I think it would be useful to allow questions on hardware, since we're the general AI forum, and this is relevant.

(In other words, it's better to have AI-specific hardware questions asked here, b/c other AI developers will likely have similar questions & issues.)

In the past we have had questions about the type of hardware various algorithms that have reached AI milestones have used.

It's important to understand the relationship of hardware to software in that strong Machine Learning was only possible once there was sufficient memory and processing power.



since we're the general AI forum

We are not a general AI forum. This forum supposedly exists to fill a certain gap. As stated in this answer

It's because the OPPOSITION against creating this site argued (correctly) that we already created sites to handle this subject explicitly. The argument FOR creating this site claimed that we have a missing socio-scientific angle that needed filling.

I believe that ALL implementation questions, such as "Can you explain the parameters of this ML program?", "Why isn't my ML program working?" or "How do you implement this model?", are OFF-TOPIC. They would be on-topic, if we merged this site with Data Science (aka applied machine learning). Similarly, there are already sites for hardware and software (which already has the tag ai) recommendations. There is absolutely no need for duplicating services, which are available somewhere else.

Therefore, I strongly suggest we focus on the social, scientific and theoretical aspects of AI, otherwise, we'd better just merge this website with other websites. Do we want to have a website that 95% overlaps with another website only because people disagree on the meaning of the expressions "artificial intelligence" (or "machine learning") and "data science"? There are so many theoretical questions that have not yet been asked. For example, there could be a lot of questions on AIXI, which is a highly mathematical and theoretical topic (that is, a perfect topic for this site), which is not easily understandable, so I would expect a lot more questions, but we only have 2 questions.

Unfortunately, this website has already taken the wrong direction, IMHO. We already have a bunch of implementation, hardware, and software-related questions, which is partially due to the fact that the community and moderators do not take action with respect to the original goals of the site, which has become quite redundant.

However, there are also topics on other sites that would be better suited for this site, such as reinforcement learning on Stats. To conclude, apparently, there is a lot of duplication of services across sites. Maybe there should be a way of migrating even old questions from one website to the other, as a way of organizing better the communities. For example, there are a lot of theoretical ML questions on Stack Overflow, which could be migrated to this site or Stats.

  • 1
    $\begingroup$ The underlying premise of this answer is sound--stick to defined parameters--but I don't think the close/merge line of justification is valid. We're a healthy stack, not at risk of closure, so it's more a matter of what makes the most sense in terms of the utility we provide. You make a fair point about hardwarerecs.stackexchange.com although that stack is still in Beta, and not performing particularly well: area51.stackexchange.com/proposals/65287/… $\endgroup$ – DukeZhou Nov 15 '19 at 22:31
  • $\begingroup$ @DukeZhou As I said before, almost all questions have utility. We cannot just base our choices on utility, otherwise, everything will be on-topic here! As I say in my answer, if we decide to accept all AI-related questions, then we'd better just merge this site with Data Science. What are your arguments against not merging this site with Data Science (if we decide to accept implementation, hardware and software-related questions)? The arguments will be that we also address philosophical and social issues, which is not the majority of the issues that we address. $\endgroup$ – nbro Nov 15 '19 at 22:39
  • $\begingroup$ Currently, we have 173 philosophical questions and 47 questions tagged with social. Of course, there are other tags related to these topics, but the tags with the highest number of questions are neural networks, machine learning, deep learning, reinforcement learning and CNNs, which are all also on-topic on Stats and Data Science. To me, this is a sign of the redundancy of one of these sites. $\endgroup$ – nbro Nov 15 '19 at 22:44
  • $\begingroup$ The scope of Data Science is much narrower than ours, and only involves a specific type of AI (statistically-based.) datascience.stackexchange.com/help/on-topic. Interesting quote from their help page: "Data science is a multi-disciplinary field and many new users wonder whether their questions are most appropriate here or on other SE sites" and "If you think a question is equally appropriate on multiple sites, ask on the site with the most users (usually Stack Overflow or Cross Validated)" $\endgroup$ – DukeZhou Nov 15 '19 at 22:45
  • $\begingroup$ Since we're still an emerging site, on the cusp or graduation but not fully graduated, I tend to respect contributor preferences. There are clearly users who like hardware, and even implementation questions, in that such questions are reliably upvoted. (By contrast, where I see multiple close votes, i usually "tip the scales" and close.) $\endgroup$ – DukeZhou Nov 15 '19 at 22:47
  • $\begingroup$ @DukeZhou Well, AI is also a multi-disciplinary field. Basically, everything on Data Science could be considered AI, based on the definition of narrow AI. I am not saying that our site is redundant. I am saying that either our site or Data Science are redundant. $\endgroup$ – nbro Nov 15 '19 at 22:48
  • $\begingroup$ Per the DS help advice, in terms of hardware specifically, we have more active users and much more traffic than Hardware Recommendations. $\endgroup$ – DukeZhou Nov 15 '19 at 22:48
  • $\begingroup$ @DukeZhou Well, I am personally not against implementation, hardware and software questions, but there's a lot of redundancy here! Why should we have 2 sites on the SE network that basically do the same thing? It is unproductive! It is like having multiple political parties for each unhappy person. If you disagree on even a tiny aspect, you may decide to create your own political party $\endgroup$ – nbro Nov 15 '19 at 22:51
  • $\begingroup$ Data Science is more general than just AI, as the field has other applications. AI ∈ DS; DS ⊂ AI. $\endgroup$ – DukeZhou Nov 15 '19 at 22:53
  • $\begingroup$ @DukeZhou How is DS more general than AI? To me, DS is just a buzz word, which essentially means statistics + AI and AI = statistics + logic + heuristics + programming, etc. $\endgroup$ – nbro Nov 15 '19 at 22:53
  • $\begingroup$ Because there are statisticians and analysts who are data scientists, but not AI developers/researchers. (The field itself is rooted in statistics and analysis, and only recently, Machine Learning.) $\endgroup$ – DukeZhou Nov 15 '19 at 22:56
  • $\begingroup$ Basically, the expression DS only exists because of (the successes of) ML. Of course, there are people in AI that do research on different sub-topics, but this happens in all fields. For example, not all physicists study particle physics. Some of them study or focus on theoretical physics. $\endgroup$ – nbro Nov 15 '19 at 22:59
  • 1
    $\begingroup$ The Data Science field may have coalesced with the advent of strong-narrow ML, but it existed in prior form in the general field of analytics. Data Mining was an exceptionally hot field for about a decade, prior to the ML revolution, per the rise of the internet and social media. But this is tangential to the question here posed. $\endgroup$ – DukeZhou Nov 15 '19 at 23:07
  • 1
    $\begingroup$ I agree with this answer. I think expanding to allow hardware-related AI questions, at this point, would make things unwieldy. Hardware-related questions are a different ballgame, and, at this point, it's probably better to remain focused. $\endgroup$ – The Pointer Dec 16 '20 at 8:44

You must log in to answer this question.

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