okay....Play Fifa 14 on the playstation 4. The players independently react to situation presented to them. That's artificial intelligence.
Yes, we have figured out AI. Yes, we did it by way of algorithms. In the same vein as how our intelligence comes by way of electrical pulses sending messages to different sections of the brain.
No. We have figured out
some aspects of AI, and we have figured out how to program some behaviours that
look intelligent and
look human.
What is Siri, Google Now, Cortana? These are all keyword-based assistants. Their "understanding" of what you are saying is entirely based on looking for keywords that trigger actions, then looking for other keywords nearby in context to satisfy the required inputs for those triggers. They try to pull out the data they need to complete their task. They ignore everything not related to the keywords detected, including "fluff" words like "the", "please", etc.
Wrap that up in a magical layer of speech recognition (so you say the sentence instead of typing it) and text-to-speech (so you hear the response instead of reading it), and you get a device that seems intelligent. And it kind of is. But not really. Play with it long enough and the cracks are soon apparent.
Try it yourself. Say "Siri, would you please be a dear and set my alarm for 7:30 tomorrow morning? Thanks so much." It worked, right? She understood a natural English sentence! Intelligent!
But you could also say "set alarm 7:30". Or something gibberish like "Mind the set and alarm the pickle 7:30" -- Siri isn't trying to comprehend anything there, it just plucks out the keywords it needs (a time of some sort) to associate with the keywords "set" and "alarm". Or say "Set my alarm for 7:30 and then tell me if I have any appointments tomorrow" - the second half of your instructions are ignored because they are not relevant to the first part. Why? Because Apple hasn't programmed Siri to handle compound statements. (In fact when I tried it, Siri pulled the old ELIZA trick of taking my whole compound phrase and trying to stuff it into the command it was trying to execute -- by calling my 7:30 alarm "my alarm for 7:30 and then tell me if I have any appointments tomorrow".
The more actions they program into the assistant (the more it can do), the smarter it looks. But that's all it is. Keyword-based fuzzy logic inference engines, API services, rules engines for matching input parameters. Then they throw in some fun, useless, "if user says X then respond with witty remark Y" rules, to make it pretend to have a personality. It's fun, but let's not think we're talking to any sort of intelligence here.
It's very slick. And pretty advanced. I don't mean to minimize the work they've done on these assistants. But AI in general? No, we have barely, barely scratched the surface.
(signed, notjustjay, M.A.Sc. thesis topic on machine learning and behaviour imitation)