Topics / Artificial Intelligence
Artificial Intelligence Podcasts
513 shows and 86,995 episodes sit under Artificial Intelligence.
An editorial survey of how the top AI podcasts are framing agents, infrastructure, alignment, and the race between labs. Updated weekly.
Photo: Brecht Corbeel on Unsplash · Unsplash License
Swipe · Artificial Intelligence only
Swipe through Artificial Intelligence
Save time discovering your next favourite podcast by listening to just the best minute of an episode.
The artificial intelligence slate · every cover a link
All 513 artificial intelligence shows →
Read this first
Picking your next Artificial Intelligence podcast
Some listeners want the lab race by Friday afternoon. Others would rather spend an hour on one hard question about intelligence, watch an engineer pull apart an agent stack, or hear what happens when AI leaves the demo and enters a workplace. Decide on your listening pace first: briefing, interview, technical walkthrough, or reported narrative. On the Artificial Intelligence shelf, that choice matters more than familiarity with any particular show.
If your question extends beyond machine learning, step up to Technology. Move sideways to Cybersecurity when security comes first, or to Software Engineering when the craft matters beyond AI. Stay here when models, agents, their builders, and their consequences are the reason you pressed play.
The week in models
ThursdAI - The top AI news from the past week moves like a live group chat among people who already know the model names. Launches, benchmarks, prices, APIs, and rumors arrive quickly, with jokes and corrections left in. Its compressed titles tell you exactly what sort of listening commitment this is: 📆 ThursdAI - Aug 21 - DeepSeek V3.1’s hybrid upset, ByteDance’s 512K Seed-OSS, Nano Banana wizardry, Agents.md standardizes agents, and more AI barely stops for breath. The equally crowded 📆 ThursdAI – Jul 31, 2025 – Qwen’s Small Models Go Big, StepFun’s Multimodal Leap, GLM-4.5’s Chart Crimes, and Runway’s Mind‑Bending Video Edits + GPT-5 soon? offers a fast entrance to Multimodal AI.
That density makes ThursdAI - The top AI news from the past week a useful launch point for Frontier Models and Open Source vs Closed AI. Follow the Chinese model names through its episode list and the route toward the US-China AI Race becomes apparent too. This is listening for people who enjoy being slightly behind the conversation and catching up at speed.
FT Tech Tonic slows the tempo. Its reported style begins inside a scene, asks short questions, and lets the significance emerge without excited narration. The run from The battle of the AI labs through AI Labs: Are Anthropic really the good guys?, AI Labs: Google DeepMind plans its comeback, and AI Labs: Sam Altman may make or break OpenAI treats the labs as institutions with distinct pressures rather than a parade of release notes. Head from there into Frontier Models or Alignment & Safety, depending on whether the competition or its consequences hold your attention.
Power, policy, and ordinary life
Decoder with Nilay Patel is built around tech-business interviews conducted as thesis tests. The host’s long questions gather several premises, then land on the point that needs defending. Restatement is part of the method: an answer gets translated into sharper language and handed back for inspection.
The episode list moves easily between company strategy and public resistance. Does Google even want to win in AI? sits near The left and right agree on one thing: no data centers, while What Apple’s OpenAI lawsuit is really about shares the run with Dr. Jill Lepore on why the AI backlash is vital for the future. Choose AI Regulation when the legal and institutional questions pull hardest. Choose AI in Society when the argument is really about people living with the systems.
Possible sounds more like a thoughtful boardroom conversation. One host frames a fresh development; the other answers in long, qualified passes, making room for upside, caveat, and practical stakes. That rhythm works especially well when the subject is a person already using AI in a specific setting. Inside an AI-powered personal dashboard | Matthew Tiemann, How Sougwen Chung teaches robots to pause, and The film director winning awards with AI | Ben Hansford form a route through applied work without turning the discussion into a tool roundup. The natural continuation is AI in Society.
Then there is Shell Game, which has the patience and dry timing of a long-form magazine story. Small scenes and recorded moments do much of the work; the host returns with a quiet observation rather than a grand verdict. Titles such as The One-Human Unicorn, Episode 1: Minimum Viable Company, and Episode 4: The Startup Chronicles signal a sustained company-building arc. Take this route when a reported narrative sounds better than another panel debating abstractions.
Minds, models, and research papers
For AI Reasoning, Brain Inspired offers the deepest conceptual on-ramp. These are leisurely academic interviews about cognition, consciousness, biological brains, and their relationship to machines. The language assumes a listener comfortable with technical terms and willing to let one thread unfold for a long time. BI 240 Cristopher Moore: Cognition and Computational Complexity and BI 233 Tom Griffiths: The Laws of Thought are good tests of whether that pace suits you. If they do, expect exploration rather than quick conclusions.
Data Skeptic is friendlier to the listener who wants technical substance with more handrails. The host knows the vocabulary but frequently translates a difficult idea into a familiar analogy, then asks the guest to correct it. Its recommender-systems run is unusually focused: Social Choice for Fair Recommendations, Give Users the Wheel, Disentanglement and Interpretability in Recommender Systems, and Healthy Friction in Job Recommender Systems. That sequence can lead toward AI Reasoning, but it also opens a practical route into AI in Society through questions of choice, fairness, and control.
Daily Paper Cast works at paper speed. Two synthetic-conversational voices alternate between explanation, reformulation, and confirmation, with little room for digression. With technical vocabulary left intact, the show suits a listener seeking distilled methods and metrics rather than a broad introduction. AI Agents is the obvious next stop after PILOT in the Loop: Live Self-Improvement for Long-Horizon Agents and What Makes Good Agentic Data? An ACE Lens on Data Generation for LLM Agents. For visual systems, continue from VGI-Bench: Probing Visual Intelligence in Video Generation Models into Multimodal AI.
