Every now and then in the discussions of AI/AGI and what not comes the central figure of that entire intellectual movement - Ray Kurzweil. And with him inevitably comes a form of an exponential chart like the one below:
Basically the curve depicts Moore's law (which is not disputable), with a few additional labels suggesting that a particular computing performance is somehow equivalent to a processing power of brains of various animals.
Superficially this looks fine, but of course the problem is hidden in how do we arrive with these equivalences? The typical answer to this question is that perhaps the labels should move around on the curve left or right but the sit there somewhere so it's fine, we might be off by a year or two, who cares. The relevance of whether or not it even makes sense to put brains along with computers on that chart typically isn't even questioned.
Since I want to keep this post short, let's cut straight to the conclusion - this chart alone shows we are off by at least 23 years from original predictions. Why?
Let's take a closer look: Kurzweil claims we should be seeing insect brain capability in $1000 computer in 2001. Alright what would we expect from an insect brain? Well the things an insect should be able to do, with perhaps the addition that we somehow issue a command of what to do.
So what can a bee do? It can autonomously navigate in novel environment for miles, recognize and approach a particular type of a flower, pick up some nectar, navigate back to the hive (without GPS, but perhaps with a magnetometer), and do it many times over in a day.
So we should expect to be able to have e.g. a combat drone that can take a grenade, navigate for miles to enemy territory, pick on an enemy target, drop the grenade, and navigate back to the base.
But we have drones like that in Ukraine no? Yes. Remotely controlled...
This is all you need to know, to understand that Kurzweil is off not by a year or two but by some two decades at least. It's 2024 at the time of writing this, and the best military technology that could benefit from an insect brain like capabilities is still on remote control. I could write a similar story about what we should expect if had robots controlled with the capability of rodents or dogs. But I don't think I have to, you get the point.
So if Kurzweil is right that whatever the brain does could be viewed as computation (which I have some serious doubts about and prefer to call "control") and even giving him all the benefit of the doubt, empirical evidence shows we are at least 2 decades behind. So we shouldn't expect human like intelligence before late 2040's. That is of course if we actually get an insect brain this year. Which we won't.
If you found an error, highlight it and press Shift + Enter or click here to inform us.
