In 30 seconds
Jeremy Howard recommends hands-on experiments, fine-tuning and combining deep learning with domain knowledge. Course examples and timing reflect the 2019 conversation.
- 1Run experimentsTrain models and inspect their behavior.
- 2Work with dataBuild, clean and inspect a relevant dataset.
- 3Choose a domainDomain knowledge helps evaluate usefulness and errors.
Takeaways
- Train models and inspect inputs and outputs.
- Use data connected to a problem you care about.
- Domain knowledge helps interpret errors.
Read the edited text by chapter
Learn by training
Jeremy Howard advises beginners in deep learning to train models and run many experiments rather than only read about them.
Jeremy Howard advises beginners in deep learning to train models and run many experiments rather than only read about them.
He recommends inspecting inputs and outputs, changing the inputs and observing how results vary to build intuition.
Howard argues for fine-tuning a model on data from a domain the learner cares about. The time and resources required vary with the task, data and hardware.
Build and inspect a dataset
He describes a fast.ai course example in which learners build an image dataset, clean examples and create a small classification application. Data collection and reuse require appropriate permissions.
He describes a fast.ai course example in which learners build an image dataset, clean examples and create a small classification application. Data collection and reuse require appropriate permissions.
He shares examples of student projects and emphasizes inspecting mistakes, not only reporting accuracy.
Develop domain expertise
Asked how to become an expert over several years, Howard again recommends training models in a domain that matters to the learner.
Asked how to become an expert over several years, Howard again recommends training models in a domain that matters to the learner.
He gives examples such as medical analysis, language and fisheries, arguing that domain expertise helps identify useful questions and evaluate results.
Fridman and Howard agree that working on a real problem helps researchers understand whether results are useful and why errors occur.
More from this channel
3:34:27EN
Lex Fridman· Oct 26, 2026Podcast · 4 speakersAI, Mathematics, and the Meaning of a ProofGuest Steven Brunton, Ankur Moitra, Alex Kontorovich · Host Lex FridmanLex Fridman and three mathematicians discuss recent AI-generated research, hard open problems and what human understanding contributes.
3:36:56ENLex Fridman· Sep 16, 2026PodcastPsychiatry Has Never Found Its PenicillinHost Lex FridmanHistorian Andrew Scull tells Lex Fridman why the field can ease symptoms, has never cured a major mental illness, and keeps paying for overconfidence with patients' lives.
5:15:51ENLex Fridman· Aug 26, 2026Podcast · 2 speakersDHH Did Not Write a Line of Omarchy QuattroGuest DHH · Host Lex FridmanThirteen months after a skeptical conversation with Lex Fridman, the Rails creator describes how Opus 4.5, a swarm of coding agents, and a one-minute Linux installer ended his era of hand-chiseled Ruby.
3:12:58ENLex Fridman· Aug 12, 2026Podcast · 2 speakersOne Cold Shower, Twenty Champions: How Khabib Nurmagomedov Was ForgedGuest Khabib Nurmagomedov · Host Lex FridmanIn a Makhachkala school gym, seventy men shared one shower, and the water was only cold.
3:46:18ENLex Fridman· Jul 28, 2026Podcast · 2 speakersIf You Take Slavery Out of the Picture, There Is No Civil WarGuest Gary Gallagher · Host Lex FridmanIn 1860, enslaved people were worth more than every railroad and factory in the United States combined.
3:51:47ENLex Fridman· Jun 30, 2026Podcast · 2 speakersForty-Six Percent of Emperors Were Overthrown. The Empire Lasted AnywayGuest Anthony Kaldellis · Host Lex FridmanHistorian Anthony Kaldellis walks Lex Fridman through 2,200 years of Roman history and argues that what we call Byzantium was Rome itself.