Thread by @NGKabra

If a student in high-school is keen on a career in AI/ML, how should they prepare for it? What subjects should they learn, and what activities should they participate in?

What basic subjects should you be strong in if you're interested in a long-term career in AI/ML? The replies to the parent tweet show that there are many different opinions.

Here's a thread on the slow and steady method, which focuses on building foundations for the long term.

3. You should have done well in maths and statistics in school. Nothing fancy, no multivariate calculus. Just basic probability and statistics and basic calculus.

You don't have to be a star student in math, and you don't have to be a topper. Just enjoying the subject is enough

4. It is important that you understand maths from the first principles. You don't need to be a fast at problem solving or calculations. And you don't need to memorize formulae. But you need to understand how the formulae were derived and where to apply them.

5. The other thing that can help at high-school level, is basic programming. Any language will do (though Python is particularly good). And it doesn't have to be any AI/ML specific programming. Just having fun with programming is good enough.

6. It is not necessary to do AI/ML at school level (or even undergraduate). You can do a robotics course, but only if you're enjoying it. The focus should be on fun, not on understanding the ML algorithms (at this stage).

7. If you're still a student, and keen on a career in AI/ML, this should be the plan. Notice how actual AI/ML shows up only during Masters. Before that the focus is on foundational maths and CS courses

8. During undergraduate, the focus should still be on foundational courses in maths and computer science/programming but which build towards AI/ML. Here's a list

9. The more of these you do, the better equipped you'll be to do good work in AI/ML in the long term. Specific algorithms (e.g. CNNs, RNNs, reinforcement learning) come and go, specific tools (e.g. TensorFlow) come and go, but the fundamentals don't change.

10. Does this mean that you're doomed if you did not start the process in high-school? No. All these courses can easily be done online. Just make a list of which areas you're weak in, and start from the basics.

11. But keep in mind that directly jumping to an ML course without good foundations will give you an illusion of learning, but you'll be cargo-culting your way through it. Things will work but you don't know why, and when they break, you won't be able to fix them easily.

12. This thread is based on a talk given recently by Dr Ashwin Rao, Vice President of AI at Target, and Adjunct Professor at Stanford. Here's a video of the full talk:

13. I've summarized the important points from the talk (in more detail than this thread) on my substack. See: futureiq.substack.com/p/training-for…

14. @pratik05101974 asks a good question: aren't we going a little overboard wanting our children to start worrying about AI/ML in school itself? Where will it stop?

https://twitter.com/pratik05101974/status/1404723769843675136

15. Should nurseries teach Neural Networks for Babies?
amazon.com/Neural-Network…

16. Certainly not. *If* your child is interested in AI/ML and actually enjoys maths or programming or robotics, then this is advice for them. And there is a small percentage of students like this.

But it is completely fine to *not* be interested in this.

17. Does every child have to learn AI/ML to survive in the coming future where software is eating the world? Certainly not.

I have written at length about how Liberal Arts is a great choice for some children:

https://twitter.com/NGKabra/status/1227878183132729347

One interesting thing about building strong foundations in maths and computer science is that you can approach these subjects from many angles (if you don't focus too much on marks in school exams):

https://twitter.com/MJDeshpande/status/1404728115599790081

Learning to spot and use patterns is one of the most important abilities for doing well in life. We don't pay enough attention to this in school.

@MJDeshpande suggested this (previous tweet).

In addition, @_lavita has 2 tweets with great suggestions

https://twitter.com/_lavita/status/1404785174815023110

"Students should construct surveys, conduct them, then analyze/visualize them to find interesting patterns". This is the second time this week that someone has suggested that to me. The first was a senior statistics prof. Do we do enough with surveys in our schools?

Continuing with "what to study to build AI/ML chops": a few people responded saying that the math is not important, and students should just directly jump into ML algorithms.

But the number of people recommending maths/linear algebra/statistics is far higher.

You do need the math, says @h_rapa Chief Data Scientist at @netcore_cloud.

"For a long time I kidded myself thinking I can “wing” it without understanding the math. It’s only after I studied the math, I could understand the “why” part of the job."

https://twitter.com/h_rapa/status/1404746592305827846

Unfortunately, most schools do a bad job of teaching maths. They make it boring and stressful, instead of challenging fun.

By contrast, studying for Olympiads is a better idea. Also, websites like artofproblemsolving.com

https://twitter.com/AmanHasNoName_2/status/1404847739167395841

Exam-oriented studies with prescribed textbooks kill the joy of learning. Many of the ML practitioners in the responses said that kids should get involved in open-ended/exploratory tinkering/programming activities.

https://twitter.com/nehalmax/status/1406846001323778050

But, with the tinkering/programming activities, make sure the kids are having *fun*. It shouldn't become another stressful activity. Read the responses to the first tweet in this thread and see how many times the word "fun" appears.

It is important for the students to have exposure to non STEM subjects.
For example:

https://twitter.com/NarenDhyani/status/1404747461923536900

and

https://twitter.com/vishalkmumbai/status/1404753217540411398

Exposure to non-STEM doesn't have to be through subjects taught in school (which are often done so badly as to bring students to tears).

Instead they can read the right books:
See:

https://twitter.com/dileep31/status/1404726476222566402

and

https://twitter.com/KrAsh0x35/status/1404733949570469892