But their "AI" does not exist. Their "AI" does not have properties. Their "AI" is not on the way.
Their "AI" is a bunch of fancies bounded by stipulations. Their "AI" stands in the loosest relation to the substance of real code and real networks and their real problems and real people doing real work on them here and now.
"AI" is a discourse, and it serves a primarily ideological function: It creates a frame -- populated with typical conceits, mobilizing customary narratives -- through which real problems and complex phenomena are being miscomprehended by technoscientific illiterates, acquiescent consumers, and wish-fulfillment fantasists. Ultimately, the assumptions and aspirations investing this frame have to do with the promotion and advertizing of commodities, software packages, media devices and the resumes of tech-talkers. At their extremity, these assumptions and aspirations mobilize and substantiate the True Belief of techno-transcendentalists given over to symptomatic fears of mortality, vulnerability, contingency, error, lack of control, but it is worth noting that the appeal to these irrational fears and passions merely amplify (in a kind of living reductio ad absurdum) the drives consumer advertizing and venture-capitalist self-promotion always cater to anyway.
Actually-existing biologically-incarnated consciousness, intelligence, and personhood look little like the feedback mechanisms of early cyberneticists and less still like the computational conceits of later neurocomputationalists. Bruce Sterling said nothing but the obvious when he pointed out that the brain is more like a gland than a computer. Living people don't look any more like the Bayesian calculators of alienated robocultic sociopaths than they look like the monomaniacal maximizers of political economy's no less sociopathic homo economicus.
So, of course, "The Forbin Project" and "War Games" and "The Terminator" and "The Lawnmower Man" and "The Matrix" are movies -- everybody knows that! Of course, our computers are not going to reach critical mass and "wake up" one day, any more than our complex and dynamic biosphere will do. Moore's Law is not spontaneously going to spit out a Robot God any more than an accumulating pile of abacuses would -- not least due to Jeron Lanier's corollary to Moore's Law: "As processors become faster and memory becomes cheaper, software becomes correspondingly slower and more bloated, using up all available resources."
Again, everybody knows all that. But can everybody be expected to talk or act like people who know these things? Sometimes, the exposure of the motives and hyperbole and deception of AI ideology will lead its advocates and enthusiasts to concessions but not to the relinquishment of the ideology itself. Even if we do not need to worry about making Hal our pal, even if AI will not assume the guise of a history-shattering super-parental Robot God... what if, they wonder, somebody codes some mindless mechanism that is satanic by accident or in the aggregate, like a vast robo-runaway bulldozer scraping the earth of its biological infestation, a software glitch that releases an ubergoo waveform transforming the solar system into computronium for crunching out pi for all eternity?
The arrant silliness of such concerns is exposed the moment one grasps that security breaches, brittle code, unfriendly interfaces, mindless algorithms resulting in catastrophic (and probably criminal) public decisions are all happening already, right now. There are people working on these problems, right now. The pet figures and formulations, the personifications, moralisms, reductions and triumphalisms of AI discourse introduce nothing illuminating or new into these efforts. If anything, AI discourse encourages its adherents to assess these developments not in terms of their actual costs, risks, and benefits to the diversity of their actual stakeholders, but to misread them as stepping stones along the road to The Future AI, signs and portents in which is glimpsed the imminence of The Future AI, thus distracting from the present reality of problems to the imagined future into which symptomatic fears and fancies are projected.
So, too, sometimes the exposure of the irrational True Belief of adherents of AI-ideology and the crass self-promotion and parochial profit-taking of its prevalent application in consumer advertizing and the pop-tech journalism will lead its advocates and enthusiasts to different concessions. Sure, it turns out that Peter Thiel and Elon Musk are hucksters who pulled insanely lucrative skim-and-scam operations over on technoscientific illiterates and now want to consolidate and justify their positions by promoting themselves as epochal protagonists of history. And, sure, Ray Kurzweil and Eliezer Yudkowsky are guru-wannabes spouting a lot of pseudo-scientific pseudo-philosophical pseudo-theological nonsense while looking for the next flock to fleece. But what if there are real scientists and entrepreneurs and experts somewhere doing real coding and risking real dangers in their corpoate-military labs, quietly lost in their equations, unaware that they are coding the lightning that will convulse the internet corpse into avid Frankensteinian life?
Of course, the very robocultic nonsense disdained in such recognitions has found its way to the respectability and moneybags of Google, DARPA, Oxford, Stanford, MIT. And so, to imagine some deeper institutional strata where the really serious techno-transcendental engines are stoked actually takes us into conspiratorial territory rather quickly. Indeed, this fancy is a mirror image of the very pining one hears from frustrated Robot Cultists who know all too well in their heart of hearts that nobody is out there materializing their daydreams/nightmares for them, and so one hears time and time again the siren call for separatist enclaves, from taking over tropical islands or building offshore pirate utopias on oil rigs to huddling bubbled under the sea or taking a buckytube space elevator to their private L5 torus or high-tailing it out to their nanobotically furnished treasure cave -slash- mad scientist lab in the asteroid belt to do some serious cosmological engineering.
Again, it is utterly wrong-headed to think there are serious technical types working on "AI" -- because there is nothing for them to be working on. Again, "AI" is just a metaphorization and narrative device that enables some folks to organize all sorts of complex technical and political developments into something that feels like sense but is much more about wishes than working. The people solving real problems with code and technique and policy aren't doing "AI" and to read what they are doing through AI discourse is fatally to misread them. It is only a prior investment in the assumptions and aspirations, figures and frames of AI discourse that would lead anybody to think one should relinquish the scrum of real-world problem solving and ascend instead to some abstract ideality the better to formulate a "roadmap" with which to retroactively imbue technoscientific vicissitudes with Manifest Destiny or to treat as "the real problem" the non-problem of crafting humanist Asimovian injunctions to constrain imaginary robots from imaginary conflicts they cause in speculative fictions.
You don't have to worry about things nobody is working on. You shouldn't pin your hopes or your fears on pseudo-philosophical fancies or pseudo-scientific follies. You don't have to ban things that don't and won't exist anyway, at any rate not in the forms techno-transcendentalists are invested in. There are real things to worry about, among them real problems of security, resilience, user-friendliness, interoperability, surveillance. "AI" talk won't help you there. That should tell you right away it works instead to help you lose your way.