Jeremy Howard is the founder of fast.ai, a research institute dedicated to making deep learning more accessible, and a former president of Kaggle. The items below are things Howard said or wrote about deep learning, learning, research and AI policy, in interviews on the Lex Fridman, Machine Learning Street Talk, Latent Space and Network State podcasts and at a PyTorch Conference fireside chat. — Lex Fridman Podcast #35: Jeremy Howard: fast.ai Deep Learning Courses and Research (Aug 2019), YouTube captions.

Part 1: The Practice of Deep Learning
- On Treating Fine-Tuning as a Continuum: On the Latent Space podcast, Howard said that we should "treat it as a continuum, and we should have much higher expectations of how much you can do with an already trained model". — Latent Space: AI Magic: Shipping 1000s of successful products with no managers and a team of 12, Jeremy Howard of Answer.ai (Aug 2024), transcript.
- On Money and the World: On the Latent Space podcast, Howard said that "99.999% of the world is driven very heavily by money, especially huge amounts of money". — Latent Space: AI Magic: Shipping 1000s of successful products with no managers and a team of 12, Jeremy Howard of Answer.ai (Aug 2024), transcript.
- On Long-Term Value Principles: On the Latent Space podcast, Howard said of Answer.AI’s second round: "we had everybody invest through a long-term SPV, which we set up where everybody had to agree to vote in line with long-term value principles". — Latent Space: AI Magic: Shipping 1000s of successful products with no managers and a team of 12, Jeremy Howard of Answer.ai (Aug 2024), transcript.
- On the People Worth Working With: On the Latent Space podcast, Howard said of the people Howard tends to hire: "they're often like risk-takers, they're often creative, they're often extremely tenacious, they're often very open-minded". — Latent Space: AI Magic: Shipping 1000s of successful products with no managers and a team of 12, Jeremy Howard of Answer.ai (Aug 2024), transcript.
- On Not Being Better at Everything: On the Latent Space podcast, Howard said that "it's not reasonable to expect you're going to be better at everything than everybody else". — Latent Space: AI Magic: Shipping 1000s of successful products with no managers and a team of 12, Jeremy Howard of Answer.ai (Aug 2024), transcript.
- On a Large Yard with Narrow Fences: On the Latent Space podcast, Howard said that, in the R&D historian Eric Gilliam’s description, successful labs have "a large yard with narrow fences". — Latent Space: AI Magic: Shipping 1000s of successful products with no managers and a team of 12, Jeremy Howard of Answer.ai (Aug 2024), transcript.
- On removing prestige: The fast.ai home page is titled "Making neural nets uncool again". — fast.ai home page.
- On Code as Expression: On the Latent Space podcast, Howard said: "I think you express a lot in the code you choose to write and how you choose to write it". — Latent Space: AI Magic: Shipping 1000s of successful products with no managers and a team of 12, Jeremy Howard of Answer.ai (Aug 2024), transcript.
- On Reading the Code: On the Latent Space podcast, Howard said that "the only really, really way to understand it properly is again, just read the code and step through the code". — Latent Space: AI Magic: Shipping 1000s of successful products with no managers and a team of 12, Jeremy Howard of Answer.ai (Aug 2024), transcript.
