Proposal: Artificial Intelligence

I find that there is a fundamental problem with this proposal: the quality of example questions is very low and won't attract experts, yet they are upvoted. Here are a few examples:

What's the difference between AI and genetic algorithms?

They are not even the same type of things. This is equivalent to asking what's the difference between aeronautics and wings: it makes no sense.

How emotional intelligence could be implemented in AI?

Way, WAY too broad and WAY too speculative. Emotional intelligence is not even well-understood for humans or animals yet. Answers to this will have to be variants of "like this according to me, based on 0 evidence".

This is a question that I wouldn't be suprised to see on WordBuilding, but not in a scientific site.

How is deep learning a step towards self aware AI ?

There are no claims that it is (if you discard pseudo-scientific clickbaits) (and self-aware is not a well-defined thing).


I could go on and on.

Moreover, all the "decent" questions could perfecly fit on CrossValidated, DataScience or Philosophy, depending on the question.

I am a Ph.D student in Machine Learning, and the only questions I would consider answering are the ones that would fit on other websites.

I strongly believe that, if the quality of questions if this site launches is the same as the quality of the example questions, the majority of users of this site will end up being non-experts that speculate without any basis on unclear questions, which would result in very low-quality content.

share
3  
But those questions still need to be answered, right? There are people who want to know the answer to these questions, and it's better to answer them rather than let them wallow in their ignorance. Somebody has to help out the non-experts with their non-expert questions. If that mean the site may have low-quality content when trying to help non-experts realize their "newbie" mistakes, so be it. – Tariq Ali Jun 10 '16 at 12:54
1  
The "emotional intelligence" question could also be useful. Consider a marketer wanting to use the buzzword 'emotional intelligence' to advertise a future AI product (maybe because it sounds new and exciting and could drive up sales). They have to build the AI product in question that would have this 'emotional intelligence', but the marketer doesn't know what 'emotional intelligence', much less how one can simulate it in an AI. So the marketer has a question that could then be asked on this site, and they can then get answers. – Tariq Ali Jun 10 '16 at 12:55
1  
@TariqAli Better questions need to be answered. You can rewrite the first example as two separate questions: "What does AI encompasses?" and "What are Genetic algorithms and what are they good for"?. My point is that most example questions on this proposal sound like thing you would see on clickbait websites. – Fatalize Jun 10 '16 at 12:56
    
Yeah, I do agree that we can write many of these questions to become better, non-clickbait questions (and so when someone asks a similar question again, we can redirect them to those previous questions then). So I guess we need to make a step to rewrite these upvoted questions and make sure they get answered properly, if and when this enters private beta. – Tariq Ali Jun 10 '16 at 12:58
    
@TariqAli "but the marketer doesn't know what 'emotional intelligence', much less how one can simulate it in an AI" Renowned experts don't know either, so this question is as meaningful as asking "What created the universe" on Physics.SE – Fatalize Jun 10 '16 at 13:00
4  
@TariqAli Two prior attempts at an AI site failed precisely because the site was full of “newbie” content and held no interests for experts. A site full of low-quality content tends to fill up with bad answers, because the people who could give good answers get bored quickly. – Gilles Jun 10 '16 at 21:48
    
I fully agree. As one data point that does not inspire confidence in this proposal, note that less than 6.1% of all comitters are active on any science site! It's unlikely that programmers and other enthusiasts can form a sizeable group of experts. (Individual examples notwithstanding; there may be AI experts active on SO but not sciencey sites, or not active on SE so far at all.) – Raphael Jun 11 '16 at 8:31
    
It's actually opposite. Our current AI proposal which started few days ago has enough experts, but not enough people to ask the right questions. – kenorb Aug 11 '16 at 1:05

This is a post in a discussion thread. Please bear with me that it's not an answer as in "definite answer". I start seeing some limitations to the standard SE format for such kind of "debate". Even the thread title is not a question.

Demystifying AI is to me the main target of this proposal. A Q&A format seems particularly relevant. I understand and agree with your concerns. Yet does it really mean the proposal and possible site are doomed?

You, as a specialist in ML, probably (pun intended) see that there are many unreasonable expectations with "AI" in the world. These expectations come from very serious people that just see AI as magic, wonderful SciFi authors, and others. I think something needs be done to better understand what to expect from "AI". This is not about "killing the dream"---we need the dream to go on. This is about clarifying the muddy situation we have been in for so long with AI. This means getting naive questions, whether we like it or not. There are a lot of naive questions, even on SO, that many want to answer with a fully capitalized RTFM.

Now I'd like we keep in mind that specialists (or experts, if you will) are the very ones who "kill the dream by making it real" (which is terrific). Getting real, McCarthy is customarily cited at this point: "As soon as it works, no one calls it AI." So we seem to end up in a situation where the best people to talk about AI-related stuff (the Experts) are also the ones who don't want to talk about "it" because it does not look expert enough (I mean, answer questions deemed naive by experts). One issue here is to mix "weak AI" that progresses fast these days and contributes to our daily routines, with "strong AI" that is just plain elusive. Isn't it still "AI", and the very term that some initiatives like this Q&A could help clarify?

So, is the seemingly very low quality of the example questions on your grading system the real issue with this proposal? I tend to think the framing is more of an issue (which is related, yes), and the expectations: AI is no magic, and this Q&A would be here to make it clear.

I really like your final remark: "the majority of users of this site will end up being non-experts that speculate without any basis on unclear questions, which would result in very low-quality content." I think that describes why showing that AI is no magic matters: To curb speculations and "correct" understandings.


A couple extra comments:

  • "What's the difference between AI and genetic algorithms?" If some people are confused with the terms, you and I would probably answer "go read Wikipedia or a text book". Fair enough. Now, if there is a confusion that is upvoted, we may also see it as a mission for "people who know" to clarify it. Disclaimer: I don't like this question.
  • "This is a question that I wouldn't be surprised to see on WordBuilding, but not in a scientific site." Great point, related to the proposal framing: Is Science the right category? Science may entail that people expect "definitive" answers (as in state of the art). I chose the Science category at first, as in scientific method. I would like to see the proposal going toward enlightening people with logical (hypothesis, try, measure, repeat) statements, in other words dealing with "non-experts that speculate", as well as "expert" questions.
  • "How is deep learning a step towards self aware AI ?" Bingo. This question is pure mud we want to clear out. So many people get confused with this one. Wouldn't something like a Q&A help with such kind of confusion? And I would be thrilled to hear about alternatives too.
share
2  
Interesting point of view, +1. – Fatalize Jun 20 '16 at 6:38
    
What typically happens in these stations is not clearing of muddy waters as you hope but endless speculations and pure subjective opinions. Idle curiosity questions do not end up creating a site with high quality answers. – Kaveh Jun 20 '16 at 15:44
1  
@Kaveh I read you on the "idle curiosity". Yet I prefer to do something rather than stay idle myself and keep on hearing a lot of non-sense about AI. This Q&A may well fail if it does not get enough momentum, so be it. If we can at least get a few people understand more what one can expect from AI, what it was, what it is, and where it is heading to, then I believe we win some brownies. And if we catalyze some good energy to get momentum, I dare hope for the brownie factory. – Eric Platon Jun 22 '16 at 2:23

You must log in to answer this question.

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