Math Job in Artificial Intelligence with a Math Degree

[tmbrwlf730]

I'm currently going for my Master's in math (bachelors in applied math) but while applying I had to make the choice between math and computer engineering. I went ahead and chose math because when looking at job postings it seemed like it was an acceptable degree for several of them. Almost all of them actually, unless the word engineer was in the title.

I'm interested in going into artificial intelligence and wanted to see if anyone with a math degree actually landed a job in the field and what kind of things they do. From those people with experience in the field, is it better if I try to switch over? My university offers more pure math courses (Topology, Lattice Theory, Differential Geometry for example) than applied ones at the graduate level. It offers some statistics courses but it's not offered as it's own degree. Even then some are are statistical theory which I'm not too familiar with.

Any suggestions or advice would be welcomed. Thank you.

 

[Hercuflea]

Can you make your master's have an emphasis on statistics and machine learning? Take courses/specializations in that? Do a few projects. That will make it more marketable.

 

[Aufbauwerk 2045]

As far as education, assuming you are interested more in AI programming than AI hardware, I think computer science courses will be most valuable. AI after all is a branch of computer science. Math courses can be useful depending on the application but programming is about algorithms, data structures, and programming languages.

Since you enjoy math, you may want to look at automated theorem verification and automated theorem proving. Automated theorem proving has some important applications such as hardware verification and software verification.

But I suspect topology is hardly used by most AI programmers! Unless maybe they are applying AI to math or physics.

Learning about heuristic search algorithms, pattern detection, and so on, is of much greater importance. As mentioned already machine learning is also an important topic.

Of course math skills such as Fourier analysis can come in very handy if you are working in speech recognition or speech synthesis. That's also true for image processing, such as facial recognition software, which is often in the news. So maybe you could look at AI jobs which are related to signal processing.

These are just a few ideas. AI is a huge and growing field.

Have you read any books on AI? Ray Kurzweil has written some interesting books on the subject.

 

[MarneMath]

AI is a rather broad field and there's many different approaches to it. There isn't one particular "best" path to it. Some people are real into the algorithm and design. Others are into the Statistical Decision Theory and deriving probabilistic graphical models. Most people tend to fall somewhere in the middle. The key aspect is to keep working towards it by taking classes in Optimization, Statistical Theory, Algorithms, Complexity, etc. I would also emphasis that if you truly want to work on AI then a graduate level education is still rather essential.

*Disclaimer naturally other people venture into the field by means of Neural Science Physics etc. The common trait I find among people who work on AI or AI related field is a wide breadth of skills they tend to have. Ie, it's rare to find someone who is strictly theoretical and it's also rare to find someone who is completely applied. There tends to be a balance that has to be mixed between theory and complexity.

 

[dumbperson]

Sorry for asking this here, and I know there's a lot of posts on the internet answering exactly this question but I'm curious about the opinion of you smart eople!
Does anyone have a preference for any sources to get into AI as a theoretical physics student? such as books, online courses, etc. :)

And with books i'd prefer a book that would have programming problems/challenges etc instead of just reading!

 

[Aufbauwerk 2045]

You might have a look at Tanimoto, The Elements of Artificial Intelligence Using Lisp. Since you want a book with programming problems, I think this one is a good place to start. One nice thing is that he only uses a relatively small subset of Lisp, which shows you just how much you can do without knowing the whole Common Lisp language.

I also like Cawsey, The Essence of Artificial Intelligence. This one gives you a brief overview. It's not oriented towards one particular programming language.

There are newer books, but I still find these valuable.

By the way, I made a detour from physics for some time to get involved in AI programming. For what it's worth, I find it more satisfying now to stick with physics. My main use of computers now is numerical programming.

There is lots of hype around the topic of AI today just as there was a few years ago when I was a student. I think it's good to be skeptical. Of course one should try to specify what one means by "AI." If it's computer algebra then this is mainstream now and if that's what we mean by AI then fine, it's OK. But some topics seem pretty speculative. For any given topic, I suggest comparing the reality of AI today with the hype before spending a lot of time on it. Think about what AI is attempting to accomplish and whether it even makes sense to achieve this through programming a computer.

IMHO nothing beats physics.

 

[Aufbauwerk 2045]

P.S. just to follow up on my previous comments, I admit I have been working in AI for several years. The more work I do in this area, the more I believe in it. But my goal is to use AI to help us solve problems in math, science, and engineering, and my favorite area is physics. This is a very difficult but still a limited and I think a realistic domain for AI. It has nothing to do with creating artificial brains out of neural networks. Not that I disparage interest in the latter, but it's a very immature and overly hyped technology.

Very recently I decided to halt a physics project I was working on, and devote all my time to AI. But for now physics is still my main interest when it comes to applications. Perhaps one day it will be biology.

We are still at the early stages of AI. It feels to me like we have developed the abacus, and now we are working on the calculating machine.

I was still a bit conflicted when I posted on Dec. 7. I enjoy solving problems with my own mind, and not needing to rely on a computer. But let's be realistic. The unaided human mind is pathetically weak compared to what we can achieve with AI.

So we may as well embrace AI, while being careful to control it. There is of course the question of whose AI will dominate.