When will AI robots become part of our daily lives?

Many AI robots look smart - as long as they work in controlled conditions. Neuroscientist and roboticist Elisa Donati explains why real-world-ready machines need more than fast software.

Why robots are clumsy in real life

Robots already perform a huge number of tasks: they work in factories, deliver packages and clean apartments. But many systems become surprisingly clumsy when faced with unpredictable reality.

Therefore, the development of AI robots requires more than just a narrow focus on artificial intelligence - it needs to consider the interaction of the brain, body and environment. Intelligence, as it turns out, is not just about software.

A smart body makes it easier for the brain to function

When a robot has a smart body, its “brain” can relax. A good example is the animal world. Many animals are much less intelligent than humans and have only a primitive nervous system. But they are able to navigate, avoid obstacles and interact with the environment. Some of this behavior arises directly from the interaction of the body and the environment.

Applying this principle to robotics, you realize that the physical design of the robot is no less important than the AI ​​that controls it. Soft robots illustrate this idea especially clearly - their pliable bodies greatly simplify some tasks.

For example, if a robot's hand needs to move a fragile porcini mushroom, a rigid structure will not work: it will require extreme precision. A soft robot, on the other hand, gives in gently upon contact, making the task much easier. Its pliable structure absorbs some of the effort, so the AI ​​brain needs much less control over the process.

A New Approach to Robot Evaluation

Donati leads a research group at the Institute of Neuroinformatics at ETH Zurich and the University of Zurich. The neuroscientist and roboticist is studying how neuromorphic AI systems—computers inspired by the architecture of the brain—can be integrated with robotics to make the robot's interactions with its environment faster, more adaptive, and more energy efficient.

If intelligence is not just a property of software, but something born from the interaction of the body, brain and environment, then evaluating robots only by speed or accuracy is no longer enough.

Everyday life is chaotic and full of surprises. To really help in such an environment, AI robots need a certain amount of flexibility in responding to changes - such as obstacles, different types of surfaces or unfamiliar tasks.

Adaptability is more important than perfection

This is why it is necessary to measure how quickly robots adapt to new conditions, how stable they are in the real world, and how efficiently they use energy. Equally important is how they interact with their environment:

do they cause damage?

can they work safely around people?

This means we need adaptability, not perfection. The future belongs not to the smartest systems, but to the most flexible ones. True, the robot's ability to react flexibly is limited by its processing power. You can’t squeeze a supercomputer into a multi-purpose household robot.

Lessons from nature for robotics

Donati proposes making robots suitable for everyday use by combining small, specialized control networks with a smart body, according to a study published in Nature Machine Intelligence. Look at nature: there are no central systems that control every movement and reaction. Reflexes and motor sequences often arise from local neural circuits that perform specific tasks with remarkable efficiency.

Some aspects of intelligence arise directly from the interaction of the body and the environment. This is a huge potential for robotics. AI-powered robots have not yet become ideal generalists - and perhaps they don’t need to. What is more important is how they cope with our chaotic world and how we learn to live side by side with them.