How accurate do you think the following diagram is?

@nbro: I think "Artificial Intelligence" is very movable, depending on definitions. Perhaps a lot of recent "AI hype" would be viewed like the diagram where AI is built over successful ML using neural networks, such as DQN. But the subject in general is more orthogonal to the stats/data/coding view, and hard to pin down in a diagram like this.

I think "big data" is coming in to AI as a way of training perceptual and acting machinery, but is not core to it. Depends partly on whether you want to show current practical skillsets required to work in areas, or conceptual overlap between subjects?

For instance, right now, if you wanted to join a self-driving car project, you might be expected to have technical knowledge necessary to process terabytes of video and sensor data. But it seems likely that learning agents with capabilities close to mammalian intelligence do not need data at that scale.

So Big Data looks like a practical requirement for some "AI today", but not necessarily a requirement for "AI in general", especially if we're talking about theorists and constructing next-generation AI

@NeilSlater Btw, what's your view regarding the difference between statistics and data science? After having read a few articles regarding the topic, it looks to me that data science is a broader subject, or, at least, some data scientist claim that. Statistics seems to be a more rigoruous subjecct from the mathematical point of view. Honestly, I think this introduction of multiple terms, which refer 99% of the times to the same subject, is only confusing.

At the end of the day, machine learning is also statistics. I don't see the point of renaming a whole field only because it's applied to a specific context.

Of course, machine learning has a slightly different approach to statistics... However, I feel that many people that work in the field of AI would only benefit from having a more solid background in (classical) statistics

I'm not sure I have a strong view on that. I agree with the idea that a lot of data science is applied statistics. Definitely good stats theory is going to help someone in data scientist or business intelligence roles. Same is true to a lesser extent in many other experimental sciences as well