Technologies & Models

The Robot in Your Home: Why Is Easy for Us Hard for It?

We accepted robots behind the factory fence, so why do we hesitate when one enters the kitchen? The task is harder to engineer and more delicate for privacy than it looks.

4 min read This article is also available in Arabic

A blue background of fine electronic circuit traces

The Nano capsuleOne minute

About 4.66 million industrial robots work in the world's factories (2024), yet a robot that tidies your kitchen remains far harder to build. The reason is Moravec's paradox: abstract reasoning is easier for a machine than the perception and movement a child masters. The cost of running language models has fallen sharply and home robots have begun to go on sale, but some rely on a human operator who sees the house remotely, so privacy is the first question to ask. In caring for the elderly a machine may fill a shortage of hands, but not the presence and devotion of one's children. So ask before you open the door: where do the images go, and who sees them?

Read the full article Collapse the article 4 min read

We accepted the robot as long as it stayed behind the fence. In 1961 the first Unimate entered a production line at a General Motors plant in New Jersey, where it handled hot metal die castings, work that was dangerous for a person. According to the International Federation of Robotics (IFR), 4,664,000 industrial robots were in operation worldwide in 2024, of which 542,000 were installed in that year alone. These machines usually work in a controlled space where every part has been arranged in advance, with no surprises and no fragile glass.

Why is the simple so hard?

The kitchen is the opposite of the factory: varied objects, shifting positions, and human hands passing through at any moment. This is where "Moravec's paradox" helps, formulated by the researcher Hans Moravec in his book Mind Children in 1988. He wrote that it is difficult or impossible to give computers the skills of a one-year-old in perception and mobility, and that what we call deliberate reasoning is the "thinnest veneer" of human thought. According to the same source, Marvin Minsky and Rodney Brooks are credited with similar observations in the 1980s.

So folding a towel or lifting a cup without breaking it may be harder for a machine than a complex mathematical problem. That is less a failure of "intelligence" than a difference in kind: the first is sense and movement in an uncontrolled world, the second is explicit rules.

What has changed?

The cost of the intelligence that directs the machine has changed. Stanford's AI Index 2025 reported that the cost of querying a language model at GPT-3.5 level fell from 20 dollars per million tokens in November 2022 to 0.07 dollars in October 2024, a reduction of more than 280-fold in about two years. That figure concerns language models, not robots themselves, but it is likely to lower the cost of understanding and planning in any machine that relies on them.

On 28 October 2025 the company 1X announced its home robot NEO at 20,000 dollars, or 500 dollars a month. The Wall Street Journal then revealed that the tasks in the demonstration were run by a human operator working remotely in a virtual-reality headset. Decrypt quoted the company's chief executive as saying that the buyer has to be comfortable with this for the product to be useful. The company says it blurs people's faces in the robot's field of view and offers user-defined "no-go zones".

A home is not a factory

We are not judging the company's honesty here, only stating the principle. A home has a sanctity in our customs, and a home robot is a roaming camera that knows where things are and when people come and go. If a person is added who sees through its eyes from outside the walls, the privacy question comes before the question of skill. We explained how secrets are inferred from data in the article "Privacy in the Age of AI".

Care: what a machine can and cannot cover

Ageing societies are often cited. The US International Trade Administration reports Japanese estimates that the need for care workers will reach 2.72 million by 2040, with a projected shortfall of about 570,000. Yet reality is narrower than the noise: the same report says adoption of toileting-support devices is only 0.5%, while monitoring and communication devices reach about 30%. In other words, what is actually used is limited-function assistive equipment, not a "servant" living in the house.

Care in our values is service and presence together. A machine may relieve children of some physical burden, such as lifting, alerting and monitoring, but companionship cannot be transferred to a device. The danger is that a machine masters the appearance of care and we feel relieved of the duty to visit and sit with our parents. This is our own judgement, not a measured result, and the best remedy is to make the machine a help to the children and not a substitute for them.

Questions before you open the door

Ask any home robot the following before buying it:

  • Where are images and sounds processed: on the device or on remote servers?
  • Can an employee or operator see my home? In what circumstances, and with my permission?
  • Can I set off-limits rooms and switch the camera off completely?
  • Who is responsible if something is damaged or data leaks?

If the seller answers clearly, the robot becomes a tool we can hold to account. If the answer is vague, the door had better stay shut until matters are clear.

References

  1. International Federation of Robotics, World Robotics 2025: Global robot demand in factories doubles over 10 years (2025)
  2. Wikipedia, Unimate
  3. Wikipedia, Moravec's paradox (quoting H. Moravec, Mind Children, 1988)
  4. Stanford HAI, AI Index 2025: The State of AI in 10 Charts
  5. Decrypt, Humanoid Hype Meets Reality: 1X's $20K NEO Home Robot Draws Skepticism, Memes (29 October 2025)
  6. U.S. International Trade Administration, Japan Healthcare Caregiving Technologies (2025)