5 ms·

Why comment when you don't have the time for more than superficial analysis?

by yread 11y ago

algorias 11y ago
According to that criterion, no one should comment on HN ever, unless they are writing at the level of a peer reviewed journal. I supposed it would be interesting for people here for me to at least point out some obvious red flags.
gjm11 11y ago
Because even the superficial analysis provides more than zero useful information. (Nothing super-obviously wrong. Nothing that looks clever enough to overcome the guess that, none the less, something's probably wrong.)
Anderkent 11y ago
'This relies on their earlier paper from May 2015, and if that paper was correct it would have made big waves' is valuable context.
Steuard 11y ago
Because experts with domain-specific knowledge tend to have better heuristics for this sort of thing than the average interested outsider. (They have to, or else they'd get nothing done except debunking flawed claim after flawed claim. Which might be worthwhile, but doesn't move the field forward.) I've seen plenty of "theories of particle physics" that evidently look fairly plausible to non-physicists (and maybe even to non-specialists) that I can recognize immediately as crackpottery by my own heuristics (but that would take at least a weekend's work to convincingly demonstrate as such, if I wanted to invest that kind of time: I've done that, too). I'm not saying that anyone should take anyone's heuristic guesses in such cases as gospel (once or twice a century we might actually get a delightful surprise), but they can serve as a useful restraining influence if you're tempted to get excited about a headline.
- anfedorov 11y ago
  Nope. An expert would be able to skim it, recognize the path of reasoning as something he's seen a before, and narrow down on the error quickly. Reasoning about something by how many "waves in the community" it has made is a deference of personal judgement to the social network an individual is embedded in, which is a useful shortcut when you are not an expert relative to your peers, but dangerous set of mind to operate in for any lengthy amount of time, as it can be the basis of cults and the like.
- anfedorov 11y ago
  An expert would be able to skim it, recognize the path of reasoning as something he's seen a before, and narrow down on the error quickly. Reasoning about something by how many "waves in the community" it has made is a deference of personal judgement to the social network an individual is embedded in, which is a useful shortcut when you are not an expert relative to your peers, but dangerous set of mind to operate in for any lengthy amount of time, as it can be the basis of cults and the like.
  - evanpw 11y ago
    > An expert would be able to skim it, recognize the path of reasoning as something he's seen a before, and narrow down on the error quickly. This is almost exactly equivalent to saying that an expert programmer should be able to determine whether an unfamiliar, complicated code base contains a bug by skimming the source. The hard bugs are going to be subtle.
    - ejk314 11y ago
      > complicated code base So then the answer is certainly "yes" with the same level of certainty that algorias is claiming. Large, complicated solutions are inherently likely to have subtle flaws.
    - anfedorov 11y ago
      Not at all equivalent, because code isn't meant to be human readable, but computer interpretable. Because of these different goals, computer code is actually quite a bit harder to read than a well structured academic paper. I think programmers have a lot to learn about logical structure from writing meant to be read by humans [1]. Note how many people on this thread seem to be understanding and discussion its contents, whereas if I were to post 13 pages of "unfamiliar, complicated code" that claimed to do something, I can't imagine anyone would help me debug it. 1. Knuth expounds on this idea here: https://en.wikipedia.org/wiki/Literate_programming https://en.wikipedia.org/wiki/Literate_programming

Why comment when you don't have the time for more than superficial analysis?

by yread 11y ago

Why comment when you don't have the time for more than superficial analysis?

According to that criterion, no one should comment on HN ever, unless they are writing at the level of a peer reviewed journal. I supposed it would be interesting for people here for me to at least point out some obvious red flags.

gjm11 11y ago

Because even the superficial analysis provides more than zero useful information. (Nothing super-obviously wrong. Nothing that looks clever enough to overcome the guess that, none the less, something's probably wrong.)

Anderkent 11y ago

'This relies on their earlier paper from May 2015, and if that paper was correct it would have made big waves' is valuable context.

Steuard 11y ago

Because experts with domain-specific knowledge tend to have better heuristics for this sort of thing than the average interested outsider. (They have to, or else they'd get nothing done except debunking flawed claim after flawed claim. Which might be worthwhile, but doesn't move the field forward.) I've seen plenty of "theories of particle physics" that evidently look fairly plausible to non-physicists (and maybe even to non-specialists) that I can recognize immediately as crackpottery by my own heuristics (but that would take at least a weekend's work to convincingly demonstrate as such, if I wanted to invest that kind of time: I've done that, too). I'm not saying that anyone should take anyone's heuristic guesses in such cases as gospel (once or twice a century we might actually get a delightful surprise), but they can serve as a useful restraining influence if you're tempted to get excited about a headline.

anfedorov 11y ago
Nope. An expert would be able to skim it, recognize the path of reasoning as something he's seen a before, and narrow down on the error quickly. Reasoning about something by how many "waves in the community" it has made is a deference of personal judgement to the social network an individual is embedded in, which is a useful shortcut when you are not an expert relative to your peers, but dangerous set of mind to operate in for any lengthy amount of time, as it can be the basis of cults and the like.
anfedorov 11y ago
An expert would be able to skim it, recognize the path of reasoning as something he's seen a before, and narrow down on the error quickly. Reasoning about something by how many "waves in the community" it has made is a deference of personal judgement to the social network an individual is embedded in, which is a useful shortcut when you are not an expert relative to your peers, but dangerous set of mind to operate in for any lengthy amount of time, as it can be the basis of cults and the like.
- evanpw 11y ago
  > An expert would be able to skim it, recognize the path of reasoning as something he's seen a before, and narrow down on the error quickly. This is almost exactly equivalent to saying that an expert programmer should be able to determine whether an unfamiliar, complicated code base contains a bug by skimming the source. The hard bugs are going to be subtle.
  - ejk314 11y ago
    > complicated code base So then the answer is certainly "yes" with the same level of certainty that algorias is claiming. Large, complicated solutions are inherently likely to have subtle flaws.
  - anfedorov 11y ago
    Not at all equivalent, because code isn't meant to be human readable, but computer interpretable. Because of these different goals, computer code is actually quite a bit harder to read than a well structured academic paper. I think programmers have a lot to learn about logical structure from writing meant to be read by humans [1]. Note how many people on this thread seem to be understanding and discussion its contents, whereas if I were to post 13 pages of "unfamiliar, complicated code" that claimed to do something, I can't imagine anyone would help me debug it. 1. Knuth expounds on this idea here: https://en.wikipedia.org/wiki/Literate_programming https://en.wikipedia.org/wiki/Literate_programming