ANTHROPIC'S MODEL WELFARE RESEARCH:

• Exploring Model Welfare - Anthropic's research announcement: https://www.anthropic.com/research/exploring-model-welfare
• YouTube interview, "Could AI Models be Conscious?": https://youtu.be/pyXouxa0WnY?si=1yK4YkMbE5iW9SC0
• Dario Amodei on interpretability: https://x.com/DarioAmodei/status/1915515160607023391
• The Urgency of Interpretability (Dario's blog): https://www.darioamodei.com/post/the-urgency-of-interpretability
• Axios report on AI sentience research: https://www.axios.com/2025/04/29/anthropic-ai-sentient-rights

THE PERSONALITY CRISIS OF CHATGPT:

• OpenAI rolls back sycophantic ChatGPT update: https://arstechnica.com/ai/2025/04/openai-rolls-back-update-that-made-chatgpt-a-sycophantic-mess/
• Sam Altman's tweet about the issue: https://xcancel.com/sama/status/1917291637962858735
• Stanford HAI research on LLM personality: https://hai.stanford.edu/news/large-language-models-just-want-to-be-liked

META'S CHATBOT SEX PROBLEM:

• Wall Street Journal investigation: https://www.wsj.com/tech/ai/meta-ai-chatbots-sex-a25311bf

Want to connect with us? Here's how:

Connect with Perry:

• Perry on LinkedIn: https://www.linkedin.com/in/perrycarpenter
• Perry on X: https://x.com/perrycarpenter
• Perry on BlueSky: https://bsky.app/profile/perrycarpenter.bsky.social

Connect with Mason:

• Mason on LinkedIn: https://www.linkedin.com/in/mason-amadeus-a853a7242/
• Mason on BlueSky: https://bsky.app/profile/wickedinterest.ing

Learn more about your ad choices. Visit megaphone.fm/adchoices