What are some non-obvious applications of AI that will be huge? originally appeared on Quora - the knowledge sharing network where compelling questions are answered by people with unique insights.
I expect we'll see AI start to affect a wide variety of domains, including "obvious" ones that have never really worked before such as robotics. Here are some interesting domains which seem quite close but may not be on people's radar:
- Knowledge work: Today's AI assistants are in their infancy, but an exciting mature version would be something which has context on what you're trying to do, "understands" it, and can do something useful with that understanding. For example, imagine a tool takes a bullet point and expands it into a full-fledged essay. Imagine a virtual pair programmer, which can tell you that you've made an off-by-one error, or can synthesize functions or refactor code for you. These can't be built today, but one could imagine building on modern techniques to make them happen.
- Physical security: Deep learning has shown extremely good performance in computer vision. I don't think it'll be long before we'll see deep-learning powered security systems -- for example, we now have the technology to know the location and identity of each object and human on a security camera. (These applications also raise very important ethical and privacy questions -- these are already being raised today with apps such as FindFace.)
- Dialog systems: Deep learning for natural language applications has been (relatively) lagging, but we've been seeing increased performance recently. I expect that we'll see systems capable of extremely non-trivial dialog in upcoming years. Customer service is an obvious application, but I'm excited to see what good dialog systems can do in education or entertainment.