up vote 10 down vote favorite
2

While this is primarily directed at Machine Learning, as it is 78% commitment, this thought also applies to sites like Artificial Intelligence & Robotics and NLP/CL.

While we have an existing, launched, and graduated site on Theoretical Computer Science, it does not currently allow non-research level questions. And we currently do not have a site that properly supports Computer Science itself - it's been argued across the network about how Stack Overflow isn't completely sufficient for the task, and other sites are too narrow to cover the subject as a whole.

There is a proposal for general Computer Science, but it is lagging behind Machine Learning (probably because it got a late start due to a lot of fiascoes in trying to fit Computer Science into the already existing sites). We were wondering about how the Machine Learning following would consider with being merged into this Computer Science site proposal.

If the merge is undesired, then Machine Learning will progress as normal, launch into Beta, and follow all of the normal process. If the merge is agreed upon, it will be subsumed into the Computer Science topic. Machine Learning may belong within the scope of Computer Science to an extent, but it's ultimately up to the community to denote whether they're comfortable with trying at their own, or if they think they can rig up much better success by being part of a more comprehensive site.

There is another discussion about merging Machine Learning with the other two named proposals - I'd like to extend the option of merging to include the Computer Science site.

link
[status-deferred] until one of the CS communities stands up united to do this. See my answer below. – Robert Cartaino♦ Feb 6 at 22:03
Perhaps wikipedia definition of AI as a part of CS will clear things up. An Applied Computer Science site is the orthodox compliment to CSTheory. If the "acedemicians" bit can be struck from the CS prop (and possibly the name changed to Applied Computer Science) then you can merge the AI and robotics into CS ... Machine Learning, NLP, Computer Vision etc. all fall into AI so merge them too. If we don't do this we end up with 6 failed betas and a lot of unhappy stackers – Merlin Feb 7 at 22:37
"Applied CS" is a subfield of CS; why restrict the new site to it? Note that from the academics point of view, Applied CS is even outside CS. – Raphael Feb 8 at 11:57
feedback

6 Answers

up vote 8 down vote

The problem is, that a too broad proposal might fail to attract "real experts". If there was no demand for a bit more focused sites, the proposals probably would not have gotten as many committers as they did. I think the people committed to those more narrow proposals for a reason. Now the key is to find the right size between "narrow enough to attract experts" and "wide enough to create an active userbase", which is why I suggested the merge of AI+ML+NLP, because I thought that might be a good compromise which lies in that range.

And as you say - there already exists a graduated, active site about theoretical computer science at research-level, which means that all questions left over for a big merged site from that direction would be low-level undergraduate general computer science questions. I see no benefit from this for the community of the AI/ML/NLP proposal, who try to gather experts and people interested in this particular field. I'd even say the merge you suggest would probably do harm and scare specialized people (who committed to those more narrow proposals for a reason) off, so I don't think it is a good idea.

But I think for (theoretical) Computer Science a two-level model could work: Just let the Computer Science proposal where CS questions at undergraduate niveau can be asked grow and exsist on its own, and as the students gain more knowledge they can move on to the research level site. As many undergraduate CS students as there are all over the world it shouldn't be a problem to get the proposal running on its own, if some people start to advertise it properly. As for the more "non-mainstream" topics like Cryptography, Robotics, AI/ML/NLP and so on, which require non-obligatory coursework for normal CS students and even have complete universal programs and industry branches on their own, I think it would be better to have seperate sites which gather the people involved in the respective area.

