9
$\begingroup$

Per Moderator @BenN's request in this thread to change the site description, we need to open a new thread and vote on new suggestions.

Please propose site description texts exactly as past users did in this older thread, and vote on the suggestions of other users. After a reasonable consensus is reached, the moderators will update the site descriptions to match the top voted answer.

For people new to this process, most of the AI SE suggestions in the past follow the convention of the descriptions of most SE sites and start with something like, "Artificial Intelligence Stack Exchange is a question and answer site for ..."

$\endgroup$
1
  • $\begingroup$ Perhaps we can summon @BenN by mentioning him? It looks as though we did reach a good consensus. $\endgroup$ Oct 11 '18 at 21:33
2
$\begingroup$

The two leaders are ...

people interested in artificial intelligence theory, design, development, practice, research, and policy.

... and ...

people interested in embedded, mathematical, cognitive, and discovery centered artificial intelligence research and development.

... so I propose the union.

people interested in AI theory, mathematics, research, discovery, design, development, practice, embedded uses, cognition, policy, and impact.


This one is inclusive and dodges the terms statistics and data science which are the explicit domains of established SE siblings.

$\endgroup$
7
  • 2
    $\begingroup$ I like this idea. It's a bit wordy, but that's the not the worst thing to compromise on. $\endgroup$ Oct 4 '18 at 23:22
  • 1
    $\begingroup$ This answer seems fine to me, it doesn't seem to exclude anything that should be on-topic, and seems to cover everything that should be on-topic imo. If people think it's too wordy, I think some of the words don't add much and could be removed though: 1) mathematics (I'd say theory and research already imply that AI-related math is on-topic), 2) embedded uses (is it necessary to specify this? I feel like questions about embedded uses of AI will already also always be covered by one of the other terms, like development or practice). $\endgroup$
    – Dennis Soemers Mod
    Oct 5 '18 at 12:05
  • 1
    $\begingroup$ I don't feel strongly about removing them though... just feel like they could be removed if it's necessary to reduce the word-count for whatever reason. $\endgroup$
    – Dennis Soemers Mod
    Oct 5 '18 at 12:05
  • 1
    $\begingroup$ I appreciate the broad expansion in scope of "cognition" unqualified, but I agree it's on-topic, and no one seems to object. I agree that if we include mathematics, it's with the understanding that we're trying to raise awareness of that aspect of the field, for beginners especially. My thought is including mathematics would encourage mathematical questions, which we definitely want. $\endgroup$
    – DukeZhou Mod
    Oct 5 '18 at 22:52
  • $\begingroup$ @DouglasDaseeco I agree it's in scope and useful. Most importantly, we seem to have some consensus and buy-in on the merged descriptions in this answer, which helps in moving forward confidently with the change. $\endgroup$
    – DukeZhou Mod
    Oct 8 '18 at 20:52
  • $\begingroup$ @DukeZhou in case we need to mention you in a thread that you were on already. Please see comment thread under the main question. $\endgroup$ Oct 11 '18 at 23:35
  • 2
    $\begingroup$ I think the first vote can be dismissed. It is probable that the electorate of the previous vote is different from the electorate of this vote because different users are active now. Our process should reflect the will of those who are currently active on, and thus running, the site. $\endgroup$ Oct 12 '18 at 0:50
3
$\begingroup$

I think we should also provide guidance to users about which questions may be more suitable for Data Science, Overflow, etc.

$\endgroup$
0
2
$\begingroup$

Artificial Intelligence Stack Exchange is a question and answer site for ...

people interested in artificial intelligence theory, design, development, practice, research, and policy.

I like @DouglasDaseeco's answer, but I'm among the users who think that practice, and even code, have a place here. Presently users post questions containing code, and I and others answer them, so I think this description is more accurate.

While the founding moderators' intent was to exclude questions that overlapped with other sites (notably Data Science & Programmers.SE), the boundaries are quite porous in practice, and if we want to claim to be a useful place for AI related Q&A on the web, I think we need to accept practical questions as well.

Some examples of coding questions with no other place to go include:

Keeping track of visited states in Breadth-first Search, which is about the proper data structures to use in a search algorithm. It doesn't belong in Data Science, since it is related to GOFAI and not machine learning. It doesn't really belong in Programmers.SE, because it isn't a generic question about programming, it's related to understanding the algorithm. It seems to clearly belong on this site, and yet it includes code and is about practice.

