Do what you love. Everything else will follow.

edit: realize this reads as "not very helpful". But you seem to know what you want to do. Job opportunities will be what they are in 5 years. Who knows. But if you try to change your natural trajectory, I think you'll be less happy, and ultimately less successful as a result.

I don't know how long it takes to finish a doctorate in machine learning, but if you claim to be able to finish in 5 years, I'll just wish you luck. Took me 5 1/2 to finish a doctorate in Algorithms and Data Structures (Theory), and I was very lucky with my problems and papers.

If you go to grad school, you are going to need to TA or teach, this will be good for you. You'll probably also need to grade homework, tests, etc. This will also be good for you.

If you pay attention, work hard, and have fun, grad school can be even better than undergrad. Depending on your workload, you may even have time to do contracting work to help pay for school, and/or earn some experience along the way. Be careful though, I knew a lot of people who dropped out to join companies, and who will never finish.

Most of the advice I've been giving out over the years I've written out here: http://www.ics.uci.edu/~jcarlson/gradschool.html

Good luck, whatever you decide.

Thanks! The 5 years estimate is on the hopeful side. I have been told my current course-load is insane by a few of my professors but it does not seem that bad yet (I am doing my bachelors in 3 years (currently on year 2), and hopefully overlapping with grad school on the last year (A honor student option here at Georgia State!)) I might be changing schools for my masters/PhD so there is a potential wrench to be thrown in gears there, a PhD might actually be harder work then I arrogantly expect it to be and take more then 4 more years. Either way, it looks like fun. On the flip-side it does have a cost, one I asked this question to help evaluate.

If you've not taken the GRE subject test for CS, then unless your program doesn't require it for entrance, it's going to be difficult or impossible to get into any of the top-tier schools (I don't know if that's your intent or not).

Unless your expectation of grad school is 16 hour days work every day for 4 years, then it will be harder than you expect it to be :P In all seriousness, work will ebb and flow. Some weeks/months you will be inspired and working whenever you're not sleeping or eating. Some weeks/months/years, you'll be uninspired, spending a few hours checking email, news sites, then sit down and play video games or watch TV.

Really though, those days intermix, and the sum total of the work that I did for my degree (aside from coursework for the masters degree) could have been completed in 6-9 months. The rest of that time was spent (in order of most to least time) relaxing, trying to figure out what to work on, campus politics, working on things that didn't make it into any paper (but which expanded my knowledge in other directions), and waiting on other people.

Edit: forgot the requisite infographic http://matt.might.net/articles/phd-school-in-pictures/

If you're interested in machine learning, it's hard to imagine how you will get beyond tinkering without an advanced degree. A PhD brings a depth and maturity when thinking about how to frame machine learning (statistics) problems, which I rarely have seen in people who haven't studied to literature at the PhD level.

If you do decide, look for the right advisor (http://www.cs.berkeley.edu/~jordan/) and try to go to a top program.

If you get a PhD doing research in machine learning, I'd be absolutely shocked if Google doesn't offer you a Research Scientist position.

I knew a half-dozen people with doctorates in machine learning at Google. None of them were research scientists.

I know lots of PhDs at Google who aren't doing research, but it's because they don't want to be doing research.

For a lot of PhDs that I knew, the problem was that they were already spending 100-200% of their time on their main work, which made it impossible to spend any 20% time working on an entrance into what they really wanted to do (typically research, not uncommonly tools that would have made theirs and others' work easier, faster, etc.).

I (and various other former Googlers) can attest to it being very difficult to do what you want to do within an organization. But that's the way it is in any company. Google just gives you access to a lot of other great people, resources, and perks that would otherwise be effectively unavailable.

The usual caveats: This is just one man's experience.

I have a PhD in Comp Sci. It took me 4.5 years after my BS to get it. (2 for MS, 2.5 more for PhD.)

I rather suspect that I don't make any more than I would have if I'd just started work 4.5 years earlier, but I do believe that the work I've done is much more interesting than what I would've found with just a BS.

OTOH, times were different then (mid-90's) than they are now. Arguably you could learn more about AI/ML if you went to work for Google or a startup right now than you would in school.

I guess my advice would be: Apply to grad school, but look for an interesting job at the same time. If you find one, take it.

No, at least not immediately. You can always go back to school, try and take a stab at work experience. My first real job was at a startup between my junior and senior year of high school. I loved it, and now I'm asking myself if my Bachelors is worth it.

Unless your pursuing something that you absolutely need the PhD for, don't worry about it for now. You might enjoy yourself more in industry.

25 is young. Many 25 year olds dont have much work experience. Just make sure when you do go for the PhD, you finish it and dont end up spending 4 years and no degree. 25 + PhD in CS = very employable.

Learn, machine, learn!