Coffee Space


What Is Thinking?

Preview Image

I recently came across a clip by Jordan Peterson that got me thinking:

It kind of rubbed me up the wrong way, as I have spent a long time thinking about exactly this. His claim is 1:

Don’t be thinking that you think, because thinking is very hard. And most people can’t think at all, and even if you can think, you’re not very good at it. I mean, think about what you have to do technically to think. First of all, you have to formulate the problem. That’s hard enough - and you have to formulate it precisely. Then, you have to generate multiple potential solutions to the problem, and then you have to let those solutions argue themselves into a hierarchy internally. And so you have to be able to tolerate that stress, right? Toy can’t just be one thing if you’re going to think - because thinking isn’t just saying that what I think is right. That’s not thinking at all. Thinking is questioning whether or not what you think is right is right, and that’s really hard. It’s very demanding.

Where Does Thinking Happen?

Let’s begin by abstracting the problem. Let’s abstract outside of the idea of a human, as animals are also capable of thinking. Let’s abstract further, to the idea of an agent RR 2. An agent acts AA, within a world WW, where it can sense SS. This looks like the following (unrolled):

..wtstrtatwt+1st+1.. .. \to w_{t} \to s_{t} \to r_{t} \to a_{t} \to w_{t+1} \to s_{t+1} \to ..

tt here represents a discrete time step (which humans arguably may not act under), but the principles are the same. Given some observation from our senses, we process this information (with memory and models MM from previous observations) and take some action.

What Is Thinking?

If you have the ability to update some model MM, such that observations from your sense SS can influence your actions AA, to meaningfully work towards some objective in the future t+nt + n (which we won’t define here), in my book you are thinking. It could be simple thought, but it is thought.

I believe that the ‘thought’ that Jordan Peterson speaks of is a specific type of ‘thought’, where based on some model(s), you evaluate how a set of actions 𝐚=[a0,..,an]\textbf{a} = [ a_0, .., a_n ] will impact your future objective at t+nt + n. He believes that such a type of thought is really costly.

To the contrary, it can be quite cheap. The simplest method of cause is to brute force all possible actions 𝐀\textbf{A}, and pick one with the best reward for your given objective. Of course, nobody really does this.

How Do We Think?

Typically, rather than check all possible action sequences 𝐀\textbf{A}, we take some shortcuts. Some of these could look like the following:

How do we know this can lead to high levels of thought? Well, a lot of chess playing algorithms will have similar methods to concentrate their processing on possibilities that are likely to yield high rewards. Similarly, a lot of the speed-up in neural networks has been seen by using the limited processing power more wisely.

Is Thinking Stressful Or Difficult?

I tend to disagree. The brain has evolved to have very efficient ways to make decisions with little cost in both resources and time. When walking around, we do not think about each step we make, considering all possible futures, desperately trying to avoid falling over and trying to reach our objective. Walking is relatively easy, we have tonnes of shortcuts available to our thinking process that mean it is very cheap to walk.

What is stressful on the other hand is decision making, the difference being that you have highlighted this particular choice of actions as significant. Suddenly, rather than relying on an almost “auto-pilot” action selection process, there is some potential cost to making the wrong decision 3.

To make decisions, you need to be aware of:

  1. There is more than one action available to you. Mostly people will be unaware that there was even a choice to be made. Retroactively, we may say “well, you took that choice”, but actually they just failed to realise there was a choice at all.
  2. That there is some difference in outcomes depending on which you take. You may be aware that there is a choice, but not know how they affect your future goals positively or negatively.
  3. Have the ability to make future predictions about how each action will affect your future. If you cannot figure out which action is better past some minimum accuracy, you are essentially making a random choice anyway, as your prediction model is too noisy. The only benefit here is that a failure to make a correct prediction here means that you can at least learn from it.
  4. Be capable of taking the action you would like to select. It may be difficult to take, especially if you took too long in processing it. You also may have incorrectly evaluated that the choice was available to you.


In conclusion, Jordan Peterson can be thought provoking on some topics, but I believe that his word salad should not be accepted without some critical thought. I believe he often dazzles people with his ability to jump around topics and play with them, losing many people in the process, only to pick them up at the end with the reward of applause.

It somewhat reminds me of arguing with religious people, where they can quote from obscure passages of religious text. They can really dance around to make their point, but of course the problem being, if you follow them, it simply doesn’t make sense. They rely on long-form arguments where they lose most people in the depths of their own domain.

In my opinion, what made one of the great thinkers of our time, the late Christopher Hitchens, so great - was his ability to chase these arguments all the way through. I highly suggest to watch him during his debates, he is what Jordan Peterson is trying to emulate, but cannot.

That’s all for now.

  1. This is my transcription from the video, some of the words may be misrepresented.↩︎

  2. Technically the RR here is for robot, but this is the context I approach the problem from.↩︎

  3. We’ll ignore the fact that control is mostly an illusion to trick ourselves into attempting agency.↩︎