From prototype to production
The InfoQ Podcast speaks engineer to engineer. Guests are allowed to qualify, digress, and tell war stories; the host translates a claim into plainer language before pushing on its trade-offs. The show assumes that terms such as context window and MCP need little ceremony.
Its episode titles draw a clean line between coding assistance and the surrounding engineering practice. Start with From MCP and Vibe Coding to Harness Engineering: How Did AI Native Engineering Evolve in One Year, then compare it with The Human Edge: Why Brownfield Codebases Need Mob Programming, Not Just AI Vibes. Coding & Dev Tools is the narrower route. When architecture, teams, and existing systems count equally with the tool itself, the better next stop is Software Engineering with AI. Strands Agents with Clare Liguori leads directly toward AI Agents, while Governance in the Age of AI: A Conversation with Sarah Wells points back toward organizational control.
No Priors: Artificial Intelligence | Technology | Startups approaches the same world from architecture and company strategy. Its two hosts arrive prepared, skip elementary introductions, and let technically fluent guests take long turns. Rethinking Legacy Data Infrastructure with Eon Co-Founders Ofir Ehrlich and Gonen Stein and Re-engineering the Semiconductor Supply Chain with Intel CEO Lip-Bu Tan make sensible entrances to AI Infrastructure. For the economics around ambitious companies, try Chasing Trillion-Dollar Companies, Founder Ambition, Token Budgets, and Regulatory Capture with Sarah & Elad, then continue to AI Startup Funding. Really Big Test-Time Compute in AI Changes Benchmarks, Safety and Research with OpenAI Research Scientist Noam Brown sits at the meeting point of Frontier Models, AI Reasoning, and Alignment & Safety.
Where to start
For a fast weekly orientation, begin with ThursdAI - The top AI news from the past week. For skeptical interviews examining business power and public consequences, turn to Decoder with Nilay Patel. Pick Brain Inspired for a long, demanding conversation about minds and intelligence. If you build software, start with The InfoQ Podcast and follow whichever thread wins: agents, tools, governance, or production engineering.
Featured shows below are part of the PlayNext catalog. Updated weekly.
Featured shows
10 shows, described rather than listed
A short write-up for each, plus the job it does in this guide.
Co-hosts Elad Gil and Sarah Guo treat AI like a system to be reasoned about, not a press-release subject. Recent episodes range from coding agents with Andrej Karpathy to materials-science labs to the financial plumbing under the agentic economy — short on hype, long on what actually has to be true.
Nilay Patel's interviews live where AI strategy collides with media, regulation, and product reality. Decoder takes the kind of question other tech shows answer in five minutes and stretches it across an hour, hunting for the structural answer instead of the headline-friendly one.
Lex Fridman's long-form interviews approach AI as both a technical artifact and a civilizational question, with conversations stretching past three hours when a guest takes him somewhere unexpected. The result is more art-house than industry-tour: scientists, founders, and contrarians who get the kind of patient runway most podcasts won't risk.
Reid Hoffman and Aria Finger ask what the upside actually looks like if AI breaks humanity's way. The conversations skew analytical — incentives, network effects, attack surfaces, labor transitions — and the recent run has been preoccupied with grid bottlenecks, autonomous payments, and what happens to SaaS when agents write the software themselves.
Paul Middlebrooks runs Brain Inspired as a long-form interview show about cognition, consciousness, and the strange overlap between brains and learning machines. The default mood is patient and warm; researchers get to follow a thread to its end without the clock running on them.
Data Skeptic comes from the position that algorithms deserve the same critical scrutiny as any other claim. Recent runs have circled recommender-system fairness, interpretability, and the social trade-offs smuggled in when a model decides what you see next.
InfoQ talks to working senior engineers and architects, not press-tour evangelists. The host's job is translation: restating dense answers in plainer terms, then pushing for the trade-off the guest just glossed over — especially as agentic systems leak into real production stacks.
Evan Ratliff turns AI agents into reporting subjects rather than abstractions. Shell Game watches what happens when voice-clone agents are unleashed on banks, exes, and customer-service hotlines — first-person, dry, and increasingly necessary listening.
Tech Tonic is the Financial Times in podcast form: scene-led narrative reporting on the AI economy, anchored to specific cases — model-as-companion, the race to Mars-adjacent capital, a death investigation that ends up pulling in alignment debates.
Naval Ravikant's show plays closer to a recorded notebook than a traditional interview. Episodes drift between technology, capital allocation, and personal philosophy in a way that treats AI less as a topic and more as the latest pressure on every other system worth thinking about.
Where to start
4 routes in, one per reader
Brain Inspired runs three-hour conversations with researchers who can actually take a question apart.
No Priors keeps the math under the conversation; the hosts treat agent claims as engineering problems, not magic.
Lex Fridman gives guests room to follow a question to its strange ending, especially around consciousness, agency, and what counts as understanding.
Shell Game is field reporting from inside the voice-clone era — funny, unnerving, ahead of where most coverage is.
Glossary
Artificial Intelligence vocabulary, decoded
An AI system that takes actions rather than just generating text — making API calls, browsing the web, executing code, or running multi-step workflows on its own.
Reinforcement Learning from Human Feedback. The training step where humans rate model outputs to teach a model what helpful, honest, and harmless answers look like.
The handful of largest, most capable AI models at any given moment — currently GPT, Claude, and Gemini families. The term implies they push the boundary of capability and risk.
The research problem of making sure an AI system pursues goals its operators actually want, instead of optimizing a proxy that breaks down at scale.
A test designed to measure a specific model capability — math, code, reasoning, refusal behavior. Public evals leak into training data, which is part of why benchmarks decay.
Running an already-trained model to produce an output. Most of the energy and cost questions in AI right now are inference questions, not training questions.