Jeremy Howard: fast.ai Deep Learning Courses and Research | Lex Fridman Podcast #35
Lex Fridman (host), Jeremy Howard (guest), Narrator, Narrator
In this episode of Lex Fridman Podcast, featuring Lex Fridman and Jeremy Howard, Jeremy Howard: fast.ai Deep Learning Courses and Research | Lex Fridman Podcast #35 explores jeremy Howard on democratizing deep learning, tools, and real impact Jeremy Howard discusses his path from early programming and music to founding fast.ai, emphasizing a lifelong focus on practical data work and useful tools. He contrasts different programming languages and paradigms, arguing current Python-based deep learning stacks are powerful but fundamentally constrained for real innovation, and outlines why he’s betting on Swift and MLIR-style compilers. A large portion of the conversation centers on making deep learning accessible—through fast.ai courses, better tooling, and cloud setups—so domain experts (especially in areas like medicine) can solve real problems with limited data and compute. He also critiques academic incentives, over-reliance on big data and big compute, and stresses the ethical responsibilities and labor implications of AI.
Jeremy Howard on democratizing deep learning, tools, and real impact
Jeremy Howard discusses his path from early programming and music to founding fast.ai, emphasizing a lifelong focus on practical data work and useful tools. He contrasts different programming languages and paradigms, arguing current Python-based deep learning stacks are powerful but fundamentally constrained for real innovation, and outlines why he’s betting on Swift and MLIR-style compilers. A large portion of the conversation centers on making deep learning accessible—through fast.ai courses, better tooling, and cloud setups—so domain experts (especially in areas like medicine) can solve real problems with limited data and compute. He also critiques academic incentives, over-reliance on big data and big compute, and stresses the ethical responsibilities and labor implications of AI.
Key Takeaways
Practical problems and domain expertise should drive deep learning work.
Howard argues the most valuable AI work comes from domain experts applying deep learning to real problems (e. ...
Get the full analysis with uListen AI
You rarely need “big data” or massive compute to get state-of-the-art results.
Through transfer learning and careful experimentation, fast. ...
Get the full analysis with uListen AI
Current Python/CUDA stacks limit innovation; better compiler and language design are crucial.
Because Python is slow and forces critical loops into low-level CUDA C, many ideas (e. ...
Get the full analysis with uListen AI
Research incentives push academics toward tiny gains on trendy problems, not impactful ideas.
He notes that fields like transfer learning and active learning, which radically reduce data and labeling needs, were largely ignored until someone showed dramatic results, while papers that improve well-studied benchmarks by small margins are rewarded.
Get the full analysis with uListen AI
The fastest path to competence in deep learning is to train many models on your own data.
Howard stresses that learners should quickly fine-tune pretrained models on personally relevant datasets, inspect inputs/outputs and errors, and iterate; this builds intuition far more effectively than reading theory alone.
Get the full analysis with uListen AI
Startups and careers benefit more from tenacity and frugality than from venture capital.
He recommends bootstrapping, keeping costs minimal, charging for small early projects, and avoiding VC pressure to overgrow; persistence and solving a problem you truly understand matter more than big funding.
Get the full analysis with uListen AI
AI practitioners have ethical duties around bias, feedback loops, and human oversight.
Given the leverage of modern models, Howard says data scientists must design for appeals processes, transparency, human-in-the-loop control, and be conscious of broader social impacts like labor displacement and inequality.
Get the full analysis with uListen AI
Notable Quotes
“Most of the research in the deep learning world is a total waste of time.”
— Jeremy Howard
“I think all the major breakthroughs in AI in the next 20 years will be doable on a single GPU.”
— Jeremy Howard
“We don’t need more experts creating slightly evolutionary research in areas that everybody is studying. We need experts using deep learning to solve real problems.”
— Jeremy Howard
“The key differentiator between people that succeed and people that fail is tenacity.”
— Jeremy Howard
“I don’t think we need a lot of new technological breakthroughs to do a lot of great work right now.”
— Jeremy Howard
Questions Answered in This Episode
How would deep learning research priorities change if academic incentives directly rewarded real-world impact over benchmark improvements?
Jeremy Howard discusses his path from early programming and music to founding fast. ...
Get the full analysis with uListen AI
What concrete steps could hospitals and regulators take today to unlock medical data for AI while preserving patient privacy and trust?
Get the full analysis with uListen AI
If Swift + MLIR-based stacks mature, how might that alter who can innovate at the algorithmic level compared to today’s Python/CUDA ecosystem?
Get the full analysis with uListen AI
How should societies prepare for and cushion the labor displacement that Howard believes AI will accelerate, especially for the middle class?
Get the full analysis with uListen AI
For a domain expert with weak coding skills, what is the most effective path to becoming self-sufficient with deep learning tools like fast.ai?
Get the full analysis with uListen AI
Transcript Preview
The following is a conversation with Jeremy Howard. He's the founder of fast.ai, a research institute dedicated to making deep learning more accessible. He's also a distinguished research scientist at the University of San Francisco, a former president of Kaggle, as well as a top-ranking competitor there. And in general, he's a successful entrepreneur, educator, researcher and an inspiring personality in the AI community. When someone asks me, "How do I get started with deep learning?" fast.ai is one of the top places I point them to. It's free. It's easy to get started. It's insightful and accessible. And if I may say so, it has very little BS that can sometimes dilute the value of educational content on popular topics like deep learning. Fast.AI has a focus on practical application of deep learning and hands-on exploration of the cutting edge that is incredibly both accessible to beginners and useful to experts. This is the Artificial Intelligence podcast. If you enjoy it, subscribe on YouTube, give it five stars on iTunes, support it on Patreon, or simply connect with me on Twitter, @lexfridman, spelled F-R-I-D-M-A-N. And now here's my conversation with Jeremy Howard. What's the first program you've ever written?
First program I wrote, that I remember, would be at high school. Um, (sighs) I did an assignment where I decided to try to find out if there were some, like, better musical scales than the normal 12 tone, 12 s- interval scale. So I wrote a program on my Commodore 64 in Basic that searched through other scale sizes to see if it could find one where there were, uh, more accurate, you know, uh, harmonies.
Like mid-tone, like finding-
Like, like you want an actual exactly three to two ratio, whereas with a 12 interval scale it's not exactly three to two, for example. So that's-
And, and the Common-
... well tempered as they say in the... (laughs)
In Basic on a Commodore 64?
Yeah.
Where was the interest in music from? Or is it just technical?
I did music all my life, so I played saxophone and clarinet and piano and guitar and drums and whatever, so...
How does that thread go through your life? Where's music today? Is it-
Uh, it's not where I wish it was. I, for various reasons, couldn't really keep it going, particularly because I had a lot of problems with RSI with my fingers, and so I had to kind of like cut back anything that used hands and fingers.
Mm-hmm.
Um, I hope one day I'll be able to get back to it healthwise.
So there's a love for music underlying it all?
For sure, yeah.
What's your favorite instrument?
Uh, saxophone.
Sax.
Baritone saxophone. Well, probably bass saxophone but they're awkward.
Well, um, I always love it when, uh, music is coupled with programming.
Mm-hmm.
There's something about a brain that utilizes those, that, uh, emerges with creative ideas. So you've used and studied quite a few programming languages.
Install uListen to search the full transcript and get AI-powered insights
Get Full TranscriptGet more from every podcast
AI summaries, searchable transcripts, and fact-checking. Free forever.
Add to Chrome