Artificial intelligence: Moravec’s paradox that explains why robots and AI find easy things difficult and difficult things easy

Will we ever be able to create a robot with the same capabilities as humans?

With the explosion of ChatGPT and other artificial intelligence programs, the question is not only becoming more relevant, but it is also fueling the imagination of engineers who want to create a robot that thinks and acts like a human.

As the processes progress, several conclusions emerge: we have succeeded, especially with artificial intelligence, in imitating the complex reasoning and even creativity systems of our brains.

But at the same time, a robot cannot tie shoelaces.

Artificial intelligence and robotics can make reasoned thoughts require less computation, while seemingly simpler human actions, such as tying your shoes or picking up a bag from the floor, require enormous computational effort.

This is known as the Moravec Paradox.

And for many experts, it’s the explanation why it hasn’t been possible to build a fully intelligent robot.

“Humans have taken hundreds of thousands of years of evolution to do things as simple as keeping balance, for example, so replicating all these processes at the computer level is almost impossible right now,” said Gonzalo Zabala, Robotics researcher at the Interamerican Open. University, in dialogue with BBC Mundo.

Zabala points out that the opposite happens with reasoned processes.

“How long have we been able to talk about the intelligent man, about reason? Compared to other evolutionary processes, the time is much shorter, so we can code and replicate this more successfully,” he says.

Hans Moravec and Alan Turing

One of the forerunners of artificial intelligence was undoubtedly the British scientist Alan Turing.

Of the many studies he published during his short but prodigious career, one deals with a series of questions that, in a theoretical case, would serve to distinguish a robot from a person.

That has been the method since its formulation in the 1950s who guided engineers and theorists in the development of artificial intelligence.

As Rodney Brooks, a robotics professor at MIT, pointed out, what happened next was that engineers focused on creating programs or artifacts that could “trick” interlocutors into correctly answering the Turing test questions.

By the late 1970s, that approach began to have a problem: that the logical answers produced nothing original, and that the path indicated by Turing began to run out of many exits.

“Even funding for the research stopped because it wasn’t clear which way to go and there was no progress,” Brooks told the BBC.

Subsequently, new alternatives were sought to advance in the development of artificial intelligence.

“The path that was followed was to create circuits similar to those of the human brain. Not a robot that reacts logically, but a circuit that manages to think,” says Zabala.

It was then that this still unresolved contradiction appeared: artificial intelligence processes were created with some ease, while basic human functions were basically impossible to recreate in a robot.

This was widely observed in the late 1980s by specialists in robotics such as the aforementioned Brooks, the Austrian Hans Moravec and the American Marvin Misnky.

But it was Moravec, a professor at Carnegie Mellon University in the US, who put it best in 1988 based on the work of the three colleagues:

“It is relatively easy to get adult-level computers to perform in intelligence tests or chess, but difficult or impossible to give them the capabilities of a one-year-old when it comes to perception and mobility.”.

In short, robots can be as smart as they are inept.

“What the Moravec paradox did was understand what was being observed. And when the problem is mentioned, the possible solutions to the problem are mentioned,” says Zabala.

“When this point is reached, something really interesting starts, which is to know ourselves better in order to be able to replicate it in robots: knowing how to keep our balance, how to learn to drive, in short,” he adds.

sentient robots

Both Moravec and Brooks and Misnky have advanced projects to clear up the paradox.

Brooks has worked with the American company Boston Dynamics and a company he founded known as iRobots.

The principle they followed is summed up in a direct premise, according to Brooks: “If we want to build a robot with human intelligence, let’s first build a robot with human anatomy.”

From this, robot projects have been developed that present an aspect closer to ours.

For example, a team of European scientists has developed a prototype known as the ECCERobot, which has a thermoplastic skeleton complete with vertebrae, phalanges and ribcage.

The ECCERobot has as many degrees of movement as a human torso, and most importantly, all these parts are packed with sensors.

But the same scientists who developed the robot point out that the drawback has not been overcome: the complexity of the ECCERobot is so great that it can barely hold a cup. Therefore, it cannot be expected to exhibit intelligent behavior.

“Building an intelligent humanoid robot, one that can interact seamlessly with humans and human environments, of course requires advances in computing and battery efficiency, not to mention a quantum leap in sensory equipment,” he said. Rolf to the BBC. Pfeifer, who coordinates the ECCERobot project.

“A really crucial development will be the skin. The skin is extremely important in the development of intelligence because it provides such rich sensory patterns: touch, temperature, pain, all at once.”

But experts point out that despite the difficulties posed by Moravec’s paradox, the possibility of a humanoid intelligent robot, while distant, is not impossible.

“What the Moravec paradox did was reveal a problem for researchers to look for solutions. One of them is undoubtedly what we are seeing with the artificial intelligence revolution, where we have taken a step towards creation, not just logical answers,” explains Zabala.

And it is clear to the connoisseur that this revolution does not pose a threat of extinction.

“I don’t think this is the end, as several analysts have suggested. It is a tool that will facilitate many processes in the future,” he concludes.