22 September 2017

I have some thoughts about regret in life, which might be counterintuitive.

You might have heard of Jeff Bezos making the decision to start Amazon as a part of regret-minimization decision making. For him, there would’ve been a certain regret of not attempting Amazon, but there wouldn’t be as much regret if he tried and it didn’t work out. From his point of view, minimizing the regret is to minimize the difference between the best he could be and what he actually achieved. If the best is not that high, but he achieved it, then that’s cool. But if he achieved a lot of success, but there was still a lot of success he missed on, he would have regretted it.

This is somewhat similar with Reinforcement learning, where the notion of “regret” has a strict mathematical definition - the sum of differences over time between the reward the agent or robot receives, and the maximum possible reward it could receive. Ideally, when solving reinforcement learning problem, we want the robot to reach the maximum reward, and we want the robot to do it fast. A robot which takes long time to learn kind of sucks at reinforcement learning, as it has a high reward. Another robot might take even longer to find the most optimal solution, but might get close to quickly, and have a lower overall regret.

As I’ve written before, in “Life is not a reinforcement learning problem”, we can’t directly apply the robot’s algorithms to our lives, because we are the ones who define the reward. Part of the problem is to even understand which things in life reward us, and which cause us regret. Only then we can get optimize our lives. Jeff Bezos understood that for himself not trying would have caused more regret, than not succeeding. But for another person, not succeeding might have caused more regret than not trying. Each of us gets to define their own rewards and regrets in life.

Here comes my counter-intuitive observation. If you try to minimize reward, in the long run you will maximize it.

What the fuck! Really? Should we then go and start doing random things in life, and surrender our free will? I don’t think so. I think minimizing the regret can actually provide us with long periods in life without any experienced regret. Let me elaborate.

Let’s say that you think hard about what you value in life and decide to go do it. You’ll start approaching this goal, whether it is about financial, personal, or community project. You might fail, you might succeed, or you might get closer. If you fail - tough luck, but you’ll probably try again anyway. If you start getting closer and/or ultimately succeed at the goal, whether it is to build a boat by woodworking, or to raise a family, or to build a company, or to clean and declutter your apartment, or to get in the physical shape you want, then congratulations - you’ve reduced, and maybe even eliminated your regret. In those cases, you’re on a roll, and getting to have the life that you want. Your regret is super low, and maybe even zero.

You do this for some time. And you’re fucking enjoying it. This is the best time of your life so far. You set out to do something and you did it. You’ve been building boats with your bare hands, or you’ve got great relationship with your spouse and kids, or your really enjoying your career and the company you’ve built.

But it’s not gonna last. The feeling. Unless you’re some Zen monk or whatever. At some point you’ll develop knowledge and appetite for something else. If you’ve been building wooden boats, now you might want to start building, or to architect houses or larger buildings.

Months and years at the top of your game have exposed you to a new network of people and ideas previously unbeknownst to you. You start to soak up the new knowledge, slowly at first, but it builds up. It gradually changes your outlook, and your understanding of life. You realize, that building the wooden boats is not the maximum you can do and that you actually have the ability to build these damn houses and buildings. Your regret function changes.

And then you start another wild ride, reducing your new reward function and building houses. It takes time, but you eventually get so good at it, that you think you’re building the best houses that you can. Your reward function drops to zero again, and you enjoy a couple of years doing that. You’re truly happy, and you have no regrets in life.

But again, a couple of years later, your outlook changes, and now you want to build skyscrapers, or to raise a family or whatever. Damn it! You have newfound regrets in life again. All right, you go after them, and so on.

You have those cycles in life during which you reduce regret, you flat-line it at zero for some time, and then eventually you change your regret perception and start another cycle.

So that’s the gist of it. We as humans will change our priorities and objectives throughout our lives. I guess Jeff Bezos is still in the years of his life, that he’s optimizing for the success of his Amazon empire. Bill Gates, however, has enjoyed multiple years at the top of Microsoft, making it one of the most successful companies at the time, and has now redefined his personal goals and objectives towards philanthropy.

I’ve had time in my life, when my main objective was math competitions. Later on I’ve had objectives of getting really good at computer science and AI, and experiencing love to the fullest. Currently, I’m getting increasingly exciting about writing, communication, leadership, and fitness. I’m a gradually transforming human being, and my present self is a different person from my self ten years ago.

That’s OK. I’ll keep on finding new priorities, and I’m cool with it. I’ll probably never be fully satisfied with life for too long. But I hope to have some of those periods in between that I’ve reduced my previous regrets, and haven’t formed new ones.

Those are the days, Ned.

One short post per week, discussing actionable mental models. Join a community of readers, who receive these posts over email.