# What differentiates us from other similar websites?

The topics of our website highly overlap with the topics of CrossValidated and Data Science, but also with the topics of Computer Science SE, Stack Overflow and Philosophy SE (in fact, they even have an AI tag with currently 145 questions, while we barely have more philosophical questions, 169).

The differences between our site, CrossValidated and Data Science seem to be the focus, the users and their background, and certain topics. I think that a new and growing website, like ours, is attractive to certain people (including me) because it may represent an opportunity to show their abilities to others and maybe rule the website, while, in websites like CrossValidated, where there are already many established users, this may be more difficult. But does it really make sense to have all these separate websites (especially, CrossValidated, Data Science and ours), only because of these small differences?

It may happen that users on one of these sites may not be able to (properly) answer a question on their own website, but users on other related sites may be able to answer such a question. In those cases, the asker may not receive the help that, in theory, is available, but not directly accessible.

In order to understand if our website deserves to live, I think we need to enumerate the topics and goals of our website that really differentiate (or not) us from the other websites. Maybe we should really focus on the topics that differentiate us from the other websites. What do you think?

### Topics

Here's a preliminary list of such topics (I am using the tags below only to emphasize that these are on-topic here, but I am referring to the topics)

For completeness, maybe we should also list the topics that are on-topic both here and on the other sites.

Feel free to add more topics and goals that distinguish (or not) us from especially CrossValidated and Data Science.

AI has always been an interdisciplinary field. It therefore should not surprise us that AI.SE's content overlaps with that of other established stacks. I think this is essentially okay.

Perhaps as an analogy: The SoftwareEngineering.SE allows programming questions, but not of the same flavor as the StackOverflow main site. If you want to know how to do X in language Y, you visit StackOverflow. If you want to know whether to do X using language Y, you are better off asking on SoftwareEngineering.SE

If you want to know how to train a deep neural network in Python, you should visit DataScience.SE. If you want to know whether to train a deep neural network in Python (or, use any of the other various approaches in AI), you should visit AI.SE.

I think this means that a tag-based approach is the wrong one. We are likely to have questions that are about, say, statistical learning theory. This is part of AI. It is maybe part of Data Science, but I'd say it's a stretch. It is maybe part of statistics, but certainly not conventional statistics. It is definitely part of AI, and has been a core part for decades. Nonetheless, it encapsulates topics like support vector machines that are widely used in Data Science. We, therefore, oughtn't to outlaw the SVM tag. I think the same kind of argument can be used for most or all duplicate tags.

I'm especially concerned to see the machine-learning tag highlighted in the duplicates. Modern AI without machine learning is... not much.

I think if we focused only on the tags that are not present on other websites, we will not be able to claim to be about AI, and the site would (and perhaps should) then cease to exist. I think we'll do much better if we instead focus on claiming the why space.

• I think you didn't really get the point of my post. First, I've used the tags only to indicate that we cover certain topics (even though not necessarily in the same way as other websites). Second, I think we should just have a better and clearer idea of what is on-topic here and other related websites. Third, there is so much overlap between AI, Stats and DS that I am questioning the legitimacy of having three separate highly overlapping sites. This is the purpose of my post. – nbro Nov 6 '19 at 1:25
• @nbro Hmm. I was focused on this part "In order to understand if our website deserves to live, I think we need to enumerate the topics and goals of our website that really differentiate (or not) us from the other websites. Maybe we should really focus on the topics that differentiate us from the other websites. What do you think?" I think that we should not focus on the topics that differentiate us from other websites. Therefore, I also think it may not be a good exercise to find all overlapping tags. – John Doucette Nov 6 '19 at 1:28
• I've just edited my comment above. – nbro Nov 6 '19 at 1:28
• Well, I disagree with you. There are so many questions on AI SE that would have been better answered e.g. on Stats or vice-versa. I think it is a good exercise to clarify the topics that are on-topic or off-topic (here and in other related websites), in order to help the users in the most appropriate way. We shouldn't create a website for every hyped expression. For example, I think that Data Science SE is just the result of this. – nbro Nov 6 '19 at 1:31
• @nbro Hmm. Now I feel more confused rather than less. If the purpose of your post is "questioning the legitimacy of having three separate highly overlapping sites", I think my answer addresses it pretty directly: "why" should be on topic in general. "How' should not be on topic, unless it is specific to understanding something about AI, rather than applying an AI technique (e.g. "How can I inspect the weights in my neural network" seems on-topic to me, "How do I train a NN in tensorflow" seems off-topic). "What" questions depend. I agree that DataScience is hyped. AI is way older though. – John Doucette Nov 6 '19 at 1:34
• I agree with you that AI interdisciplinary field. However, I am wondering if it is really worth having these 3 highly overlapping websites (if I had to estimate, they overlap for like 90% or more). For example, if we merged Data Science and AI, wouldn't the community benefit more? There would be more interaction between the users and users could potentially receive more help. – nbro Nov 6 '19 at 1:45
• @nbro I think not. AI is broader than Data Science. In fact, in may ways, DataScience could be described as "Applied Machine Learning". It's relationship is this more akin to that between, say, Civil Engineering and Physics. Many AI researchers, even some quite famous ones, have said that AI has nothing to do with data (I disagree, but that doesn't mean they aren't part of the field). Likewise, CV is pretty clearly about statistics. Machine learning grew out of AI, and increasingly into statistics, but AI isn't either of them. – John Doucette Nov 6 '19 at 1:50
• CV is not just about "pure" statistics. It is also about machine learning and they accept questions related to reinforcement learning, causation (which is also important in AI), etc. Currently, "AI research" is highly based on data, (applied) machine learning, so maybe we should not have the DS site. – nbro Nov 6 '19 at 2:00
• Good point about SE:SofwareEngineering. When we revise the guidelines, we should probably mention that site as the preferred for "what language to use" questions in general (although I still think they should be on-topic here as well, in that good answers ideally would be factual, mentioning specific functions and capabilities.) – DukeZhou Nov 27 '19 at 21:09

Does it really make sense to have all these separate websites (especially, CrossValidated, Data Science and ours), only because of these small differences

No, it doesn't make any sense, because whatever people are saying on meta, in practice if you look at the questions posted on AI.SE, over 90% of them are on-topic on CrossValidated and Data Science. This creates plenty of crossnetwork question duplicates, which personally kills my motivation to participate.

• We can address futurism (one of the leading drivers of misinformation about AI!) and serve an important function of myth-busting.

• We deal with social impacts in general, which other related stacks don't address.

• We can take pyschology/cognitive/neuroscience questions related to AI, which may unwelcome on those stacks.

• We can treate AI milestones in general, not just those related to statistical AI.

re: Philosophy, although we have few questions formally containing that tag, a search of the most voted SE:AI questions reveals the subject to be popular and well-treated on this Stack. (Compare to the relative lack of activity for AI questions on SE: Philosophy, especially in recent years.) SE:Philosophy also lack a "neoluddism" tag, which is the more relevant philosophy tag, in that it relates to material effects of AI implementation, including bias.