4
$\begingroup$

I've been active on SE from some time and only recently entered in ai.SE community. In very large communities like SO questions about resources are not welcomed and usually are closed after a short while. In smallest or not very large communities are usually welcomed since they could be helpful to beginners.

So, are this type of questions welcomed here?

For the sake of the question, I've already posted a question but then the doubt comes up and I thought that it was better to ask instead of closing in advance my question. The post itself isn't opinion based or too broad but, as I said, not all communities welcome list type question.


Just to be clear, what I means for resources isn't links to external sites that could easily expire. I mean books, articles and so on. Of course links to external resources like tensorflow/keras/caffe/etc. manuals, tutorials or documentation are welcomed.

$\endgroup$
  • $\begingroup$ As you have an opinion that your question might suits the site (due to its small size), then I suggest you also write an answer. It is votes on the answers which will gauge community opinion. Votes on the question are not agreements with your stance, they are agreements that it is a worthwhile Meta question. In fact your first upvote is from me - I wish more Beta sites engaged with this question, it crops up repeatedly. $\endgroup$ – Neil Slater Jul 15 '18 at 21:08
3
$\begingroup$

The problems with generic external resource requests don't really change due to the size of the site.

  • Links can fail, or go out of date. An answer that is mostly links could degrade so that it is not usable, unless it was actively maintained. This is also why link-only answers are discouraged.

  • A "correct" answer is hard to assess.

  • There is a strong element of opinion on what to include or exclude when compiling "comprehensive" lists. There is an implied "and the list should be reviewed for relevance and curated" which is hard to objectify, but if it wasn't present then clearly just Googling e.g. "Reinforcement Learning tutorials and MOOCs" would be enough for the OP.

  • No-one will actually read or use a comprehensive list of introductory material. It becomes like a restaurant menu where a reader has to attempt to pick out the 2 or 3 items from the answer that would be most useful to them.

  • I don't think that technical avoidance of actual hyperlinks, and use of ISBNs, course codes etc changes the nature of this at all. Some external references have a long shelf life. E.g. "Origin of Species" is still relevant today. But this does not apply to all books, just because they are books.

Just to be clear, what I means for resources isn't links to external sites that could easily expire.

Perhaps if you made it clear what the nature of these non-link resources would look like in an answer, it could help move it out of being a request for generic resources, and become a more focused question. E.g. "What are the must have introductory books in subject area, and what prior knowledge do they assume?" is a lot more focused than "I'm looking for a comprehensive list of MOOCs, books, tutorial and good resources" which is essentially asking for anything and everything that might be useful, without bounds.

$\endgroup$
  • $\begingroup$ I think users can request materials in chat, especially from those who top answered in a tag. $\endgroup$ – DuttaA Jul 15 '18 at 23:22
  • $\begingroup$ @DuttaA: Yes of course, chat can accept a much wider range of topics. In addition, when someone has asked for links in good faith, I also will usually post a comment that both suggests edit/closing the question and gives a useful starting link. Because sharing a link that I know is very easy. I did that in this case on OP's question. $\endgroup$ – Neil Slater Jul 16 '18 at 6:55
  • $\begingroup$ I agree with some of your points and don't agree with others. e.g. you don't need to have dozens of answer, you could also have one wiki answer in this case (also the question should be a wiki one IMHO). $\endgroup$ – gvgramazio Jul 16 '18 at 11:33
  • 2
    $\begingroup$ @gvgramazio Yes a Wiki answer might work better for creating managed lists. One problem then is users, mostly well-meaning, adding low quality links and references - which is harder to spot/manage compared to other edits because they are just links and references. $\endgroup$ – Neil Slater Jul 16 '18 at 11:48
  • $\begingroup$ Yes, you're right. But the post could be protected, in this way only expert or semi-expert users could add links and references. $\endgroup$ – gvgramazio Jul 16 '18 at 11:53
  • $\begingroup$ @gvgramazio the problem is that expert or semi expert level users Generally don't answer such questions on this stack at least...From personal experience, when I was new here I answered resource recommendation questions in a nice comprehensive manner, after a certain time similar questions with a little variations started coming in and I stopped answering, I will assume it is the same case with experienced users they get tired of new users who ask similar questions without exploring the stack. $\endgroup$ – DuttaA Jul 16 '18 at 20:28
2
$\begingroup$

When I think about this question in the context of research papers, for instance, I can't see a real issue.

