Learning Artificial Intelligence

I'm learning artificial intelligence (AI), and I thought it might be helpful for others if I document my process. (Of course, it might not.) This is just the start, and I hope to write more along the way to form a series of articles.

Observations

As I was thinking about this article, I made three observations.

First, this is so meta. I'm learning artificial intelligence, where people are trying to learn how humans learn so that they can make machines learn like us. While computers can calculate things much faster than us, there are basic things that we do, such as identifying objects correctly, that computers struggle to do. Because of this, it feels useful to document and understand how I'm learning AI and how I learn in general. There might be things that turn out to be helpful insights in the future.

Second, when I first started, I was looking for the best, or at least a good, way of learning AI. This meant I have made an assumption that we had discovered the best ways of learning (which by the way sounds like rule-based learning in AI). It's probably not too farfetched to think that way, given there are many really smart people who are knowledgeable about AI. But as the AlphaGo vs Lee Sedol episode has shown how there are interesting moves and strategies that even the very best has not thought of, there might be even better ways of learning AI than what we know about today. The chances of me, someone with no prior experience, discovering an even better way is low but if more people continue to try different ways, we might eventually discover better strategies after many tries. This is one of the reasons AI attracted me—that computers can try many things much faster than we humans can and likely discover better ways of doing things faster than we can.

Third, algorithms deal with well-defined problems with a well-defined goal, for example, to win a Go match. I don't really have a well-defined goal. I want to learn AI because I think it's fascinating and has so much potential for us. Maybe I should set a well-defined goal. Or maybe in the process of learning, I figure out the or a goal. Life certainly feels like the latter, where we slowly discover our goals or purpose as we experience things. Such philosophical discussions are a big part of AI, which is another thing I find attractive about it.

My learning process so far

Anyway, back to my learning process. It has been less than a week when I got serious about this, so there isn't much to share yet. Here are a few things that I found helpful:

• The AlphaGo movie - In 2016, AlphaGo, an AI developed by DeepMind, went against the best Go player, Lee Sedol, who at that point had won 18 world championships. Director Greg Kohs did an amazing job telling the story through the movie and got me excited.
• T-Minus AI - Author Michael Kanaan gave a great overview of the evolution and the state of AI. The book isn't technical but Michael Kanaan did briefly introduce several technical concepts, which helped me feel more comfortable when I encountered those concepts elsewhere later. It feels like a good book for beginners like me who are trying to understand the general picture of AI as it is now.
• David Silver on Lex Fridman Podcast - I personally like to learn about the people in the field, on top of the technical stuff, perhaps because it puts a face to the ideas and helps me remember better. David Silver was the lead researcher for AlphaGo. He did a Ph.D. in reinforcement learning, which, together with supervised learning and unsupervised learning, forms the three main machine learning systems. Reinforcement learning was the approach used by AlphaGo.
• Introduction to Algorithms - I picked up this 1,000+ page textbook thinking it'll help me learn the basics. A friend who is knowledgeable about AI said it has great information but isn't a prerequisite to learning AI. I was planning to try and read the entire textbook but I'm now undecided. I might still try or at least go through a set of lectures on it.
• Friends - While this is last on the list, it's probably the most important. The friend mentioned above spent four hours on a Saturday to give me a broad introduction to AI, which I found so helpful. The best thing is that I could ask him questions and get a quick response, rather than trying to search for the answers online myself when I don't even know the right words to use to search at the moment. Another friend, a mathematician, has been helping me with mathematics. This is also why I want to write about my journey on my blog. I hope I can find more like-minded people to learn together.

That's all for now. I'll try to document my learnings as much as I can so that I can remember better, and I'll link them from here. Here we go!