This reflects my biggest concerns at the moment. AI is far from new. Way past infancy. We've hit a degree of it's usse that has caused it to start to be in the public eye, but it wasn't a breakthrough. It's just another iteration of the self learning algorithms. One famous, well known one of those is the YouTube algorithm, that helps you (and each of us, individually, with whatever "preferences" it thinks we have), find our way through that site. That site has enough videos, that if you scrolled your scroll wheel as fast as you could, too fast to even read, you'd never get to the end of the list, because the list of titles grows faster than you can get it to move up the screen. YouTube and Google have no control over that algorithm, except to install a new limit (like "don't show how to eat tide pods"), or to shut it off completely, but It's hard to put it into terms like that, because unlike an anthropomorphized life cycle, AI has no limits to life span or brain capacity. It is self learning, self growing, it writes it's own code. If it runs out of "brain power", it just needs more processors plugged in, more memory plugged in. The law makers (and law enforcers) don't have a lot to work with here, because they do not know what it is. Neither do the people who turned it loose. It literally learns on it's own. It literally writes it's own code. After some time in the wild (it's far from new right now), all the best, brightest, most accomplished computer scientists in the world don't know much more about their program than the 20 year old whiz kid who read about it in a magazine. The only restraints on what AI can do is what is set in it as a hard limit when it's set loose on a project. And that is and can only be based on what can be foreseen at the time.
Right now (with exceptions I'm sure) AI is being used for "finite" tasks, where it's "missioin" is to do a specific thing. What becomes scary is when someone turns one loose that does NOT have a specific goal, and/or it's goal is not one that's desirable. Or it's not precluded from accessing it's own "OFF" switch, which in the end is just more software in the program. Or when it's not restrained from updating it's own mission parameters. The stuff sci-fi movies are made from. We ARE there, and for now, we're relying on programmers and end users of this software to keep there from being any "holes" in the safety part of the programs.
We are not (to my knowledge) having any messes like that right now, but it will come. The more this stuff is available, the more bad actors will use it. It's been five or six years now since I got my first AI phone call. A very poor one. We think we don't get those any more, but what's actually happened is the AI has "learned" how to interact in such a way that it's not recognizable from an actual person. So when you get a "spam" phone call, it might yet be a person, and it might not be. You really can't tell any more. That's why AI calls have "gone away", it's not that they're gone, it's that they're unrecognizable. Due to the countries of origin that are popular for these types of fraud, law enforcement is pretty much out of the loop. Just as it is when it's a human calling to talk you out of your money. I see no reason why that sort of thing has any reason to stop with technology growing to larger and larger potential.
Yes, it's just a program running on microchips. What you're missing is the scale at which it's used by the internet giants. This doesn't run on a desktop comouter. Billions of instances of the same program running at the same time. Literally, billions, with a "B". More efficiently than that many individual laptops, but that power is real, and does add up.
That's not even a question. It's an evolution that's been growing geometrically since the 1950's, (60's? not positive), when IBM gave it the name "machine learning". Because of it's gometrical groth pattern (as is typical of computer software), it's advancements that left it's predecessors in the dust, particularly it's having foresight, predictive power, and the power to GENERATE it's own code, rather than "react" with the changing of some available variables, it's been "upgraded" to artificial intelligence, and it's kind of become more well known. It's not new at all. It's capabilities, it's implementations, and it's public awareness are what's new.
Dancing robots are not AI. I mean, AI could be in there, but dancing robots that have learned to balance themselves, react to unexpected outside stimulus (someone trying to knock them over), That's reactions. That's the realm of machine learning. The transition from machine learning to AI is not a "hard line", although there are some distinctive differences, but the advancement from "self learning" to "virtually thinking" is a change that's 30 plus years old. Of course the "visionary" computer folks from the earliest days had "intelligence" in mind, but it took quite some time to get to where it was even on the table to try....
What has changed with AI as of late, with it being more well known and publically visible, is the power of these programs, the scale at which they're being used, and the availability of "off the shelf" programming, to set up your own intelligent computer, sent off to do whatever you want it to do...
