Will hard work matter?

In February this year, I started an attempt to automate my investing. I used to spend my nights falling asleep to earnings calls and taking the bus would mean reading financial statements. Effectively, doing homework was something that was all consuming. Within an hour, AI can do more work than I could have done in years. Occasionally, I will still do what I used to do, but now the act of performing equity analysis is purely for enjoyment and curiosity. These days, I am mostly auditing the output of my AI system, looking for questions it’s been asking, ways it has been attempting to answer them, and so on. I can say with confidence, the system I have will invest better than I could. 

One way to view values is to see it as a framework of how to navigate life. One value that has been useful for all of the history of mankind is the value of hard work. When the wheat needed to be harvested or the spreadsheet needed to be filled out, a human needed to act. Today, even with AI, there is an abundant need for human knowledge work. AI has deflated the cost of so much work but humans are still required to audit, judge, and provide the frameworks that AI uses. However, in 20 years, when my son joins the workforce, I wonder if that will still be true. 

AI is excellent at following an algorithm/framework. If you ask an AI system to research a management team’s incentives, background, childhood, etc. it can do a much better job than a human. But when you ask AI to create that algorithm/framework, it does a poor job. I tend to think about the performance of AI as a function of the problem space it needs to live in. If the problem space is near infinite (ie. an ambiguous problem or a problem where the solution cannot be easily measured), AI will tend to be lacking compared to humans. And vice versa, if the problem is well defined, AI does well. At least that’s the story today. 

We still have wars, famine, children dying, and we don’t yet have humans living on Mars. Clearly, there is still an infinite demand for labour and solutions. What’s less clear to me is that a human needs to be the one labouring or implementing those solutions. At least in my lifetime, the future has never looked less uncertain. But there’s no uncertainty that our lives will revolve around AI. We don’t know if we’ll need human labour forever, but to the degree that we do, it’s likely going to be in areas where AI is insufficient. AI can do work. It cannot be creative nor does it deal well with ambiguity. That’s the story today. Who knows if that’ll be true tomorrow. 

The world has playbooks on how to think. Operating systems, if you will. For example, there’s a Chinese operating system for childhood education. Work intensely hard, set the bar intensely high. You have 4 options in life: doctor, lawyer, engineer, or failure. Those playbooks were formed in a time where the assumptions were different. Many of those assumptions are probably not true anymore. 

Time to rewrite those playbooks.