Suppose my goal is to collaborate and create an advanced AI, for instance, one that resembles a human being and the project would be on the frontier of AI research. What kind of skills would I need?

I am talking about specific things, like what university program should I complete to enter and be competent in the field.

Here are some of the things that I thought about, just to exemplify what I mean:

  • Computer sciences: obviously, the AI is built on computers, it wouldn't hurt to know how computers work, but some low-level stuff and machine-specific things do not seem essential, I may be wrong of course.

  • Psychology: if AI resembles human beings, knowledge of human cognition would probably be useful, although I do not imagine neurology on a cellular level or complicated psychological quirks typical to human beings like the Oedipus complex would be relevant, but again, I may be wrong.

  • $\begingroup$ This question seems to be very similar to this, although I am not sure if you focused on AGI or not, but you certainly mentioned "human-like AI". $\endgroup$
    – nbro
    Commented Dec 21, 2021 at 15:30

2 Answers 2


As a full-time AI researcher myself, I'd say that a PhD in machine learning would certainly be one useful option.

However, in order make much-needed progress, AI needs avoid falling into the trap of thinking that currently fashionable methods are any kind of 'silver bullet'. There's some danger that a PhD that heads straight into (say) some sub-sub-sub area of DL would end up imposing too much bias on the student's subsequent perspective.

AI research is an essentially multi-disciplinary activity. Other possible backgrounds therefore include:

  • Mathematics or physics (to first degree or PhD level). A strong background in either of these never did anyone any harm. People who are competent in these fields tend to be able to turn their abilities to new domains relatively easily.

  • Software Engineering. One of the things that AI needs are integrative architectures for knowledge engineering. Here's why. I believe that one of the reasons that we haven't yet managed to do OCR at the level of a 5 year old is that we've yet to accept that we have to 'build a sledgehammer to crack a nut'. Software architects are used to managing large-scale complexity, so they may be able to help.

  • Cognitive Science, Psychology, Cognitive Linguistics. The reasons here are obvious.

Above all, I personally think that a good AI researcher should be creative, inquisitive and prepared to question received wisdom, all of which are more important in practice than specifics of their background.


Research on AI seems to be getting wider these days (2016). First, "obvious" few departments (no order):

  • Computer Science (e.g. computation theory, algorithms): AI researchers there assume that intelligence is a kind of computation, under various forms (e.g. a neural network, a logic system).
  • Software Engineering: Assuming we find a good model for AI, how do you make it? This is what the engineer will want to figure out. And it can be hard to map mathematical models to an engineered piece.
  • Statistics and Probabilities (more specific than just Mathematics, which is also close to Computer Science): This is about Data Science, notably as a foundation to Machine Learning, the most active branch in AI---which "just" covers the learning part.
  • Physics: This is particularly relevant now for hardware (see below).
  • Neuro Science: Understand how the brain works, as an inspiration to create an artificial one, is the home for Connectionists. Recently, Hassabis and his team at Google Deepmind made several breakthroughs related to reinforcement learning, memory, attention, etc.

Recently Electric Engineering is getting a lot of light, together with the related branches of Physics. Several public and private laboratories focus on "brain chips". To name a few: IBM (who's working on that for some time already), Nvidia, and Facebook. Circa 2010, it became clear that techniques like deep learning require horsepower, thus an increasing focus on creating more powerful, smaller, more energy efficient chips. And on top of that, there is all the work in Quantum Computing.

But the thing is, there seems to be many more fields that are getting involved in AI research. We should mention Chemistry and Biology, as both inspiration and tools to make new models or hardware (e.g. chips that do not use silicon, so they can get smaller).

As for 2016, the above fields are the most active, and promise to remain very active for quite some time. Pick your own depending on your interest, skills, or mere intuition!

To finish, we may be surprised in a few years when we look back at where AI has come from. I believe that if we manage to build an AGI, it will leverage all these fields anyway. I guess the thrill is to be part of the story.


You must log in to answer this question.

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