If you’re one of the people who think AI research is infatuated with the terminology and all the things AI hasn’t done yet, you’re not alone. The other hobby, funded into the billions of dollars bracket, is doomsaying.
So far, the story is this – A group of researchers has mystically come up with the verdict that a super-intelligent AI, which (doesn’t yet exist), can’t be controlled. Actually, this verdict is a bit more rational than that. The trouble with the statement is that they’re trying to predict something which doesn’t have any operating systems, operational parameters, or inputs.
The research information, such as it is, refers to a “Turing machine”, a device named after one of the parents of modern computing, Alan Turing. This is a computer that calculates the limits of what can be computed, another helpful definition. The mere idea of the Turing machine is much more advanced than modern thinking, and way beyond actual computing capacity at the moment. So that’s not really much help, either.
AI and the fear of AI
As a PR exercise for AI, it’s more than a bit grim. It’s very off-putting about tech which is quite inevitable. People are scared enough of the unknown elements of technologies that do exist. AI is at the top of the moving target list of unknowns.
On the more cynical side of the ledger, creating baseline problems that don’t yet exist for technologies that don’t yet exist is also a bit off. What use is that?
If like many people you spent years being told “the Bogeyman will get you” and haven’t yet noticed that he didn’t, you’ve now got something else to worry about. So that’s a plus.
Practical issues
The main problem identified by the researchers is that control will be impossible. None of the “rules”, even Asimov’s Laws of Robotics, will work. That may have something to do with the fact that the whole idea is hypothetical, or simply reflect the fact that researchers get given some pretty weird things to research. …Whether they’re researchable or not.
The argument advanced by the researchers is that “total containment is, in principle, impossible, due to fundamental limits inherent to computing itself.” Isn’t exactly definitive. Most computing operations are predictable sequences. 1 + 1 may occasionally equal 2. 0 and 1 are pretty readable down to the basic switch level.
….So what’s uncontrollable? The fact is that we need AI to process the gigantic amounts of data produced by the sciences and daily life. These AIs are all function-specific. It’s therefore unlikely that a super-AI doing water management will suddenly decide to do a B movie script in which delightfully photogenic raccoons replace the human race, for example. Pity, really.
Could a super AI of the future take over all the other computers? Well, your current phone could probably do that for previous computer technologies. The missing point in this argument is that a super AI might have better things to do than take over comparatively bovine old 2020s computers.
Humanity and AI
There’s quite a lot of disingenuity, probably unintentional, in this argument. AI will become humanity’s dream – A technology to blame for everything. There it is, folks, prim and pristine, a technology nobody will admit to understanding which can be the built-in excuse for anything.
Further disingenuity – People talk a lot about machine learning and neural networks. What are these things based on? Human learning and neural networks. Future AI is likely to become a self-portrait of humanity with the added advantage that you can turn it off.
The fear is based on human development of technologies based on human ideas. …So the risk of super intelligence is minimal. Humanity can return to its drear and drab comas secure in the knowledge someone will screw it up anyway.
Thus frolic thee, O Mighty Schmucks of the Wastelands, mid the futile, filthy feckless realities of politics, greed, and total disorganized inefficiencies wondrous and glib. You’ll love it.
Meanwhile, do try to find out what you think you’re doing with AI. It might help.
_______________________________________________________
Disclaimer
The opinions expressed in this Op-Ed are those of the author. They do not purport to reflect the opinions or views of the Digital Journal or its members.
