Practical Advice for Getting Started and Becoming an Expert in Deep Learning with fast.ai

This podcast clip features Jeremy Howard offering concise, actionable advice for individuals looking to enter or advance in deep learning. The core recommendation is to "train lots of models," emphasizing hands-on experimentation over passive learning. He highlights the fast.ai course, which recently won the CogX award for best AI course, as a free and effective resource. The course teaches practical skills like creating custom datasets from scratch using tools like Google Image Search scripting and deploying web applications, enabling students to build functional deep learning models quickly.

A crucial distinction made is between merely running models for inference and actively training or fine-tuning them. Howard stresses that true learning and utility come from fine-tuning models with one's own data, which can take as little as five minutes. This process allows individuals to adapt existing models to their specific domain areas, making the technology personally relevant and powerful. He cites examples of students achieving state-of-the-art results or creating unique classifiers for niche interests, demonstrating the accessibility and impact of this practical approach.

For those aspiring to become deep learning experts, Howard advises against solely pursuing "evolutionary research" in crowded academic areas. Instead, he advocates for becoming an expert in applying deep learning to a specific "passion area" or real-world problem. Examples include diagnosing malaria, analyzing media bias, or studying fisheries. This approach leverages an individual's existing domain expertise, allowing them to identify meaningful problems and evaluate the effectiveness of their deep learning solutions more accurately.

The broader implication is a call for problem-driven innovation in AI. Howard argues that the most interesting and valuable research stems from efforts to solve actual problems effectively. By combining deep learning tools with deep domain knowledge, individuals can create solutions that are not only technically sound but also genuinely useful and impactful, fostering a more diverse and applied landscape of AI expertise beyond traditional academic research. The fast.ai platform, being entirely free, serves as a significant enabler for this democratized, practical approach to AI education and application.

train lots of models that's that's how you that's how you learn it

come to our course cost up faster day I and the thing I keep on harping on in my lessons is train models print out the inputs to the models print out to the outputs to the models like study you know change change the inputs a bit look at how the outputs vary just run lots of experiments to get a you know an intuitive understanding of what's going on to get hooked

you've got to find true in the models so that's that's that's the critical thing because at that point you now have a model that's in your domain area

it only takes five minutes to fine-tune a model for the data you care about

somebody said oh I tried looking at different gary characters couldn't believe it the thing that came out was more accurate than the best academic paper after Lesson one

Everything we do is free we have no revenue sources of any kind it's just a service to the community

We need experts at using deep learning to diagnose malaria well we need experts at using deep learning to analyze language to study media bias

become the expert in your passion area and this is a tool which you can use just about anything and you'll be able to do that thing better than other people particularly by combining it with your passion and domain expertise

if you're not working on a real problem that you understand how do you know if you're doing it any good

the vast majority of interesting research is like try and solve an actual problem and solve it really well