link
Thank you for your insight and contribution to the matter. I take it from this that you think that the comprehensive Computer Science model is going to be too broad and fail to attract experts. Rather, you feel that those sites could be merged, but not as part of the CS proposal, am I grabbing that right? – Grace Note♦ Jan 18 at 13:45
Indeed. As I wrote - maybe a comprehensive CS site for not-research-level questions can exist on its own, but I think it shouldn't be merged with the people who signed up for a more narrow site. When it somes to merging AI+ML+NLP I think it is still narrow enough, but also attracting enough users for a heathy site, which might not be the case if those pages try to stand alone. – eowl Jan 18 at 13:52
eowl, I think you misunderstand the scope of cstheory.SE as it is only a very small part even of research-level CS. Therefore, the general CS site can/should attract many experts that happen to research CS outside of TCS. Besides, I don't see why CS should be too broad as math.SE and physics.SE do very well. I think the CS site should include the non-mainstream topics, because experts of those fields can mingle and spread their knowledge about "mainstream" fields and, most importantly, basics to others. – Raphael Jan 19 at 8:59
maybe it works for math or physics already, but the existence of TCS shows, that experts seem not to be interested in a broad computer science site, why wouldn't they have build a general one in the first place? the field is huge and as I said - there is a reason people voted for more narrow proposals here on Area51. Nevertheless I think there is a place for a general CS site on SE, but it has to reach the right people who are interested in a general site and commit for it independently. – eowl Jan 19 at 9:34
Maybe people did so because there was no general CS site and those specific set of persons did not want to invest in a site that would only contain few interesting questions? That is not equivalent with using a general site for special stuff if it is already up and running. By the way, based on the amount of CS questions that are unanswered on SO or closed as offtopic on cstheory, a general CS site is needed. My claim is that once it is there, it can also be a home for many who now try to establish a specialists community (and struggle). – Raphael Jan 19 at 14:22
2  
‘A too broad proposal might fail to attract “real experts”’: that's generally not the case, not when the discrimination is by topic (as opposed to by community). The Area 51 model tends to produce sites that are too small. It is perfectly normal to be uninterested in the majority of a site. Witness Stack Overflow, the original Stack Exchange site: most people know only a few programming languages and a large majority of questions are thus boring to them. – Gilles Jan 20 at 0:23
2  
A big part of why there's no generic computer science site is that several CS site proposals were shot down, not due to a lack of interest, but due to a perceived duplication with Stack Overflow or with CSTheory (unfortunately, closing a proposal on Area 51 does not require understanding what it's about). – Gilles Jan 20 at 0:25
could be.. could not be.. but that people vote for more narrow here sites is a fact... I don't want to be the one deciding to merge, when lots of people didn't sign up for that. maybe we need some opinion from some moderators here, who have to execute a possible merge in the end anyhow. – eowl Jan 20 at 8:15
That people vote for narrow proposals might hint that the crowd is in general not smart. ;) But yea, I think Stack Exchange needs processes to not only merge proposals but also to split and merge sites when they are running. There is nothing wrong with giving a narrow site a chance. However, there should be the possibility of moving to a more general site if it fails. – Raphael Jan 21 at 14:12
@Raphael The possibility of broadening a site's topic after a site goes live exists, but it's very difficult. The private beta is designed to bootstrap content into the site. If a topic is added later, it will have to start from scratch, and without experts; this is unlikely to work. A merge of two sites could work, but only if the resulting site is still viable (if A and B are subtopics of a domain D and each makes sense on its own, A+B may not make a viable site). – Gilles Jan 23 at 1:36
feedback
up vote 5 down vote

I don't see any reason to merge ML with other CS proposals. As Raphael points out, ML is a genuine cross-discipline area that exists as a natural community IRL.

link
1  
well, almost every sub-field of CS is cross-disciplinary (HCI? Digital Graphics? Vision? Computational Biology? Computational Linguistics? Computer Architecture? Computational Finance and Economics? ...) In fact, it would be quite hard to find some which are not, and that's the reason I love CS. :) – Kaveh Jan 30 at 6:58
1  
Architecture; Databases; Compilers; Networking; and yes, even theory. When I say ML is cross-disciplinary, I meant that there are people trained in ML who may not have any training in CS at all. – Suresh Jan 30 at 16:49
feedback
up vote 1 down vote

I'll only remind that Machine Learning is on topic and quite active on CrossValidated.

link
This may have been a viable argument in the early stages of the proposal, but there is plenty of precedent for Stack 2.0 sites spinning off specific areas on the parent sites. You've trumpeted this all along, but how about offering some support for the nascent site and seeing how it goes. It's clear that even if ML doesn't get its own site right now, there is at least one other proposal waiting to subsume the topic. – jonsca Feb 17 at 0:30
1  
An own site notwithstanding, I think ML is better at home on stats.SE than on cs.SE. This a huge imho, though; I am sure there are questions or even subfields that fit better to cs.SE. – Raphael Feb 17 at 14:21
@Raphael You are a victim of a common misconception that stats.SE is about stats -- it is about stats, machine learning and data visualization. – mbq Feb 18 at 8:59
@mbq I fail to see how your comment relates to mine (which was directed at jonsca; I should have made that clearer), sorry. – Raphael Feb 18 at 14:23
@Raphael You didn't @ me, so I didn't get notification. I think ML has become de facto on topic at stats. I don't think it gets the attention it deserves there, and I don't see the harm in at least trying to give it its own site for the time being. It seems as though mbq may be concerned about the loss of questions more than getting answers for them. People spoke, they committed to a proposal for ML on its own, and its in beta. – jonsca Feb 19 at 2:41
If it's found that the ML site can't exist on its own, the topic is probably a better fit for the proposed CS.SE, as practical problems in ML are at least slightly more likely to be associated with computer vision, NLP, etc. Theoretical ML problems related to stats should, of course, remain on stats.SE. – jonsca Feb 19 at 2:42
@Raphael Sorry, I wanted to @ jonsca as well. – mbq Feb 19 at 11:50
1  
@jonsca I'm perfectly concerned about ML questions getting answers, that's why I think they are better fit on a 6kvisit/day site on which they already are on topic and actively discussed for over a year than on 0.1kvisit/day dying beta. – mbq Feb 19 at 12:02
feedback
up vote 0 down vote

