Nidhish Alex

Artificial Intelligence is not about just scientific distributions as most people may think. They even have no clue about the difference between data science and AI. R is good for these kind of scientific computations/analytics and building models for data science. For other areas of AI like NLP, Machine Learning, Text Mining, Information Retrieval/Extraction, Speech Processing etc Python and Java are the best to go for. When Java has a lot of good libraries Python is good in terms of ease of programming. 

I would suggeest Python because it is simple and there are Python wrappers for those Java libraries as well nowadays.

There is no definite answer to this. It depends on the type of webs service you would like to serve, what are the backend modules to be loaded etc.

If your backend modules run as JAR files then it is simply irrelevant to write a Django module above that. Same is true for other languages as well. But still it depends on your comfort level. In our startup we use Python front ends even for these JAR modules since both the teams work independently and we feel comfortable to write server modules in Python framewroks.

Not many organizations have moved out of Python 2. In our organization most of our programs are in P2. This is true for other organizations as well. 

Even if you learn Python 3 you will still have to do a lot of your work in Python2 if you are working on old libraries. So it is always good to learn Python 3 along with Python2. For a beginner there will not be a big difference between those two versions other than this simple code changes like putting a bracket in print statement etc. This is pretty easy to remember. If you are using distributions like Panda or Anaconda then start learning P3 and migrate to P2 whenever required. Since these libraries are more compatible with P3.