4
$\begingroup$

If we look at the curriculum of AI in any university, we will find as topics: agents, neural nets, supervised/unsupervised learning ... . In a few ones, an small part about AGI or mind.

It is not surprising that this site receives (and answers) a lot of questions about previous subjects. All the more because only a 0.0001% of people (not me) will thing in a site called "cross validated" to ask about neural nets.

However, the description of this site talks about "conceptual questions about life and challenges in a world where cognitive functions can be mimicked in purely digital environment." that not only excludes all previous, but even AGI seems off-topic.

This fact means that we should reject 99% of the questions and flag them as off-topic instead of answer them doing in this way an off-topic answer.

In my opinion, there are a disagreement between name and site description. More opinions are welcome ...

$\endgroup$
  • $\begingroup$ I don't think you can have an AI site without questions on NN or ML. Although a separate site exists does not mean we cannot admit questions on ML. The same question arose on CV.se and datascience.se. And if we see stats.se, dataS.se and Cv.se are much more overlapping than AI.se. $\endgroup$ – DuttaA Mar 12 '18 at 4:09
4
$\begingroup$

When AI.SE was about to be created, there was a divide. A few wanted ML and implementation details to be part of the site; most wanted to exclude them. The final agreement was to exclude---so the current topic list excluding explicitly implementation details and so on.

The reasoning back then was that popular frameworks like Tensorflow, etc. were explicitly asking to question on SO. Questions about Data Science were perceived as much better fits for Cross Validated, etc. So no need to duplicate them here, and take the risk of killing AI.SE before it gets momentum. That was an interesting thinking, and I think the result is okay.

Now it seems that many AI.SE questions are about ANN, including popular subcategories like CNN, DNN, RNN, etc. The community here seems to expect xNN Q&A, so the description and topic list do seem like a mismatch---I concur (and that is how I found your question, searching for such a discussion). Another way to put it: How can we "make a mind™" without talking about techniques and tools we have at present?

IMHO it may be time to update description and topic list, to define a posteriori the scope of AI.SE.


Note:

  • The current description does include AGI, though. Terminology is quite vague, but an AGI could be a "set" of "cognitive functions [...] mimicked in purely digital environment." As for theory, models, notions, concepts, the current scope has been carefuly thought through, IMO.
  • Let's keep in mind that AI.SE is the 3rd attempt to create an AI-related SE site, and the most successfull so far (the previous attempts topped at 6 months before closing). This AI.SE is on something (and the current "market" makes it easier with the DL wave).

Disclaimer: I was part of the first group, so I am biased in my agreement. However the sheer volume of xNN questions might be a data-backed confirmation the site needs an update (I did not go beyond lazily listing and eye-balling unanswered questions).

$\endgroup$
  • 1
    $\begingroup$ I think sometimes to be successful we have to be versatile and do the cumbersome jobs also...that is why this site is going on well...even area 51 tells that the questions per day is increasing $\endgroup$ – DuttaA Apr 13 '18 at 7:02
  • $\begingroup$ More than sometimes... $\endgroup$ – Eric Platon Apr 14 '18 at 3:32
  • $\begingroup$ Deep Learning must be added (I think Machine Learning as well). $\endgroup$ – Royi May 11 '18 at 14:13

You must log in to answer this question.

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