Types of smart people

One of the many wonders of the internet is that it allows one to listen to a bunch of smart people talk. On Youtube, there are hundreds of thousands of hours of smart people from a wide range of disciplines talking about their ideas in an even wider range of fields.

When you hear a mathematician speak, versus a tech CEO speak, the way they think about problems is very different (obviously). But I think there’s an even clearer distinctions between the two groups. Without defining it too clearly, for now, let’s just call them Type A smart people and Type B smart people. I bring this distinction to light because I think one type of intelligence is much better understood conventionally, while the other type is harder to define, and is sometimes only visible in outcomes.

1. For example, I’ve known some Type A smart people who had impressive performances in high school Olympiads or early internship experience but struggled to find jobs or opportunities later, while many relatively average classmates land big tech offers. Could this just be variance?

2. You also see a lot of engineers who don’t seem very technically impressive (like with competitive programming or GitHub projects), but they nevertheless find a lot of success, such as being promoted to senior or staff very quickly or landing roles at quant firms. Be it nepotism, luck, or other factors, maybe you attribute it to some form of luck and dismiss their unusual success as statistical variance.

General, Abstract Definition

To put a name to the two categories, the distinction between the two is almost that of problem-focused (such as being fast at mental math or being good at chess) and direction-focused (identifying opportunities and knowing what matters).

Strategies for the former, Type A, are often taught to us at a young age — in school, we are taught deliberate practice and repetition when drilling the times tables or calculus. It seems like Type A smart people excel at tasks with clear rules and a fast feedback loop. When I was younger, I would watch Big Bang Theory every Wednesday at 7:00, and Sheldon Cooper was my idea of the “Form of Intelligence”, at the time.

Type A questions are like: “How do I solve this algorithms question?” or “What is the best move to make in this chess position?”

Type B people fall into more of a “meta” category, but I wouldn’t classify it as meta-learning in the sense of finding mnemonic strategies or using spaced repetition. Type B people don’t learn things faster; they are just better at identifying and working on the things that matter. So less so about raw intelligence, and more so about strategy. This is also why I think Type B intelligence can be learned.

The Type B questions are along the lines of “How do I get a promotion?” or “How do I land this internship?” or “How do I handle this relationship?”

We can formalize this into:

Type A:

• Well-defined problem and solution constraints
• Has many examples and is drillable

Type B:

• Large solution space
• Problem is not well-defined
• Sparse and long-horizon rewards

Properties

Now that we have the two groups defined, let’s see if we can find any interesting properties between the two!

• Observation: Type B people tend to be knowledgeable about many things, but it’s not that their focus is on being a generalist; it’s more that they need to search widely to find a solution.

• Another weird observation: it seems like Type A people in tech chase technical complexity or hard technical challenges, while Type B people chase business impact.

• Sometimes success from Type B people can be dismissed as luck.¹

I’ve been presenting Type A and Type B people as distinct groups, but in reality, they’re closer to distinct axes. For example, a mathematician still needs to find good research directions and articulate his thoughts well, which falls more into Type B.

If you had to plot them on a graph, they’d be slightly correlated vectors, a few degrees off orthogonality. Although difficult, you can be good at problem and direction. But if we already have a relatively confident picture that Type A seems to be a function of deliberate practice, meta learning, and raw intelligence, then what is Type B a function of?

The characteristics of Type B people seem to be scattered because there is more than one way to be effective, but some commonalities are roughly as follows:

• Have new ideas often
• Are very open to feedback and to giving feedback, or are good at communicating
• Usually read and write quite a lot, and some keep a personal blog page to get feedback on their ideas (or maybe this is just survivorship bias)
• Are usually relatively coherent about their ideas and can give impromptu talks or describe their position on various things clearly²
• Are proactive in trying out new things and aren’t afraid to fail
• Are clear about their goals and have a well-defined long-term strategy
• (Not too sure on this) Tend to think about systems more and develop rules and principles.

From these observations, a general framework is: when feedback is delayed, you can’t rely on clean repetitions anymore. Instead, you have to have explicit models, tests of those models with explicit predictions, outside information (from people in similar situations, social/observational learning), higher-level principles, and smaller experiments that give earlier feedback.

Footnotes

1. I think Type B intelligence is very hard to define, and what you can’t define is usually attributed to luck (almost as a catch-all explanation for unexplained success). I think this is why many people attribute founders’ or CEOs’ success to luck (which I acknowledge is a part of it), but they also have many qualities that you could not replace them with anyone else and have the same success. It’s luck in the sense that if they tried it multiple times, they might not succeed every time — but it’s intelligence and deliberate strategy in the sense that only a handful of other people could swap places with them and have any chance of replication. ↩

2. Thoughts are like a big, mushy ball of consciousness; they gather in the head and float around like gas. Talking is like a realization: it’s like opening the box in Schrödinger’s cat, or like inference with LLMs, in that it takes a big ball of probability and coalesces it into a solid idea. ↩