Just mull the power and influence of deep learning: Yann LeCun is in FB, Geoffrey Hinton and part of his lab are affiliated with Google, Ruslan Salakhutdinov is now with Apple. In addition to this, there are other alumni of LeCun, Hinton, and Bengio dispersed across different notable companies like OpenAI, etc. Aside from the post-WW2 semiconductors-spurt, I can't think of a 'technology' that has become so suddenly so important as deep learning. You have a few scientists who spawned the fields of neural networks and deep learning in charge of what appear to be significant research efforts at the top tech companies (by market cap). Like silicon semiconductors and integrated circuits, deep learning approaches are likely to be the primary set of algorithms underlying many future 'intelligent' products and services. You will likely see a similar thing in biology/biotechnology with CRISPR in the near future.
I don't know who any of these people are. I am assuming they are all at the top of their field. Can you recommend some blogs or new sites geared specifically to AI news that you follow. Its a subject I am interested in. Thanks.
If you are really new to AI and want a broad understanding of the topic, I really recommend this book as a quick read: "Artificial Intelligence: What Everyone Needs to Know" by Jerry Kaplan
How about a book for a more technical, practical introduction?
I'm (hopefully) just starting a new position where some of the team is using machine learning techniques in an applied sense (i.e. it's not a CS position, they're not researching new methods, just applying existing ones). I'd like to get a good overview of common mistakes, pitfalls, etc. to look out for from a newcomer's perspective. I have used neural networks naively and briefly in a project about 6 years ago, but I know a lot(!) has changed since then.
A fair amount. I had a lot of math courses in university, quite a bit more than I needed for quantum mechanics and that sort of thing. But I'm also by no means a mathematician, and I'm definitely out of practice.
If your comfortable with calculus and linear algebra, then that is plenty to get started. Bonus points if you know probability.
Plenty of top universities put their ML material online, so I would pick your favorite and check it out. My school teaches roughly based on the Bishop book (I think he keeps a free pdf online). It's dense reading but has information on a huge array of topics. Someone else may be able to suggest other books that are a little more focused.
Really though, I would just pick a school, look at their course website, get their textbook, and work through the posted material at whatever pace you feel provides the value you are looking for.
Hinton, Bengio and LeCun are the most well known deep learning researchers. They wrote an introductory paper about deep learning for Nature and they also gave a talk at NIPS last year. That would be a good place to start I think.
There are not many blogs out there about the topic as far as I know, but I can recommend Chris Olah's blog.
What is really striking to me is that I encountered and read Hinton's earlier work in 1994 after coming across it in the stacks at the University. I was doing a lot of quantitative social science research in my grad program.
Once I left the university and started doing market and customer research in business that kind of raw prediction/black box modeling didn't fly very well. Companies preferred simpler models with explainable connections between independent variables that they could leverage and dependent variables (business outcomes).
