14 ms·

Some on this thread have recommended Norvig's PAIP, but that's kind of an old school AI book in that it focuses on heuristic search and logic (implementing prol

by krosaen 10y ago

Some on this thread have recommended Norvig's PAIP, but that's kind of an old school AI book in that it focuses on heuristic search and logic (implementing prolog in lisp at one point, very impressive stuff actually); but is lacking any coverage of statistical machine learning, which is the approach that underlies most of the cool stuff these days. It's still a great book, but I'd instead recommend a path that focuses on machine learning: - The Master Algorithm: made for a general audience, gives you a lay of the land - Python Machine Learning by Sebastian Raschka: gives you practical skills using python, scikit-learn, numpy, jupyter notebooks, pandas etc. From zero to kaggle in 4 chapters, goes deeper after that. Also goes into enough theory you aren't flying completely blind. After that, I'm afraid I think you do need to go "academic", if by that you mean learning some of the underlying math to approach AI / ML from a more rigorous probabilistic perspective. I'd recommend studying probability theory and then working your way through Bishop's Pattern Recognition and Machine Learning. After that a lot more doors open up too more specialized topics like computer vision, reinforcement learning etc. I've written up a lot more about this here: http://karlrosaen.com/ml/ http://karlrosaen.com/ml/

JamilD 10y ago
You don't need to focus on just one side or the other. AlphaGo wasn't made by just hacking together a couple neural nets, it built on heuristic search, MCTS, and the "old-school" AI. I agree that basic statistics + Bishop's book is a great way to start getting into machine learning -- but AI is a much broader field than that.
- taktoa 10y ago
  "old-school" AI is just a pretentious name for algorithms/graph theory/combinatorics, plus a bit of (now very outdated) PLT mixed in. "new-school" AI (machine learning) is just a more pretentious name for statistics/control theory/randomized algorithms.
  - nojvek 10y ago
    What is to say the human brain doesn't engage in graph combinotrics with some learned heuristics. Some of the old ideas with enough computation power are actually pretty amazing.
    - goatlover 10y ago
      Because the human brain isn't a symbolic machine. It's a living organism made up of cells trying to survive and reproduce in the real world. All our symbol manipulation capabilities are built up on older abilities based on feeling, experiencing, socializing, and moving about.
dominotw 10y ago
Hi Karl, I just read through your very intersting post about the sabbatical. I 've been thinking about this a while too. I see that you are currently in a A.I role at umich, I was wondering you have written anything about how you went about actually landing a role in this particular field. Thank you.
- krosaen 10y ago
  Long story short, I became aware of the position through networking in the tech/startup community in Ann Arbor, and eventually got introduced to an ML prof who knows the profs who are actually running the lab. I applied to the position with a resume and cover letter customized to relevance of the role (compressed a lot of startup / product / leadership stuff down, went into more details on tech stuff, emphasized ML studies, excitement for mission of their lab) and met with the profs a few times and it worked out. If I can extract any advice from this, it's that putting yourself out there and letting the world / your network know that you are interested in something or working towards something (in my case: a transition to a career applying ML), things might turn up. Also: if you have more experience you should feel comfortable completely customizing your resume to the role so you have one page jam packed with relevance; it's ok if they don't see (or won't care) about some of your experience. I also notice that in some of your previous HN discussion you lamenting companies not really looking at your open source work; it's annoying that they wouldn't take the chance to look. But if I were you I might highlight specific projects on your resume relevant to the role you are applying to if you haven't been doing this already; this could elevate your open source work to job experience in its emphasis. Assume 99% of people will only see your resume, everything else should be supporting resources in case they get interested enough to look (or wish to validate your claims).
EternalData 10y ago
+1 to Python Machine Learning, I feel like some of artificial intelligence is understanding a lot of finer (but critical) mathematical nuances, but some of it is just getting your hands dirty, and the second is definitely more accessible for somebody who is starting to learn.

by krosaen 10y ago

JamilD 10y ago

You don't need to focus on just one side or the other. AlphaGo wasn't made by just hacking together a couple neural nets, it built on heuristic search, MCTS, and the "old-school" AI. I agree that basic statistics + Bishop's book is a great way to start getting into machine learning -- but AI is a much broader field than that.

taktoa 10y ago
"old-school" AI is just a pretentious name for algorithms/graph theory/combinatorics, plus a bit of (now very outdated) PLT mixed in. "new-school" AI (machine learning) is just a more pretentious name for statistics/control theory/randomized algorithms.
- nojvek 10y ago
  What is to say the human brain doesn't engage in graph combinotrics with some learned heuristics. Some of the old ideas with enough computation power are actually pretty amazing.
  - goatlover 10y ago
    Because the human brain isn't a symbolic machine. It's a living organism made up of cells trying to survive and reproduce in the real world. All our symbol manipulation capabilities are built up on older abilities based on feeling, experiencing, socializing, and moving about.

dominotw 10y ago

Hi Karl, I just read through your very intersting post about the sabbatical. I 've been thinking about this a while too. I see that you are currently in a A.I role at umich, I was wondering you have written anything about how you went about actually landing a role in this particular field. Thank you.

krosaen 10y ago
Long story short, I became aware of the position through networking in the tech/startup community in Ann Arbor, and eventually got introduced to an ML prof who knows the profs who are actually running the lab. I applied to the position with a resume and cover letter customized to relevance of the role (compressed a lot of startup / product / leadership stuff down, went into more details on tech stuff, emphasized ML studies, excitement for mission of their lab) and met with the profs a few times and it worked out. If I can extract any advice from this, it's that putting yourself out there and letting the world / your network know that you are interested in something or working towards something (in my case: a transition to a career applying ML), things might turn up. Also: if you have more experience you should feel comfortable completely customizing your resume to the role so you have one page jam packed with relevance; it's ok if they don't see (or won't care) about some of your experience. I also notice that in some of your previous HN discussion you lamenting companies not really looking at your open source work; it's annoying that they wouldn't take the chance to look. But if I were you I might highlight specific projects on your resume relevant to the role you are applying to if you haven't been doing this already; this could elevate your open source work to job experience in its emphasis. Assume 99% of people will only see your resume, everything else should be supporting resources in case they get interested enough to look (or wish to validate your claims).

EternalData 10y ago

+1 to Python Machine Learning, I feel like some of artificial intelligence is understanding a lot of finer (but critical) mathematical nuances, but some of it is just getting your hands dirty, and the second is definitely more accessible for somebody who is starting to learn.