What it is
Machine Learning Street Talk is an interview podcast in which Tim Scarfe, with Keith Duggar, talks to working AI researchers. Its Apple Podcasts description says it covers "current affairs in AI, cognitive science, neuroscience and philosophy of mind" "with the hype surgically removed", and the RSS feed says "removing hype from mainstream narratives". It is free on Apple Podcasts (£0.00 on the GB storefront).
What an episode looks like
Each episode is one guest and one topic, named in the title, for example "Who Checks a Proof No Human Can Read?" with Leo de Moura, or "How Deep Learning Finally Cracked Messy Tables" with Frank Hutter. In the RSS feed, September 2026 episodes ran from 25 minutes to 2 hours 14 minutes. Expect researcher-level discussion of scaling, speech recognition, physical AI, interpretability and AI forecasting rather than product news.
Why this one
What sets it apart from news-style AI shows is the format: a single researcher explaining their own work at length, plus a scope that reaches into cognitive science, neuroscience and philosophy of mind.
Cadence and back catalogue
The RSS feed showed nine episodes dated 2 to 30 September 2026, so it is clearly active, at roughly two a week in that window. Apple Podcasts UK listed 269 episodes on 30 September 2026 (the iTunes lookup without a country parameter returned 268 the same day). The Apple page rated it 4.7 out of 5 from 34 ratings. Apple Podcasts UK page: https://podcasts.apple.com/gb/podcast/machine-learning-street-talk-mlst/id1510472996
Who it suits
Listeners with some technical grounding who want guests explaining their own research at length. In this reviewer's judgement, beginners may find it heavy going. Some episodes carry named sponsors or partnerships, which the feed discloses in episode notes.
