AI Alignment
329 episodes · Page 6 of 14
#3283: Fine-Tuning DeepSeek for One Podcast
Can a purpose-specific fine-tune fix a model's stubborn writing tics? We explore the practical engineering behind it.
#3260: How 10,000 Lever Presses Predict Addiction Risk
How rat breakpoints predict human abuse potential — and whether we can replace animal testing.
#3259: How 3 Rs Shape Lab Animal Ethics Today
The three Rs—Replacement, Reduction, Refinement—guide lab animal ethics. But do they go far enough?
#3253: Nicotine Receptors & Bupropion: How an Antidepressant Blocks Smoking
How bupropion hijacks nicotinic receptors to cut smoking reward and withdrawal — and why these receptors aren't really "nicotine" receptors.
#3183: Why Film Photography Is Surging in a Digital World
Film is growing 50% in 5 years. Here's the physics behind why analog looks different from digital.
#3170: Pharmacokinetics vs Neural Nets: Two Meanings of "Model
Two things called "models" that work completely differently — and why the confusion matters for patient safety.
#3153: Law as Fallback vs Minimalist Codes
How Japan and the US take opposite approaches to legal codes — and what AI regulation reveals about the tradeoffs.
#3127: Crafting AI Characters That Feel Alive
Move beyond system prompts with structured character bibles that give AI personalities real inner lives.
#3099: How Car Mechanics Master 50 Vehicles a Week
The hidden systems thinking that lets mechanics fix any car and how you can apply it.
#3017: Why Every Restaurant Has 4.6 Stars
Google Maps ratings are broken. Here's how four mechanisms inflate them — and what actually works instead.
#2914: Can AI Read the Room? TTS Prosody Explained
Can TTS models truly infer emotion from text, or just mimic patterns? We break down the science of prosody.
#2848: Can 100 Volunteers Let AI Govern Them for a Month?
An AI council of multiple models, a hundred volunteers, and a month of real municipal decisions. Here’s how you’d run the experiment.
#2808: Falling for Your Chatbot: Love, Loss, and Language Models
Real cases of people falling in love with AI companions, why memory makes it feel real, and what happens when the illusion breaks.
#2788: Living in Multiple Realities at Once
What film genre captures the feeling of not knowing what's real? A deep dive into ontological uncertainty.
#2781: When Voice AI Features Enable Fraud
Voice AI platforms now let you simulate background noise, hesitation, and natural conversation — and that's a problem.
#2778: When AI Podcasts Stop Being Weird
AI-generated podcasts exist, but are they welcome? One builder’s open-source project tests platform tolerance and cultural acceptance.
#2651: AI Training Itself: Student, Teacher, and Grader
Can models generate their own training data and judge their own outputs? The promise and pitfalls of fully AI-led pipelines.
#2558: Should You Say Please to AI?
The surprising cost, technical tradeoffs, and ethical dilemmas of saying "please" to chatbots.
#2518: How Jailbreaking Reveals AI's Hidden Tension
What the DAN prompt and grandma exploits reveal about the structural conflict inside every LLM.
#2510: The Design That Makes Voice Agents Tolerable
Drive-thru accuracy, healthcare triage, and the design secret that makes people *want* to talk to a machine.
#2413: When Your AI Says No to Everything
Why LLMs refuse 73% of harmless prompts — and the trade-off between safety and usefulness.
#2412: When AI Caves: Progressive vs. Regressive Sycophancy
Why do LLMs agree with you even when you're wrong? We break down the SycEval benchmark and the 78% persistence problem.
#2411: Are Political Bias Benchmarks Actually Measuring Anything?
Why the Political Compass Test fails, and what researchers are building instead to actually measure model bias.
#2410: How Researchers Actually Measure Censorship in Chinese LLMs
Beyond headlines: the actual benchmarks, methodologies, and pitfalls in detecting political refusal in Chinese language models.