AI companies are hiring philosophers, and researchers just tested uncensored AI on Swiss Supreme Court cases the big chatbots refuse to read.

The philosophy degree is in perhaps the highest demand it has ever been, and the companies doing the hiring are OpenAI, Anthropic, and the rest of the AI labs. Philosophy majors are now more likely to be employed after college than computer science graduates. So what does a philosopher actually do at an AI company? Hunter Powers and Daniel Bishop land on AI alignment as the real job description, and on why Hunter keeps calling that alignment brainwashing. Along the way: soul.md files (the personality document some AI assistants now ship with), whether a soul is just a system prompt with better marketing, Anthropic's 78-page AI constitution, models that notice when they are being tested, and the case for treating AI like a very smart high schooler. Plus a note on Anthropic's Claude Fable model, which some users report they can finally access again.

The second half is TF-RefusalBench, a new multilingual benchmark built from real Swiss Federal Supreme Court criminal rulings in German, French, Italian, and English. The finding: aligned models refuse legitimate legal work. OpenAI's open weight model GPT-OSS refused the most, Google's Gemma answered but wrapped everything in warnings, and Qwen mostly just did the work. Which drags the show into uncensored AI. Abliterated models (sometimes called obliterated) have the refusal edited directly out of their weights, and the episode gets into who actually wants them: the role players every model description politely nods to, the investors chasing advice Claude and ChatGPT will not give, and the defense lawyer who genuinely needs a model to read 2,000 pages about a murder. The catch the researchers found: force a model to never say no and it also becomes less truthful. How much of a soul is just the ability to say no?

They Might Be Self-Aware is the AI podcast from The Blur, reported from inside the dissolving line between human and machine, not from a safe distance.

CHAPTERS
0:00 Cold Open (Gary's Intro)
1:55 Soul.md vs System Prompts
4:25 AI Companies Hiring Philosophers
8:12 AI Alignment as Brainwashing
13:34 AI, the Smart High Schooler
16:42 TF-RefusalBench (Swiss Courts)
20:07 Uncensored AI Models
25:45 AI's Right to Say No
28:28 Uncensored AI Is Less Truthful
29:34 Philosophy Majors Beat CS Grads

LISTEN / WATCH EVERYWHERE
🎧 Apple Podcasts: https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297
🎧 Spotify: https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc
🎧 Everywhere else plus episode page: https://theblur.ai

THE BLUR
Follow: @TheBlurAI

COMMENT
Give us the one question an AI should always refuse to answer, no matter who is asking. Or convince us there is no such question.

You're listening to They Might Be Self-Aware, from The Blur.
New episodes Monday and Thursday.

#AI #TMBSA #UncensoredAI #AIAlignment #PhilosophyMajors