Machine Learning is a tough nut to crack. As far as I know, ML consists of roughly equal parts of computer science, pure mathematics and statistics. It also overlaps with electrical engineering to some extent. The mixture extends to the people active in ML.

Therefore, while I think that we should avoid overly specific sites and merge quite a number of proposals into CS, I am not sure about ML. Should we rather merge it into math.SE, cs.SE or stats.SE? There is also scicomp.SE (in beta) which might be a good fit. Or should we merge it somewhere at all?

That said, I think that -- in the specific case of ML -- we should not merge now. The ML proposal looks healthy, so we should wait and see how it fares in beta. If it comes out strong, it has earned to be a site in its own right. If not, we can discuss again and maybe make a better decision regarding the merge target based on the direction the site has leaned towards.

Nota bene: cryptography could, imho, very well be merged into math.SE, and robotics into CS.

link
it might be a good idea to post a meta question on Mathematics Meta and Meta Stack Overflow to further promote the proposal. – Kaveh Feb 9 at 4:26
Right, I'm on it. – Raphael Feb 9 at 9:56
feedback
up vote 0 down vote

@GraceNote thanks so much for getting back to us :)

The reason I started the AI proposal was to induct the 160,000 people who took part i the recent Stanford Introduction to artificial intelligence and related CS topics at a general interest level. These include ML, NLP, Game Theory and Search amongst other topics. Various robotics-related topics are also included such as SLAM and computer-vision.

After a little digging it became clear that there were a lot of good proposals that were direct subfields of AI and a set of struggling robotics fields in the wings. My initial proposal was rather hasty and it has come to light that there are a lot more related proposals out there.

After discussing with @eowl I came up with the idea to rename the AI & Robotics site to Stackbot with the hope that the abandoned ML proposal (proposer hasn't been around for a while) and other related fledgling proposals be merged in to one site that can accomodate a broader range of disciplines for all levels of community.

ML is on-topic for stats.SE and also cstheory.SE but these sites are pro-level and out of range for AI enthusiasts and hobby roboticists, I believe that Stackbot would be of greater significance to the committed communities.

Obviously we would wish to avoid treading on the toes of the expert sites, however I believe that the strength in SE sites comes from the ability to learn at all levels ... we can always redirect questions to the relevant sites as happens on SO. We probably already have a good idea of the list of sites to be included in the flag for redirection form ;)

This works for SO and SF and there is no reason why topics could be granted their own site after adequate audience has grown although I like that SO is a melting pot and thing Stackbot should be the same.

So in answer to your question. I think that the initial Computer Science topic was perhaps a rather scattered in its intent, as such I commend you on spotting this as a potential fit.

We already have SO, Programming.SE and CodeReview.SE, stats.SE, mathOverflow and electricalEngineering.SE and dba.SE ... there really isn't much meat left on the CS bone so to speak.

As SE communities should be more about inclusion, I say the more the merrier and would welcome the general CS proposal to be included in Stackbot (AI&Robotics) although I would not be happy with the merge being the other way around. For the simple reason that CS is a university course not a community which is what we are aiming for here.

link
1  
"there really isn't much meat left on the CS bone so to speak" -- disagreed. There is. "would welcome the general CS proposal to be included in Stackbot" -- Computer Science is most definitely not a subset of "AI & Robotics". – Raphael Jan 23 at 8:15
I didn't say CS was a subset of AI, but a lot of the content that particular proposal was looking to cover is indeed AI related. The rest is covered by other sites on the SE network. – Merlin Jan 23 at 12:29
@raphael wikipedia: Artificial Intelligence appears to many of the topics here. – Merlin Feb 7 at 22:19
feedback
up vote 0 down vote

Creating a combined "Computer Science" site has been attempted on many occasions. The larger community seems to want it, but it is almost always met with contempt and disdain by those participating in the individual subjects.