Ideally, when posting research papers links, the title of the paper will be used in addition to the link, so if the link goes bad, people can still search for the paper.

Russel & Norvig's Artificial Intelligence: A Modern Approach is heavily cited on SE:AI, and the text was originally published in 1995. The book is in its 3rd edition now, (which is not always noted when cited,) but even the 3rd edition dates from 2009, earlier than the recent Machine Learning milestones (~2016) yet the textbook is still relevant and heavily utilized.

List questions do have some issues (see Neil Slater's answer) and seem to be off-topic in general across Stack exchange.

However, I'd still think lists of research papers on a given topic, ideally peer-reviewed, would provide utility and carry archival value. In the same way, lists of well-regarded textbooks could be useful.


Second Consideration: Contemporary Hacker Culture and Youtube

In some sense we're the "General AI" site, covering the full scope of the field, as opposed to focusing on any given specific aspect (distinct from stacks like Data Science.)

We seem to be the stack where beginners typically come to first. I created a getting-started tag because there are so many of these questions.

Many people today are learning the basics today via youtube videos. Where the videos are solid, they seem to provide benefit, but they tend to be more ephemeral, especially when they come from non-academic sources. (Erik Demaine's lectures on Time Complexity will likely be available for a very long time indeed, where a random youtuber using click-baitey titles subject matter to generate ad-revenues may not be.)

My feeling is, re: videos, is that anything commercial should be avoided, but anything coming from accredited academic institutions is reliable and suitable.

$\endgroup$
  • $\begingroup$ I agree with your arguments but I don't get your conclusions. Should question regarding a specific topic of artificial intelligence be considered off-topic or not in this site? $\endgroup$ – gvgramazio Jul 22 '18 at 14:05
  • $\begingroup$ @gvgramazio Personally, if the resources are vetted (peer reviewed, standard texts, videos from academic institutions and other reliable sources) I tend to be in favor. $\endgroup$ – DukeZhou Jul 23 '18 at 19:20
0
$\begingroup$

After a week and after reading your comments and answers I still think that ai.SE could benefit from resource request questions. However, I think that my original question on main ai.SE is badly posed.

Some reasons why ai.SE could benefit from resource request questions are:

  • They attract visitors. This type of questions have usually a lot of views and could attract new visitors from web search engines.
  • They could prevent some users to post dumb questions. This could be only a personal thought but one of the main reasons why I choose to join this community is that, as a self-thought beginner, I don't know which sources I should consider trustworthy.
  • They could be helpful even for non-beginners. Even at the semi-professional level, one could find new resources interesting.
  • They could condense a lot of similar questions that ask for resources about some topic.

Of course, there are some disadvantages to this type of question. One of them is how to choose which answer should be accepted. I think that we could have one answer for each resource suggested and one accepted community answer that keeps track of the top resources linked. The community answer could be edit by anyone that has a certain reputation.

$\endgroup$
  • 1
    $\begingroup$ Thanks for following my suggestion and posting it. I don't personally think these points are enough, but it's a well-thought out answer, so not going to downvote just because I disagree :-) $\endgroup$ – Neil Slater Jul 16 '18 at 11:53
  • 1
    $\begingroup$ @NeilSlater I have no problems if you down-vote it because you don't agree. I neither upvoted nor downvoted yours because I agree with some points and disagree with others. ;) $\endgroup$ – gvgramazio Jul 16 '18 at 11:56

You must log in to answer this question.

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