- On Human-Friendly Tools: At a PyTorch Conference fireside chat, Howard said that "TensorFlow never felt very human-friendly to me". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
Part 2: Data Science and Kaggle Strategy
- On Hundreds of Hours with PyTorch: At a PyTorch Conference fireside chat, Howard said: "when I saw what Soumith and Adam had done with PyTorch, we spent hundreds and hundreds of hours working with it and we were just like, this is so much more human-friendly". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On Ideas Taking Time: At a PyTorch Conference fireside chat, Howard said: "So it takes a long time for an idea to percolate.". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On ChatGPT and the Three Stages: At a PyTorch Conference fireside chat, Howard said that "it wasn't until ChatGPT came out that used that exact approach, the exact same three-stage approach". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On AI as a Source of Power: At a PyTorch Conference fireside chat, Howard said that "AI is now a source of power in the world, and it's becoming an increasingly important source of power". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On AI in the Hands of the Powerful: At a PyTorch Conference fireside chat, Howard said that "if we let AI just be in the hands of rich and powerful people, the best AI, it could literally tear apart democracy at its foundations". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On Recommitting to a Principle: At a PyTorch Conference fireside chat, Howard said that "we all have to recommit to this principle that, okay, some people will use this technology to do bad, but most people will use it to do good". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On Getting Faster: At a PyTorch Conference fireside chat, Howard said: "When software engineering is going well, you should be getting faster and faster and things should be getting easier and easier.". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On Agents Doing Your Work: At a PyTorch Conference fireside chat, Howard said that "I would be very, very careful of spending too much time trying to get agents to do your work for you". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On Losing Competence: At a PyTorch Conference fireside chat, Howard said: "I've seen people becoming just depressed that they're no longer competent and they are no longer in control". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On Using AI to Improve: At a PyTorch Conference fireside chat, Howard said: "I'm a much better developer than I was two years ago because I'm all about using AI to help me get better". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
Part 3: Learning How to Learn
- On Skills Growing Faster: At a PyTorch Conference fireside chat, Howard said that, as AI gets better, "it's more and more important that you are, that your skills are growing faster than the AI's skills". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On George Pólya: At a PyTorch Conference fireside chat, Howard said that Solve It is based on "principles from a 1945 math book by a guy called George Pólya". — PyTorch Conference fireside chat with Jeremy Howard and Anna Tong, transcript.
- On pedagogical sequencing: On the fast.ai course page, Howard wrote: "I start by showing how to use a complete, working, very usable, state-of-the-art deep learning network to solve real-world problems, using simple, expressive tools.". — fast.ai: Practical Deep Learning for Coders.
- On Over-Resourced Labs: On the Network State Podcast, Howard said that "these big labs suffer from being overresourced". — Network State Podcast: Jeremy Howard on Fast.ai (Jul 2025), transcript.
- On Being Rewarded for Compute: On the Network State Podcast, Howard said that at Google "you were explicitly rewarded for using more compute". — Network State Podcast: Jeremy Howard on Fast.ai (Jul 2025), transcript.
- On Not Every Kid Learning at the Same Speed: On the Network State Podcast, Howard said that "not every kid learns everything at the same speed". — Network State Podcast: Jeremy Howard on Fast.ai (Jul 2025), transcript.
- On Reaching Out: On the Network State Podcast, Howard said that "we really reached out to parts of India and Africa and stuff that had nothing". — Network State Podcast: Jeremy Howard on Fast.ai (Jul 2025), transcript.
Part 4: Medical AI and Real-World Applications
Part 5: Open Source and the AI Balance of Power
- On Believing in Humanity: In the essay "AI Safety and the Age of Dislightenment," Howard wrote: "The alternative to craving the safety and certainty of control and centralization is to once again take the risk we took hundreds of years ago: the risk of believing in the power and good of humanity and society". — fast.ai: AI Safety and the Age of Dislightenment (Jeremy Howard, 2023).
- On Not Rushing to Regulate: In the essay "AI Safety and the Age of Dislightenment," Howard wrote that "we should not be in a rush to regulate this technology, and should be careful to avoid a cure which is worse than the disease". — fast.ai: AI Safety and the Age of Dislightenment (Jeremy Howard, 2023).
- On the Age of Dislightenment: In the essay "AI Safety and the Age of Dislightenment," Howard wrote: "If we regulate now in a way that increases centralisation of power in the name of “safety”, we risk rolling back the gains made from the Age of Enlightenment, and instead entering a new age: the Age of Dislightenment". — fast.ai: AI Safety and the Age of Dislightenment (Jeremy Howard, 2023).
- On Licensing and Surveillance: In the abstract of "AI Safety and the Age of Dislightenment," Howard wrote: "Proposals for stringent AI model licensing and surveillance will likely be ineffective or counterproductive, concentrating power in unsustainable ways". — fast.ai: AI Safety and the Age of Dislightenment (Jeremy Howard, 2023).
