The first impression might be, that Artificial Intelligence is hot topic in the internet and it’s easy to identify relevant blogs. The problem is that most of so called AI blogs are discussing a slightly different topic or are forming their own community. A typical example are the blogs of phd students. Many of them are available in the internet. The typical blog contains of a large publication section in which the phd student has posted the papers he has published at arxiv in the last 5 years. The problem is, that the phd community is forming it’s own group which isn’t working with blogs but with academic reputation.
The second example are large newspapers blog in which journalists have written article about the social role of AI. They are announcing, that Google is using a new sort of neural network or that robots will become important in the industry. I would call this sort of information mainstream AI, because the aim is to explain to public how the world looks like. The intention of such blogposts is not to discuss the details.
The third form of AI related blogs are not located in computer science but in the humanities. The idea is to describe AI from a philosophical point of view. A typical question is, if machines can think, if language is separated from consciousness or what Singularity means. This is also not real AI because, the aim is not program software.
After this negative example it make sense to define how the perfect AI blog will look like. The best example I’ve found in the internet was a python tutorial website, in which a programmer has explained how to realize the game AI for a snake game. In this post everything was made right: AI was seen as a programming topic, secondly the intention was not to publish a paper at arxiv but to write a blog post and third the subject was explained to the newbies who are not familiar with the technology already.
Before a blog can write about AI, the previous step is to write about Python game programming. This kind of amateur community is trying to realize short programs on their desktop PC which looks like Snake, pong or Tetris. The community who is doing so is large, and they have published lots of tutorials what they are doing.
If this idea is put to the next step the next question is how to realize the AI for such a minigame. This topic is not very often discussed in the internet and only a small amount of people are talking about it. If we can answer who the problem is solved, we can identify the relevant blogs who are doing so well and not so well. These blogs are equal to the AI community. Their aim is not to discuss singularity, not to write Arxiv papers, not to talk about philosophy, but the focus is much more simpler: how to write on top of a python game the AI?
Sure there are many so called papers available from the universities how have answered this question on a theoretical level. But the authors of these papers are not equal to the ai community. What they have done is to create content in an academic environment. Which means they are in their own world outside the blogosphere. The same problem is located around blogs who are dedicated programming blogs. They are explaining what Python is, and how to program a game in C++. This has also nothing to do with the AI Community. Because programming is different from AI programming.
Definition of the AI Community
As a work hypothesis, the definition of the AI Community is, that it’s about weblogs in which tutorials are published about implementing Game AI in amateur games, written with Python. This definition is based on the following elements: weblog, tutorial, Game AI, Python. Everything which doesn’t fit to the definition is not part of the AI community.
Let us make some thought experiments. A new arxiv paper in which a neural network is explained doesn’t fit to the definition, because it’s not a weblog but a peer reviewed paper. A blog in which amateur philosophers are discussing about Artificial Intelligence in general doesn’t fit too because it’s not about Python nor Game AI. The barrier to enter the AI community is high and the number of blogs which are belonging to this definition is small.
The good news is, if the definition is very strict it is much easier to get an overview over all the blogs in the world who are fitting to this definition. It is realistic to identify them all and make a small list. I would guess that the amount of blogs on the list is smaller than 100, perhaps the number is between 10-20. I don’t know.
But why is the definition so strict? Wouldn’t it make sense to include more content which has to do with AI in a more general way? Let us analyze what will happen if we see the Arxiv papers as part of the AI community. The problem is, that the papers submitted to the academic community have a different kind of aim than a blogpost about Game AI. In most cases, the papers doesn’t describe Narrow AI but they are created around mathematical formalism. If a papers contains lots of university related equation the reputation of the author gets much higher. Unfortunately this reduces the number of people who understand the paper. This contradicts the idea of a community to teach a topic to newbies who are not familiar with AI before. The entry barrier of academic journals is too high to become part of the AI Community.