Hacker News Reader

Comment by 6ren | Hacker News Reader

> Without communicating with the field’s experts, Zhang started thinking about the problem. After three years, however, he had made no progress. “I was so tired,” he said.

> To take a break, Zhang visited a friend in Colorado last summer. There, on July 3, during a half-hour lull in his friend’s backyard before leaving for a concert, the solution suddenly came to him. “I immediately realized that it would work,” he said.

http://c2.com/cgi/wiki?FeynmanAlgorithm

  Write down the problem.
  Think real hard.
  Write down the solution.

View on HN

This sounds glib on the surface but it's really surprisingly deep. The key is to realize that the problem solving doesn't happen in the second step, but rather in the first. Writing down your problem means you have to structure it in some way, to frame it in a way that at least can be represented on paper. Perhaps it's a formalization, or simply a brief explanation. In any case, step 2 is only possible because of step 1 for most interesting problems. Even for those of us who aren't Feynman.

My undergrad advisor had a similar story. He was working on an old open problem for his dissertation, and his advisor (quite a well-known figure in the field) kept giving him ideas. Nothing worked. This went on for 4 years.

Then one day he decided to stop listening to others, and just work on his own. Six months later, he had a novel solution, closing the problem. His entire dissertation (including title, abstract, etc.) was 22 pages (single sided, double-spaced).

Been there, done that.

It is not as glorious as it sounds. I had some very strong intuitions about some obscure computer science problem (maybe 10 people worked on it worldwide, but I think they gave up already). I was thinking about this problem for at least one year, with some periods of very intensive thinking. I could have really used that time more productively, because that was already quite a busy period for me, but the solution seemed so close.

After about one year, I finally had a solution mapped out, which I think would work. However, there were some major implementation problems (memory complexity). That was the time when I was finally able to let it go.

So the advise that I would give to people is: focus on real-world problem, that real people have.

> focus on real-world problem, that real people have.

Because the world would be a much better place today if we hadn’t bothered about quantum mechanics, special and general relativity or even advanced mathematics to handle these topics. Sure, this sort of thing won’t solve the problems of this specific time period, but it might well be relevant to a future generation’s problems.

Quantum mechanics solved a bunch of real world problems that were relevant at the time. As one example, people were trying to predict heat capacities of gases (for things like chemical engineering industrial processes) and the theory they had, based on equipartition, produced results that didn't match experiment at room temperature. And we're not talking a few percent mismatch; we're talking significant differences.

Special relativity likewise addressed specific experimental problems people were having.

General relativity you may have more of a case for.

I disagree. We often don't know what the consequences could be of solving abstract problems by enthusiasts. Whole avenues of science were opened this way.

The issue is that we celebrate the people who follow Feynman's approach and succeed, but forget all those who (like yourself) end up failing. It's a simple matter of survivorship bias. For each person who "thinks really hard" about a particular problem of this magnitude and finds a solution, there are hundreds, if not thousands, who end up failing.

If we knew ahead of time which tough problems have elegant solutions, it would save us so much time and effort. Alas, that's not the nature of research.

My point is simply that we should treat Feynman's algorithm as a magic bullet. You have to recognize that it fails a lot more than it succeeds.

You should remember that the Feynman algorithm is a joke told by one of his colleagues (Murray Gell-Mann) to mock Feynman's story-telling. It has never been intended as a real algorithm that should be applied by anybody.

Previously on HN: https://news.ycombinator.com/item?id=1520075

Is it even an algorithm though? It's so imprecise a rule of thumb that evaluating it as a failure or a success is utterly arbitrary.

Reply on news.ycombinator.com