- On a Feudal Future: In the essay "AI Safety and the Age of Dislightenment," Howard wrote: "What if, in our attempts to avoid an AI apocalypse, we centralize control of the world’s most powerful technology, dooming future society to a return to a feudal state in which the most valuable commodity, compute, is owned by an elite few". — fast.ai: AI Safety and the Age of Dislightenment (Jeremy Howard, 2023).
- On Broad Participation: In the essay "AI Safety and the Age of Dislightenment," Howard wrote: "Broad participation could allow more people with a wider variety of expertise to help identify and counter threats, thus increasing overall safety". — fast.ai: AI Safety and the Age of Dislightenment (Jeremy Howard, 2023).
Part 6: Artificial General Intelligence and True Risks
- On the actual threat of AI: "The real risk of AI is how it concentrates power." — Source: fast.ai Blog
- On Centralizing Power: On Machine Learning Street Talk in 2026, of giving all the power to rich companies or the government, Howard said "that's the worst possible thing you can do because you've centralized the ability to control in one place". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
- On Software Engineering: On Machine Learning Street Talk in 2026, of LLMs, Howard said: "They're really bad at software engineering". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
- On Inhumane Ways of Working: On Machine Learning Street Talk in 2026, of traditional software engineering, Howard said: "I think it's inhumane and I just don't think it works very well". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
Part 7: Productivity, Tools, and the Cost of Vibe Coding
- On Never Using a Debugger: On Machine Learning Street Talk in 2026, Howard said: "I basically never have to use a debugger because I basically never have bugs". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
- On Small Steps: On Machine Learning Street Talk in 2026, of how software gets built, Howard said that it builds up in small steps and "each step works and I can see it working". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
- On Slot Machines: On Machine Learning Street Talk in 2026, Howard said: "The thing about AI based coding is that it's like a slot machine". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
- On Cosplaying Understanding: On Machine Learning Street Talk in 2026, Howard said: "LLM's cosplay understanding things" and "They pretend to understand things". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
- On a Path to Obsolescence: On Machine Learning Street Talk in 2026, of focusing only on AI output, Howard said "it's basically a path to obsolescence through a company and the people who are in it". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
- On Productivity Claims: On Machine Learning Street Talk in 2026, Howard said: "No one's actually creating 50 times more high-quality software than they were before". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
- On Muscles: On Machine Learning Street Talk in 2026, of AI coding, Howard said that "when you're not actively using your design and engineering and coding muscles, you don't grow". — Machine Learning Street Talk: "Vibe Coding is a Slot Machine" - Jeremy Howard (Mar 2026), captions.
Part 8: The Reality of Academic Research and The Future
- On academic incentives: On the Lex Fridman Podcast, Howard said: "most of the research in the deep mining world is a total waste of time". — Lex Fridman Podcast #35: Jeremy Howard: fast.ai Deep Learning Courses and Research (Aug 2019), YouTube captions.
- On the power of transfer learning: On the Lex Fridman Podcast, Howard said: "I only want to teach people practical stuff and I think the only practical stuff is transfer learning". — Lex Fridman Podcast #35: Jeremy Howard: fast.ai Deep Learning Courses and Research (Aug 2019), YouTube captions.
- On ignoring boundaries: On the Lex Fridman Podcast, Howard said: "I couldn't find any examples of transfer learning and NLP so I just did it". — Lex Fridman Podcast #35: Jeremy Howard: fast.ai Deep Learning Courses and Research (Aug 2019), YouTube captions.
- On practical reinvention: On the Lex Fridman Podcast, Howard said: "everybody kind of reinvents active learning when they actually have to work in practice because they start labeling things and they think gosh this is taking a long time and it's very expensive". — Lex Fridman Podcast #35: Jeremy Howard: fast.ai Deep Learning Courses and Research (Aug 2019), YouTube captions.