3
$\begingroup$

This question was recently bumped to the front page. On some other SEs, the words "latest methods" would be a red flag and potentially cause it to be closed because any answer would have to be updated to continue to be true. The general philosophy on those other sites is that questions should be answerable in a fashion that is going to be useful not just to the questioner, but also to future visitors to the site. Indeed, one of the major uses of the SE network is as a long-standing repository of knowledge for future questioners.

Although new things can always be discovered that invalidate old answers that pre-date them, questions like the one I link to are going to attract answers that are always going to be going out of date. There is no way for an answerer to answer this question in a way that will be useful and true for a reader in six months or a year.

Is that an issue here, or are we fine with questions whose answers will inevitably become out of date?

$\endgroup$
  • $\begingroup$ Stella, thanks for raising this issue! I've provisionally retitled that question to include the year, but this is a topic requiring serious discussion. I definitely feel there is value in historical record and historical methods, but, by the same token, we don't want visitors unaware that a technique or approach may be outdated. $\endgroup$ – DukeZhou♦ Sep 7 '18 at 2:06
  • 1
    $\begingroup$ I’ve expanded my question a bit, but it’s worth noting that any policy about this should also take into account the fact that a question similar to “is there a method for doing X efficiently” is likely to one day be out-of-date if the answer is “no.” I do see a substantial difference between these two cases (and the later seems far more akin to the observation “well new facts could be discovered that invalidate answers”), but I think it’s nevertheless an important thing to note. $\endgroup$ – Stella Biderman Sep 7 '18 at 2:14
  • 2
    $\begingroup$ Regardless of how the answer comes down from the community on this issue, it’s very likely a good idea to encourage users to update answers they post as they become obsolete due to technological progression, and to encourage users to post later answers to questions when the old answers are not updated. A good example of this working well is StackOverflow where it is common to see older answers in Python 2.x, and later answers in 3.1 and even later ones in 3.5. I think this is a good way to preserve the historical value @DukeZhou raises in the context of answers effected by new discoveries. $\endgroup$ – Stella Biderman Sep 7 '18 at 2:18
3
$\begingroup$

I'd personally hesitate to declare questions off-topic just because the "correct" answers to them are highly likely to change over time. Indeed, this is going to be the case for "state-of-the-art" questions, especially considering how rapidly research in the field is moving and how rapidly the state of the art changes. I personally still feel like such questions aren't overly problematic because:

  1. They are likely a class of questions that is the most interesting for some people. In a field that moves this rapidly, and where young people are newly entering the field also at increasing rates, there is a lot of interest in knowing "what is the state of the art right now?". If there is a lot of demand for such questions, it makes sense to have a place where they can be asked to me.

  2. The web interface of the site already puts timestamps (dates) on questions and answers. Future visitors will be able to see (if they pay attention) if an answer they've run into is rather old.

  3. In the future, if the state of the art has significantly changed, people are free to provide new answers or add comments to existing (outdated) answers. If the people who wrote the original outdated answers are still around, they can also edit them. See, for example, this old question on stackoverflow. It was originally asked 7 years ago, and was about the feasibility of training an Artificial Neural Network to play a complex video game like Diablo 2. At the time, this was highly unlikely to be feasible. However, we see some answers being written a few years later, and also see many edits in the question itself and in older answers, to reflect progress in the field.

$\endgroup$
  • 1
    $\begingroup$ +1 for the ability to provide new answers. We should offer a "best effort" level of service, but that will still be very useful for 99% of users. The academics and industry researchers who care about true the state-of-the-art should looking in journals and conference proceedings anyway, not here. $\endgroup$ – John Doucette Sep 10 '18 at 14:58
0
$\begingroup$

There is a difference between disproven and out of vogue. What is proven false, if the proof stands up to thorough scrutiny is unlikely to have any future value other than to demonstrate how some things that were once widely believed may be later disproven. These are some examples.

  • The sun travels around the earth.
  • Air, earth, water, and fire are the four elements.
  • All propositions within a mathematical system can be proven or disproven.
  • All natural phenomena can be placed in algebraic closed form.

Many things that were thought absurd have been proven.

  • Mercury is travelling too fast for its orbital path to be predicted by Newton's laws.
  • Humans can survive in space and return alive.
  • Computers can accurately and reliably sort mail with handwritten addresses.
  • Computers can be trained to generate pictures of interior designs.

However, very little in a Q&A community are theorem that can be proven or disproven though. Much of what is discussed is technique (in the Jaques Ellul sense of the word) that may fall into voge and then out of vogue more than once. Something that is thought to be obsolete (but not formally disproven) may rise back to common use or may return slowly from obsolescence over decades. Here are just a few examples of this toggling evident today.

  • Earth is round -> earth is flat -> earth is round
  • Vector graphics -> raster graphics -> vector graphics
  • Ether -> no ether -> ether
  • Turmeric -> chemotherapy -> turmeric
  • AI via imitating biology -> AI via logic -> AI via imitating biology

Given the history of science and technology, unless we can flat out disprove an answer we cannot, solely on the basis of its current apparent obsolescence, assert that it will never return to the forefront. It is rather highly probable that some thing we now consider obsolete will become a key element in the furtherance of one of the sub-fields of Artificial Intelligence.

Others in the future may look back and consider us ignorant for our current belief that some solution or approach is obsolete.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .