People often ask me what I think of AI. I find myself repeating this without much effect because it's not what they want to hear:
- We're not anywhere near the birth of an artificial intelligence.
- We're producing convincing illusions for the sake of distraction and feeble hope. Desperate credulity plays a large part.
- Currently dominant approaches to artificial intelligence contain no real scientific breakthrough. Most of the ideas are at least 50 years old.
- All recent progress in AI has been powered by improvements to hardware, not the algorithms.
- All innovations have been engineering challenges to overcome the relative stupidity of the approach.
- This approach is essentially brute force. We increase the dimensions to the limit of what the hardware can manage.
- I believe that current approaches running on von Neumann architecture are already plateauing: eventually there will be no more hardware headroom unless quantum computing becomes practical.
- There is no one single insight or scientific theory behind it all. What you see is much like a modern automobile: an old idea wrapped in innumerable luxurious extras, designed to smooth out the ride and disguise the lack of originality.
It would be the height of arrogance and foolishness to assume that we are now using the ultimate technology for computation, namely silicon based integrated circuits.
Brooks, Intelligence Without Reason, ยง3
Consider the space race. What got us to the moon? Was it science, or was it engineering? Getting three dudes to chum around on our natural satellite involved an unthinkable stack of engineering challenges, but the fundamental science was Newtonian physics and good chemistry. Essentially brute force: the Saturn V rocket was a 3,000,000 kg fuel tank carrying 40,000 kg of payload - an efficiency probably better than what our chatbots accomplish, since a moonwalk is arguably an order of magnitude more impressive than churning out a string of characters to be displayed on a 2D grid of pixels. This is a highly impoverished domain of intelligence: just because we exchange such strings all day long, have learned to hallucinate a human presence behind every such string, and for many of us this exchange of bytes represents our livelihood, we've come to believe that this task must represent one of the fundamental challenges of intelligence - "if writing emails and sloppy javascript all day long constitutes my essential value, a computer that can do it autonomously and convincingly must be pretty smart".
When they create something as wildly successful, agile, and tenacious as a housefly, I'll be impressed. When we're able to create something which demonstrates navigation and success in the real world of space and time, that will qualify as genuine intelligence - until then, it's merely impoverished simulation and childish digital games.
Having said all that, here are the caveats:
Truly neuromorphic hardware is on the way. When genetics and silicon meet, when we're growing engineered tissue to suit our needs, and when we're able to write software for that tissue-hybrid, things may progress rapidly. Genetics is the most dangerous and potent 21st century science. No matter how many GPUs they stack, no matter how many vector spaces they intertwine, machine learning will never achieve anything like the awesome power of relatively simple genetic editing: the recent flatulence of virology may only be the beginning.
We're determined. Not only because of the profit involved, not only because of its entertainment value, but because we need something to worship. We want a god.
We want a god. Something to pray to, something to admire, something to grant us meaning, hope, and purpose. The more our lives are filled with petty sedentary activities - sitting, typing, staring, thinking, worrying, consuming, shopping, collating, organizing, labeling, bickering, scrolling - the more urgent this need. We want the golden calf: we beg our brightest children to forge it for us. We lavish praise and reward on those who create the anticipatory illusions. We want a priesthood of scientist-engineer to give us this god: an illusion so overwhelming, so universally admired that we can all agree to bow down and feel the presence of the sacred.
The Egyptians understood how to create things so incredible that no one could doubt their divinity: a perfect gleaming geometry rising from the chaos of the sand. A three thousand year illusion, a special effect built on unthinkable misery and sacrifice, an insurance policy against revolt.
Notice however, that as antiquity dragged on, the big gods no longer sufficed. Personal gods proliferated, becoming indispensable by late antiquity: the birth of the individual necessitates the personal divinity. It's the psychosomatic profile of this origin we should be paying attention to: anxiety, alienation, an ever shrinking horizon of concern - the fretting worried human creature who needs to be told, at any given moment, that it's going to be okay. The need to quiet the voices in our head: this is the origin of modern morality, modern consciousness, and ultimately modern religion.
In 1900, two nobodies from Ohio were able to push the boundaries of aeronautics by building their own airplane. 50 years after the Wright brothers, and it required thousands of engineers to design and build the jet fighter, but the fundamental theory hadn't changed. And since then, aeronautics has hardly budged: commercial flights use a design essentially unchanged since the 1970s.
But the comparison with the space race is even more apt: as soon as we achieved it, we quit. Because there's nothing there. There's nothing on the moon that isn't represented a million times over here at home. Earth is the most interesting planet. Just as the immense well of life here is a million times more interesting and intelligent than everything "artificial".
The problem is not that there's nothing to learn from this squirming dirty bowl, it's that we're unequal to it: we lack not only the humility but the imagination - and perhaps the intelligence to study life.
We've got something called the History Channel whose primary message is "everything worthy in humanity is extraterrestrial". Obviously this is what sells, but consider what it means that this is the kind of history continuously fed to the lower classes, that this and other conspiracy theories forms their sacrament, in those gaps where only a generation ago the church would have been. I know because I come from this class: they need gods. Humanity hungers for mythology, divinity, theogony. Some of the hipsters have even become aware of the need, and labor to resuscitate Christ and mold their conscience around some rebaptized theology. There is an increasingly common consensus that the gods must return one way or another, and the wiseacre is positioning himself in anticipation of the priesthoods of the future.
Because of my background I've had the opportunity to talk with some of the players near the leading edge of AI. In their less guarded moments, they'll come close to admitting most of what I've said. Especially those systems engineers, whose job it is to duct-tape together these expensive illusions, have seen too much of the internals to believe in the magic. They won't admit it publicly because it's their livelihood: the illusion sells. Those in leadership positions understand implicitly, that the task at hand is to form a new priesthood: this accounts for the concern for an "ethical AI". The goal is to produce a deployable moral police backed with the authority of science, which will enforce those values supportive of this emergent technocratic priesthood.
But don't get me wrong: there's no grand conspiracy. These are just rich kids and Stanford alums trying to give the crowds what they want - desperately, cloyingly, urgently. These are the artisans of the gods - in another era they would have just been predatory lawyers. The sooner they realize the priestly role, the sooner they'll develop a more compelling aesthetic: they haven't yet learned to capitalize on mystery and authority. When they stop trying to appease investors with yet another kindergarten-HR-morality-clerk, and learn to craft an enigmatic godchild, a more interesting era can begin.
