Learning AI in ten years

I recently stumbled upon an incredible functional computer science curriculum [1], and saw the value in documenting my progress. In this ever-growing, on-the-fly, and infinite game article, I’ll give my roadmap of what I’m going through, went through, what worked for me, what didn’t, and hopefully encourage some others who are captivated by artificial intelligence, but have been intimidated by the technical barrier to entry.

The goal is not to get a layman into fluency, but actually someone into being able to pose good research questions and make progress on them.

This is a lofty goal, but right now it’s my focus. My advice to you, just as much to myself is this: Pace yourself. This is a marathon. You don’t need to have all this down to be able to do things. Each item helps. Knowledge and productivity are like compound interest [2]. Each piece of human representation learning, intuition, chunking, trick, or method is a benefit to you that keeps on giving back to you over and over. If you get too frustrated, take a break from it, and manage your psychology such that you’ll come back to it eventually and in a longer timeframe you will have done more.

Curriculum

Pattern Recognition and Machine Learning (PRML) Bishop book

• Chapter 1-3 reading, implementation, and problems in the margins.
• Follow the outline from http://www.cs.toronto.edu/~rsalakhu/sta4273_2013/
• Watch the lectures http://www.fields.utoronto.ca/video-archive/event/323/2014

Footnotes

[1] I’m following a bit of this - the original guy is ridiculous in his learning drive and progress. It’s obvious from his post and discussions he’s been in the game for a while. A lot of this post follows his structure.

[2] Check out all of Richard Hamming’s “Learning how to learn”. The meta-insight and utility density-per-unit of the book and lectures is just asymmetrical.

[3] Also inspired by Norvig’s “Teach yourself programming in ten years”