The energy barrier
The main hurdle [to building an autonomous, mobile robot] is actually not its intelligence but its energy, its power.
The human body is an incredibly highly evolved machine that runs on very little energy. Our own brains operate on less energy than a light-bulb. There’s no computer in the world that can come close to performing like our brains do with so little energy. And our own bodies are basically about a quarter horsepower. It’s a very very small engine that runs our bodies that allows us to run all day. And there’s no machine that we’ve made that can do things like walk stairs or walk around that is not many many many horsepower and requires a huge amount of energy. So we simply don’t know how to do a truly autonomous mobile robot that works all day long on very little energy.
On specialised robots
We’re going to find some job – maybe it’s flipping hamburgers – and there’ll be a robot that will make hamburgers better, quicker than a human, but it won’t be able to cook spaghetti. It’s going to only do one thing. Maybe someday we’ll have a robot that will be able to do lots of different kinds of cooking, but it won’t be able to have a conversation, it won’t be able to play soccer.
So we’re going to make these very very specific species of robots, more and more of them, that have varying degrees of automation versus autonomy.
On robot maintenance
… the really difficult thing is keeping them running all the time. Very few robots made today run more than a few hours at a time – maybe eight hours at the most. And then they have to “sleep”, they have to be taken down, they have to be repaired. The human body, the human animal body is a phenomenal engine and machine that’s able to repair itself, to go all day long, keep cool, find fuel and re-energise itself, something that our machines are just incapable of doing at all right now.
So I just want to nix the dream or the fantasy of mobile robots walking around serving us as something far, far, far in the future.
On robots like us
One of the things we should expect though, is that these machines will as much as possible be engineered and be created to be as human as we can make them, even though they aren’t…
They’re going to be made as much as possible to look like humans, not because they need to in terms of their activities, in terms of their jobs, but because we need them to… we want to be able to look at a machine – a robot – and kind of tell which way it’s going to go, so it needs to give us signals somehow, earlier, just indicating where it’s going to move to. So sometimes maybe there’s a little bit like ears, or eyes, that may just drift a little bit to one side, so even though the robot doesn’t really need eyeballs, eyeballs are actually signals that tell us where it’s going to go. So when you’re seeing someone you’re looking at their eyes and their eyes are telling you where they’re going.
In conclusion
There is some fantasy that we can make some kind of AI robot that is superior to humans in all dimensions. That’s just technically, engineeringly impossible. We cannot make a machine that excels humans in all dimensions. You can make a machine that can run faster than the human, you can make a machine that can jump higher, you can make a machine that could crawl lower, but you can’t make a machine that does all those things at the same time, because there’s an engineering trade off.
…
So the point is that there’s no reason to try to make a machine that’s totally like us – flexible, strong, fast, long duration, low-powered – because we can make more of us very easily. Most of the things we will make, most of the robots will be different from us in many many ways. They will be better than us in certain narrow ways, but they’re going to be different from us overall, and that is their benefit… that’s why we’re making them.
Kevin Kelly