5

Or is a computer science phd required? Is having a statistics phd a disadvantage for AI research positions or does the research you publish and work on during the phd matter more?

Although artificial intelligence research is interdisciplinary (math+cs+stats+logic, etc), I see mainly computer scientists working as AI researchers in industry at places like Google Brain, Open AI, DeepMid which makes sense since AI is an area in CS. Is it more because statistics phds aren't interested in working in AI or are these positions are mainly for computer scientists?

I am deciding whether to accept a statistics phd from a top school where a lot of good AI research is happening and the CS and stats departments are traditionally close. I believe I could work with an advisor who is in both departments or have one advisor in the stats department and the other in the CS.

Thanks!

  • 4
    Statistics PhD's can certainly do machine learning research, and machine learning is a big part of AI. A statistics PhD from a top school is a perfect background for machine learning, and you can collaborate with CS people who are doing machine learning. – littleO Mar 1 '17 at 6:12
  • Sounds like a great question to ask a statistics PhD, or a Machine Learning PhD, if your university has one or the other. – David Mar 1 '17 at 17:33
  • Sounds like a great question to ask somebody who hires into AI jobs. Statistics PhDs can say whatever, but if they are not hired into these jobs, then they are lacking something -- and worse, they are probably oblivious to what it is that they are lacking. It could also be that AI hiring managers are a..holes who don't know that other disciplines may be qualified for the job. Kinda hard to say, but it in some markets, cross-boundary hiring is a usual thing: Wall Street does not have any problems hiring theoretical physicists to solve stochastic PDEs. – StasK Mar 12 '17 at 21:45
  • Consider asking on math.stackexchange.com – Dave Kanter Aug 14 '18 at 19:25
  • It certainly will help. Some decisions are made based on probability of outcomes. But AI is much more than probabilities. AI is about analyzing disjointed facts and be able to infer certain things from them based on a set of basic rules. If you want to master AI, you must understand how Rete Algorithm works. As a mathematician, you should be able understand the mathematical aspects of machine learning probably better than a computer scientist. – hfontanez Aug 30 '18 at 18:24
7

Nobody cares what species of PhD you have. If you publish research in machine learning venues, you are a machine learning researcher. In particular, if you publish good research in good machine learning venues, you are a good machine learning researcher.

Just at my own university, I can think of machine learning researchers with PhDs in computer science, statistics, applied mathematics, pure mathematics, computer engineering, electrical engineering, mechanical engineering, astronomy, linguistics, library science, operations research, aeronautics, and economics. I'm sure I've missed a few. They all do different types of machine learning, but whatever.

The correct question to ask is Will a statistics PhD from [department] working with [advisor] support my ambition to work in machine learning? And the answer to that question depends on [department] and [advisor].

2

Late answer, but will take a stab at it. I can't speak for the big powerhouses you mention (Google, etc.), though I am a ML researcher and am involved in hiring for such positions.

Yes, a statistics PhD is a great background. Recall, not all AI/ML is neural networks -- Bayesian Statistics (e.g., Graphic Probabilistic Models) are highly useful and poorly understood. Further, there are many programmers who can easily learn how to use ML packages; however, statistics PhDs have more mathematical maturity. So, even if you want to work on deep learning, the mathematical maturity is a great selling point, so long as you also have the software skills.

Comparing a statistics PhD to a CS PhD is more difficult; however, I would focus on relevant experience / projects / skills / publications / interests rather than weighing the PhDs against each other. Larger companies with HR departments might not realize this, however.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.