home

Poor man’s strategy for the age of AI

To get somewhere in life, one must aim ahead as if shooting at a moving target.

My (not particularly original) guesstimate about where things are headed:

  1. Even if the development of artificial intelligence slowed down or stopped, what we already have is enough to cause profound changes in society.

  2. Apps like ChatGPT will follow a trajectory similar to AIs specialized to play games - towards capabilities surpassing the best of humans.

So, it might be advantageous to understand AI. This also applies to people like me, who didn’t study computer science or math, but social sciences.

There are plenty of self-study AI plans on the internet. They usually approach it systematically - math first (algebra, trigonometry, geometry, probability, calculus), then computer science, machine learning, etc. Here is such a plan generated by GPT-4.

The systematic approach sounds sensible. I’d definitely recommend it to someone with a lot of free time. However, this is roughly the knowledge volume of a college degree, and I, unfortunately, have my student years behind me. I don’t have the time.

I have to to tackle it in a less than ideal way, backwards. I will try to solve some problems the AI way, starting with the simplest way I can think of. When I get stuck, I will learn that specific thing needed to continue. If something puzzles me, I will write a post about it and give the experts on the internet the pleasure of pointing out the gaps in my knowledge.

I’ll be like a unit of paratroopers airdropped behind enemy lines for a reconnaissance mission. These guys do not know what’s waiting for them. They just have to look around and take it from there.

The advantage of the ‘recon mission’ approach to a field of study is that you can find something useful fast if you get lucky. The downside is that you are likely to make foolish mistakes. Which is a lot better than what can happen to real scouts behind enemy lines. I hate to feel like a moron, but it’s survivable. And if my predictions about AI come true, this is a feeling that will be useful to get used to.

I looked around (on Hacker News, where else?) to find a decent starting point:


That should be enough for now.