The real AI Threat
By now, we have all read some version of the story. AI is vastly underestimated, even if it’s pretty dumb now, the innovation is accelerating.
We’re in for some tough times.
Extrapolating free of real world constraints, sounds so much better than speculating.
This long read here is one example.
The dystopian versions take many forms. At best they scare everyone into thinking we won’t have jobs to do in a year. The keyboard warriors have never had to repair a fence or replace an engine’s water pump. They also don’t realize every technological advance in the past came with the same warnings from ‘pundits’ and created more opportunities than existed before. So they think the best way is to implement some Universal Basic Income (UBI) so we can all live off of each other’s production. You’re not supposed to follow this idea too far and realize if robots do everything, whose production will you allegedly be receiving ? But the program itself, will surely be administered by well-intentioned people under the direction of - you guessed it, psychopaths who will want to use it to control you, make you obey and perhaps leverage a nice social credit system the kind the WEF people praise constantly when looking at China’s version.
I’m no fan of UBI, because it ignores most lessons of economics. If you need a short book on this for your friends, this is a good one: Universal Basic Income, for and against by my friend Antony Sammeroff.
You know that after AI has replaced all the heart surgeons, the UBI administrators still won’t trust AI to run the program, because where would the fun be in that ?
But losing our jobs is the mild outcome, which they inevitably serve you right after scaring you with predictions that are much more in line with the movie-plot dystopia we all know. If the Terminator threat doesn’t paralyze your brain from using common sense, by the time you get to the UBI pitch, you’re already exhausted and incapable of asking the dumb questions.
So we’re supposed to think AI will one day decide - a hallucination of course, that it needs all the resources available in the world to engage in its interstellar expansion and will decide to turn us all into hydrocarbon goo. Few have considered the economic calculation that would be required of AI to perform and understand in order to reach this conclusion, no doubt because so many of us are completely ignorant of economics. Alternative uses for humans, the opportunity cost, is not considered and the conclusion is straightforward: to avoid human extinction, we must elect some of us to act responsibly by cracking the skulls of whoever gets in the way of the smart regulations that will allegedly prevent this. You see where I’m going.
You must assume the smart AI is just like you, ignorant of economics, territorial and cold, perhaps even spiteful, jealous and sadistic. The movies or religious books will give you plenty of inspiration for your fears. Absent from its thinking are the concepts of comparative advantage, opportunity cost and the socialist calculation problem and the knowledge problem. If you grasp some of those, perhaps the vision of a Star Trek ‘Borg’ world will still make you react.
Not many of these alarmist have used the fact that AI currently has no concept of time, no decreasing marginal utility, zero incentive to choose and no experience of pain or awareness of death and likely won’t - ever.
These things are all central to what brings economic laws into action. No death implies eternal procrastination. No pain implies never having to choose. No time implies no time preference which cascades into no opportunity cost.
Does that mean AI isn’t dangerous, then ?
Of course not.
It just means AI won’t be a threat from AGI and some singularity, and instead it will only be a threat as a tool used by the wrong hands.
But here we go again.
Who would possibly be capable of ordering an AI to turn a bunch of humans into lobotomized slaves ? You guessed it: a psychopath.
That’s a far off threat of course and down here in the real world, we have more pressing matters to worry about, such as some hacker using AI to impersonate your voice and call your mother in an attempt to extract ransom money.
It’s sociopath behavior of course, but by then hopefully your phone will have been upgraded to have a defensive AI filter for these things.
My point is that AI is neutral in the long run. In the short term, there might be some scary stories about job losses and deep fake hacking disasters. But as a tool used by psychopaths, it’s no different than nuclear bombs or drone armies.
The problem is still the same.