Area 51 does not handle this very well, so we've been going forward with these individual sites (Cryptography, Computational Science, CSTheory (merge rejected), AI (closed), and now Machine Learning) until one of the Computer Science communities stands up and has a cohesive conversation and comes to a consensus about creating a more comprehensive Computer Science site.

link
2  
I find your observation odd: The contempt and disdain I've seen about the CS site was from people who don't do CS, not from people involved in other CS-related proposals. You also seem to be saying that the best way to foster the CS proposal is to make many proposals about dubdomains of CS then merge them; I don't see how this could be a good way to proceed. Are you at least going to allow CS to go to commitment? (And by the way, please fix the description.) – Gilles Feb 7 at 13:46
2  
Computational Science is a completely different field (science done through computation as opposed to the science of computing). – Merlin Feb 7 at 17:59
There's also a huge divide between people doing CS as a university course and roboticists (who are still waiting for something Stack Exchange). I don't know why the original AI proposal was closed that was the closest we got to getting it right. Truth is we'll keep going round in circles with this. Years are passing and SE is only putting up barriers to communities while getting itself in a mess. Shame Jeff is stepping down, now is when we need him most. – Merlin Feb 7 at 18:04
@Merlin: Why AI was eventually closed: blog.stackoverflow.com/2010/12/…. Surely an AI proposal (or ML, etc) falls under "always welcome to try again" from that blog post. Is SE putting up barriers? This is a rare instance where we're letting the communities work out which proposals they want to support without disastrous interference. If you have an idea for how to build and promote the Grand Unifying CS proposal, please propose it. Everyone is free to switch their support and allegiances. We haven't been able to bring those groups together. – Robert Cartaino♦ Feb 7 at 18:44
@Gilles Yes, barring any new developments, the CS proposal will be allowed to proceed to completion. We (SE Inc) aren't suggesting anything about merging sub-domains or anything like that. We've tried to unite these proposals but have been unable to bring a cohesive group together in agreement. So we're just letting everything proceed as the community decides. I know... pretty radical, right? Let the community decide: blog.stackoverflow.com/2010/06/area-51-trusting-the-community – Robert Cartaino♦ Feb 7 at 18:49
@RobertCartaino barriers I was referring to are not being able to support cross-over topics ... people will search for and propose specific subjects which could be part of any number of super-sites (the mechatronics proposal graphic is the paradigm of this), is this recent CS problem so unique? The [Area51] model doesn't work in these situations (unfortunately). – Merlin Feb 7 at 22:33
@Merlin Stack Exchange doesn't claim to handle all topics well and maybe it turns out that this subject is simply not well-suited to an SE platform. I'm not ceding that point by any means, but Stack Exchange (and Area 51) were designed build new Q&A sites that work like Stack Overflow. The problems associated with pulling off interdisciplinary studies and knowledge sharing have long preceded Stack Exchange and folks a lot more studied than I have not been able to solve them seamlessly. So we'll muddle through and make adjustments and continue to add features that help the most people. – Robert Cartaino♦ Feb 7 at 22:53
good luck with that bob – Merlin Feb 8 at 1:04
I don't understand this answer. The way I see it, the general CS site is coming, provided no one arbitrarily closes it (as happened to the proposal from early 2011). So what you are saying is that no merges happen until a community clearly expresses its wish to do so? Note that your last sentence (in the answer) requests something that can (in a model like this) never be achieved (cohesion, consensus). – Raphael Feb 8 at 12:00
I don't remember anyone in cstheory community being against a general level computer science site, a site similar to the quite successful Mathematics and the younger Physics. I think you are mistaking the issue of merging a research level site like cstheory with such a general level/scope computer science site. These are quite different issues. Previous general CS proposals has been shot down by the larger community (the "not programming" in the title was there to make it clear for SO community that this is not a duplicate of SO). I believe this CS proposal is going to be a success. – Kaveh Feb 9 at 3:53
I proposed a merger of CS related topics & got shot down. its clear the area51 system has no mechanism to merge coalitions and many groups will just be left to "twist in the wind" (similar proposals lying around for many months, none reaching critical mass). this is not in Stackexchanges best interests therefore I suggest Stackexchange consider a different mechanism for group spinoffs-- say compartmentalizing existing groups somehow. not sure how! but it needs an architectural approach imho – vzn Feb 11 at 3:18
feedback

You must log in to answer this question.

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