8th Layer Insights

The FAIK Files | Holiday Special: AI Safety Update

Dec 27, 2024

40:24

Note: We're posting Perry's new show, "The FAIK Files", to this feed through the end of 2024. This will give you a chance to get a feel for the new show and subscribe to the new feed if you want to keep following in 2025.

Welcome back to the show that keeps you informed on all things artificial intelligence and natural nonsense.

In our holiday episode, Mason opens a rather unique Christmas present from Perry, we invite a special guest to help explain the infamous "Paperclip Maximizer" thought experiment, and we discuss an interesting (and somewhat disturbing) new AI Safety paper from Apollo Research.

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://discord.gg/cThqEnMhJz

*** NOTES AND REFERENCES ***

An interesting cluster of new AI safety research papers:

Apollo research: Frontier Models are Capable of In-context Scheming (Dec 5, 2024)
YouTube Video: Apollo Research - AI Models Are Capable Of In Context Scheming Dec 2024
YouTube Video: Cognitive Revolution - Emergency Pod: o1 Schemes Against Users, with Alexander Meinke from Apollo Research
OpenAI o1 System Card (Dec 5, 2024)
Anthropic: Alignment Faking in Large Language Models (Dec 18, 2024)
Anthropic: Sycophancy to subterfuge: Investigating reward tampering in language models (June 17, 2024)
Fudan University: Frontier AI systems have surpassed the self-replicating red line (Dec 9, 2024)

Other Interesting Bits:

The Paperclip Maximizer thought experiment explanation
Theory of Instrumental Convergence
iPhone Game: Universal Paperclips
VoxEU: AI and the paperclip problem
Real Paperclips! 500 Pack Paper Clips (assorted sizes)

OpenAI Announces New o3 Reasoning Model:

OpenAI's "12 Days of Ship-mas" announcement page
YouTube video: OpenAI's announcement of their o3 Model
TechCrunch: OpenAI announces new o3 models
Wired: OpenAI Upgrades Its Smartest AI Model With Improved Reasoning Skills
TechCrunch: OpenAI trained o1 and o3 to ‘think’ about its safety policy
Matthew Berman YouTube video: OpenAI Unveils o3! AGI ACHIEVED!
NewScientist: OpenAI's o3 model aced a test of AI reasoning – but it's still not AGI
Yahoo Finance: OpenAI considers AGI clause removal for Microsoft investment

*** THE BOILERPLATE ***

About The FAIK Files:

The FAIK Files is an offshoot project from Perry Carpenter's most recent book, FAIK: A Practical Guide to Living in a World of Deepfakes, Disinformation, and AI-Generated Deceptions.