Snake game: snake converges to going in the same direction every time This question was about the implementation of a reinforcement learning algorithm. The question again has nothing to do with Data Science. It involves programming, but the users' problems were not related to understanding how to program, but to understanding the algorithm (and, as it turned out, the exact behaviour of a particular algorithm for training neural networks). This user is not likely to get useful answers on Programmers.SE. It seems to clearly belong on this site, and yet it also includes code and is about practice.

$\endgroup$
11
  • 1
    $\begingroup$ maybe also add research to the list? $\endgroup$
    – Dennis Soemers Mod
    Sep 25 '18 at 8:12
  • 1
    $\begingroup$ @DennisSoemers Good idea. $\endgroup$ Sep 25 '18 at 12:01
  • 2
    $\begingroup$ I would like practice questions to include code as relates to e.g. A*, RL and other high-level AI constructs, but would like exclude code (and questions) that gets bogged down in minutiae of how to implement neural networks, other supervised/unsupervised learners, or Python syntax etc. Likewise, "why does my NN only get a loss of X on this data" type question. This is already a key area of confusion between Cross Validated, Data Science and Stack Overflow. Bundling AI in with those three will further dilute things. The difficulty is phrasing things, because AI is a superset including ML. $\endgroup$ Sep 26 '18 at 8:55
  • $\begingroup$ Strongly agree that we have to include practice. I also think Neil's points are important, as we do need find the right boundaries with the stacks where there is overlap. PS I wonder if "history" might also have a place. (There is a history of science & mathematics stack, but I feel that the history of AI is most usefully covered here.) $\endgroup$
    – DukeZhou Mod
    Sep 26 '18 at 23:37
  • $\begingroup$ @NeilSlater & JohnDoucette What are your thoughts on providing a little guidance on questions that would be more suitable for Data Science, Overflow, etc. $\endgroup$
    – DukeZhou Mod
    Sep 28 '18 at 18:27
  • 1
    $\begingroup$ @DukeZhou: I am torn on that. Although there is clearly a difference between the sites that seem relevant, even experts on multiples of those sites find it hard to express what the actual differences are. It is next to impossible to word a short enough clear statement that a new question asker could follow, even assuming they take the time. The current trend to conflate "Deep X" using NNs with AI is going to cause a lot of misplaced questions. I think the guidance is going to be best focussed on the experts at each site so we can gently push questions around . . . $\endgroup$ Sep 28 '18 at 19:01
  • 3
    $\begingroup$ . . . but the first barrier to that is getting a bunch of experts to agree on any course of action when there is no clear consensus. My default position right now is to answer questions wherever they land, and only comment on suitability of another site when a question would very clearly be welcomed on an alternative according to history of similar questions that I have seen. That's definitely how I have been treating technical questions in Reinforcement Learning. $\endgroup$ Sep 28 '18 at 19:03
  • 2
    $\begingroup$ @DukeZhou I agree with Neil. I think that exactly which technical questions are "on-topic" will be very hard to define, but that moderators can "know it when they see it". A question that is primarily about how to program something is probably (though not always) off topic. A question about what to program (i.e. what the steps in an algorithm actually mean) is probably on topic. But there are lots that fall in between those questions. $\endgroup$ Sep 28 '18 at 19:36
  • 1
    $\begingroup$ @NeilSlater "Answer where you see it" seems to best serve the mission of Stack. (Pedantic adherence to guideline minutiae, especially as there is legitimate fuzziness, has definitely not been helpful.) Thanks for sharing your thinking on this and your approach. $\endgroup$
    – DukeZhou Mod
    Sep 28 '18 at 20:11
  • $\begingroup$ Snake game: snake converges to going in the same direction every time is on-topic on both CV and DS. $\endgroup$ Sep 30 '18 at 18:30
  • 2
    $\begingroup$ It may be on topic in CV or DS (I suspect it would be closed on either of those, despite the site policies). A bigger question is: why would any person expect it to be on topic there? The user certainly isn't doing anything I'd recognize as data science. They are doing a kind of machine learning, but it's not anything a regular statistician would expect to see. If I were going to ask about it, I'd expect to get an answer on the "AI" stack, and if I was searching for an answer, that's where I'd expect to find one too. $\endgroup$ Sep 30 '18 at 23:55

You must log in to answer this question.

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