Smarter than smart

There are 2 ways to be “smart”, and one way is better than the other.

Smart Version 1: Know the answer to question X.

Smart Version 2: Know how to find the answer to question X.

If someone were to build 2 giant computers, one that was Smart Version 1 (SV1) and one that was Smart Version 2 (SV2), what differences would they have?

Computer 1 (C1) would need a database of answers to all questions X. For every question it received, it would parse it, interpret it, look up the corresponding answer, and return the answer.

Computer 2 (C2) would need a set of programs that could answer all questions X. For every question it received it would need to parse it, interpret it, look up the corresponding program that can answer that type of question, run the program, get the answer, and return the answer.

Subtle difference but important.

The benefit of C2 over C1 is that the there are probably fewer answer-finding programs required than answers. So the step where the look up happens would probably require many orders of magnitude fewer rows of information to look through.

On the other hand, the downside of C2 is that getting the answer would be much slower, since it would require a program to be run… and so far we haven’t promised any minimum amount of time that the program can run before it needs to return an answer.


Now switching from computers to humans, when we think about our education system and the skills that we are given during our schooling years, are they training us to be humans that know all of the answers to question X (H1), or humans that know how to find the answers to all questions X (H2)?

Think about the kind of people who succeed at school/college/academics? Think about the kind of people we look for when hiring. What kind of questions do we ask in the SAT? What kind of questions do we ask in interviews?

Testing for existing knowledge is a lot easier than testing for answer-finding programs.

What is the difference in strategy between becoming an H1 versus an H2?

Becoming an H1 requires training to store a large set of data. Answers.

Becoming an H2 requires being willing to forget answers and to instead work towards general strategies for finding answers.

I think they are actually two very different learning strategies, and that given our minimal short and long term memory capacities and time spent in education, you can’t hedge your bets and do both… you have to pick a side.


Businesses have developed their own versions of becoming smart. Some are H1 methods, some are H2 methods.


Evolution is a version of C2 and H2. It has no ability to store answers to questions, but has developed a system of solving problems through programs (natural selection) that is way more efficient than C1/H1.


What are the processes that support H1 versus the process that support H2?

H1 processes: experiment data, user research data, market research data, highly qualified and experienced people who have been thinking about your problem area for a long time.

H2 processes: state your core problem and the desires it fulfills, find people who have the desire/problem, propose solutions, get feedback on the solutions, build better solutions.

H1 requires building a catalog of answers. H2 requires running a process. I strongly believe that H2 is more efficient, and less likely to become outdated, than H1. H1 is faster and has better short-term ROI than H2, but H2 is better in the long term.


H2’s strategy in a nutshell:

  1. Define the problem in terms of desires that are currently unfulfilled
  2. Propose a variety of solutions to fill unfulfilled desire
  3. Gather feedback from people who have unfulfilled desires on whether or not the proposed solution helps them
  4. If desire fulfilled, go to step 5, else go to step 2.
  5. Market your proposed solution to people who have unfulfilled desires

If you think about it, H2’s strategy doesn’t actually require answers to any questions. It only requires quick feedback loops between people who have a problem and possible solutions to the problem.

Similarly, natural selection doesn’t require answers. A specific species doesn’t need to know how to survive, they simply need to have quick enough reproduction cycles and sufficient mutation between generations to find a solution to an environment’s changed circumstances before they run out of the ability to generate answers.

Natural selection is a program that, evolutionarily-speaking, has solved enough questions on our planet to lead to a species of animals that can understand the program that ran to generate them.


Nassim Nicholas Taleb calls optimal H2 strategy “optionality” and “antifragility”. The ability to execute a simple program (mutate, gather feedback, select best, repeat) that is optimized at finding a solution before options run out.


tl;dr; The best way to be smarter than smart is to have a set of antifragile programs (mutate, gather feedback, select best, repeat) instead of trying to store objective solutions to future problems. In other words:

SV2 and C2 and H2 > SV1 and C1 and H1

To be smarter than smart, let go of specific answers and embrace answer finding programs.

 
127
Kudos
 
127
Kudos

Now read this

Experimenting with subjectivity

After going to the Quantified Self conference last month and thinking about it for a few weeks, I’ve decided to start a new self-tracking experiment… or rather to adapt my existing one to new thoughts on what’s worth tracking. So,... Continue →