When the Roomba robot vacuum cleaner exceeded two and a half million units sold in 2008, it seemed like the beginning of a new era: one in which robots would populate our homes, taking care of daily chores. On the other hand, the first robot vacuum cleaner (for the record it was called Trilobite and was produced by Electrolux), made its appearance in 1996. Yet, almost 30 years after its debut and after the prototypes launched by companies such as Tesla, Unitree and the Norwegian 1X, that long-awaited future has not yet arrived. Domestic robots have remained confined to a few repetitive tasks, such as cleaning the floors, cutting the grass or cleaning the swimming pool, while the most ambitious projects – those of all-purpose robots or so-called humanoids – struggle to overcome the threshold of the laboratory. If you are wondering why robots have not yet become a pop technology, know that the answer must be sought in a variety of factors: on the one hand there are technical limits and production costs that are still too high, on the other hand there continue to be prejudices towards these tools and the lack of real usefulness in various domestic tasks.
Where are we after 30 years of robots
As things currently stand, “companion” robots are advanced enough to arouse empathy, but still too primitive to be indispensable in everyday life. Even if those who own a robot vacuum cleaner tend to treat it as a sort of “robotic pet”, even giving these devices a name, we have not yet managed to make these machines objects of common use. One of the reasons behind this failure of robotics is the inability of robots to perform multiple tasks in complex environments, such as our homes and offices. Vacuuming and washing the floor is one thing; folding clothes, loading a dishwasher, cooking or walking the dog are much more complex and more difficult to automate tasks.
And although in recent years companies such as Tesla, Unitree and the Norwegian 1X have shown prototypes capable – at least on paper, in promotional videos – of carrying out a variety of complex tasks, as things currently stand they have not had mass diffusion. Also because very often, behind those demonstrations, there are sometimes hidden systems operated remotely and still far from the absolute autonomy promised by the manufacturers.
A possible solution to the problem could be the acceleration that is taking place in the world of artificial intelligence, particularly in LLMs (Large Language Model), the large language models that allow machines to understand natural language commands and convert them into actions. Google, with the Gemini Robotics project also visible in the videos below, has shown how a robotic arm can receive the instruction “take object
It is a step forward, but the road is long: the foundations of what experts call physical intelligence, that is, the ability to understand weight, strength, friction and fragility of objects, are still missing. To do this, enormous amounts of data are needed, which no one has yet collected. Some researchers propose to “teach” robots by observing online videos or simulating millions of interactions in virtual environments, but we are still in the experimental phase.
Even if these problems are resolved, practical obstacles remain: short-lasting batteries, high maintenance costs and safety issues. A humanoid robot, being a complex and heavy machine, must comply with strict rules: if it turns off while it is balanced, it can fall and be damaged. For this reason, the first realistic applications will probably be limited to industrial or logistics contexts, far from homes.
To be useful, a robot does not necessarily have to be humanoid
At this point a question arises: is it really necessary for a robot to have human features? Legs are energy inefficient, while structures on wheels or tracks provide greater stability. Focusing on robots with humanoid features derives more from a collective imagination strongly influenced by the world of cinema and comics than from a real functional need. A robot that walks and moves its arms is more likely to gain media attention than an anonymous, albeit efficient, model that walks on wheels. Yet, perhaps, precisely in this distance between form and function lies one of the biggest reasons for their lack of success. As long as we continue to chase the fantasy of the “human” robot in the strict sense, we risk forgetting what we really need: simple, reliable and accessible machines, capable of making life more comfortable without necessarily having to imitate humans.
