Invaluable advice for Machine Learning beginners

Andrej Karpathy was the director of artificial intelligence at Tesla and currently works for OpenAI. If you’re tired of self-proclaimed AI experts promising you a 2000% increase in your productivity if you add these 15 AI tools to your workflow, he would be a refreshing and reliable resource. Here’s what he had to say when asked for the best advice to machine learning beginners :

1 — Put in the work :

Many beginners often focus on what to do, but the focus should be more on how much you do. I am a believer, on a high-level, of the ten thousand hour concept, where you just have to just pick the things you’re interested in, and you literally have to put in ten thousand hours of work. It doesn’t matter as much where you put it, you’ll iterate and you’ll improve and you’ll waste some time, but I don’t know if there’s a better way, you need to put in the time.

• So the first thing I would focus on is : are you spending ten thousand hours ?

2 — Don’t get discouraged by more advanced people :

One of the biggest obstacles that machine learning beginners face is the feeling of inadequacy compared to more advanced people in the field. It’s easy to get discouraged when you see someone who seems to know everything about machine learning. The only person you should compare yourself to, is yourself from some time ago.

• Are you better better than you were a year ago ? Then you can see your progress, and that’s very motivating.

3 — Teache :

One thing I like about teaching is that it strenghtens your understanding, it’s not a purely altuiristic activity, it’s a way to learn. If you have to explain something to someone, you realize you have gaps in knowledge.

• Did you just learn something new ? The best way to see if you truly understand it, is by trying to explain it to someone, no matter how simple it is.

Sounds a lot like machine learning, aren’t we machines, learning, too ?

What he says can be applied not only to machine learning, but to almost any subject. Do you agree with his point of view, or do you believe that there are ways around the system, to get there faster ?

The interview is from his pidcast with Lex Fridman, you can find it here.