The Unexpected Places Where Artificial Intelligence Will Emerge

These places don't strike me as very "surprising." If we wanted to talk "surprising," we should be looking at the more offbeat, but nevertheless plausible sources:

1) Logistics & transportation. Inventory/warehouse management, traffic management systems, shipping, autonomous cars and even roads, etc.

2) Intelligence. Both industrial and government-oriented. Heavy investment is being made in semantic/parsing algorithms, pattern matching, and analysis.

3) Medical research. New drugs are already being invented by combinatory programs. But there are still a lot of inefficiencies in the system, because it's basically the throw-spaghetti-at-the-wall-and-see-what-sticks approach. There are billions of dollars to be made by the company who can develop a smarter combinatory program, and no doubt billions being invested in the effort to do so. Also: robotic surgery and medicine -- though that's more robotics than AI, per se.

4) Weather modeling and prediction. As the global climate and weather become less predictable, the challenge of prediction grows increasingly computationally complex (and more economically valuable). I would expect huge investment in this area in the future, if it's not happening already.

At the end of the day, AI seems most likely to emerge in the places where a) the difficulty of the problem at hand far surpasses humans' best efforts; b) the need for speed or scale cannot be accommodated by human effort; c) the economic upside is enormous (ergo, incentives for R&D expenditures); d) the nature of the problem is best solved by a "superhuman" intelligence model, i.e., a model of computation resembling the way the human brain thinks about problems, but more robust -- something requiring creativity, problem solving, and working memory at enormous scale.

> working on creating military robots that are as lifelike as possible

Ars wrote a while back about the linux-powered rifle,[1] it seems clear we still have some healthy reluctance for letting AI decide when to fire guns. So what could you use a battlefield infantryman for if no one wants to go down the road of arming AI?

Sure, the dog is one thing, to carry supplies or retrieve someone for medical assistance. Another though, might be behavioral camouflage.

A sniper has a group pinned down. What if you could have a bunch of humanoid AI trained to make furtive movements, pop a head out briefly, dart between cover, to make it more difficult for the sniper to tell when you were moving? What if self-driving humvees were going up and down every road with tinted windows, so that roadside bombers never know which convoy is worth blowing up? What if you have to escort an MVP to safety, but you have the option to do so in a crowd of other apparent people?

Wait, maybe that's the more interesting question. This humanoid AI is great and all, but how long until we have an AI crowd?

[1] http://arstechnica.com/gadgets/2013/03/bullseye-from-1000-ya...

Also known as the places where everybody who thought five minutes about the topic would expect AI to come from.

Why autonomous war machines shouldn't be allowed:

http://www.youtube.com/watch?v=pMYYx_im5QI

Not unexpected, nor futuristic. AI has already emerged in these fields. Only through the AI Effect can the author deny this.

That headline is a contradiction.