Loading slide

Loading contents...

[░░░░░░░░░░░░░░░░░░][░░░░░░░░░░░░░░░░░░░░░░░░░░░░]0 / 16
<back>next

Module 3 Chapter 3

What the Cold Taught Us

Nothing that failed in the winters was worthless. Search still routes your journey. Explicit rules still run payroll and hold aircraft in the sky. A method meeting its limit is not a method being wrong, and the tools that came before did not stop working when something newer arrived.

What the cold exposed was narrower and more damaging than any particular failure. It was that the whole approach rested on a person being able to say what they know, and people cannot. You recognise a friend's face in a fraction of a second and could not write the rulebook for doing it if you were given a year. The knowledge is real, you use it constantly, and it will not come out onto paper.

That is a wall no amount of effort gets over, and it forces a genuinely different bet: stop trying to state the rule, hand over a pile of examples instead, and let the machine work out the rule for itself.

It is worth being honest about what that gives up. Nobody who takes this route ends up with an explanation they can read. The bet is that a rule you cannot state is better than a rule you cannot write, and that trade has consequences that run all the way to the present.

In this chapter

  • Using a method where it fitswhy search and rules remain useful within clear boundaries
  • What experts cannot fully explainwhy expertise can resist complete description
  • Letting examples shape the rulehow data can change a system's decisions
  • What people still decidewhich parts of a learning system still come from humans
  • Making the rules adjustablewhy changeable settings became the next path forward
# citations