Allison Okamura is a Hoover science fellow and the Richard W. Weiland Professor of Engineering at Stanford University, Professor of mechanical engineering, with a courtesy appointment in computer science. She discusses robotics with Hoover’s Technology Policy Accelerator, whose research and insights help government and business leaders better understand emerging technology and its geopolitical implications.

Allison, you’re working on a new generation of robots that can help care for people in their homes as they get older. What factors are driving demand for them?

People have always imagined it would be nice to have robots in their homes. You know, in popular culture, there’s long been the idea of machines like Rosey the robot maid from The Jetsons. However, the idea of a robot that can provide in-home care to people is quite different. Demand for these robots is being driven by various societal trends, including aging populations, falling birth rates, immigration policies, better medical care that boosts longevity, and so on. There are just not enough caregivers – whether professionals or family members – to look after people as they age. A country like Japan, which has an aging population and limited immigration, tends to be more accepting of the idea of robots caring for the elderly than, say, the U.S. But their use is likely to grow here, too.

What kinds of things will these robots do?

To help with aging in place, these robots will assist with self-care activities, like helping someone use the bathroom. I think it’s fair to say that many older people don’t want to have their children or other human carers help them take a bath or use the toilet. Assistive methods that don’t involve another human might be more easily accepted from a psychological perspective.

Could you describe the technological changes occurring in robotics that make it possible to envisage a future with robot carers?

There are two main types of changes taking place. The first is in hardware. Many robots used today are in manufacturing, and they tend to be heavy and stiff so they can move materials exactly the right way and do it very quickly. For robots that help someone get out of bed, you need a softer interface to the human body. There’s a growing field of “soft robotics” that’s helping us understand how to use novel materials and physical structures to make robots that are inherently safe when it comes to interacting with humans, but that still have the strength and power to provide the physical assistance people need. Soft robots use materials like knitted fabric and rubber that reduce the risk of injury to humans.

Beyond the robot body, there are also changes happening in the robot brain. The next generation of artificial intelligence, including the development of foundation models and generative AI, is giving us a glimpse of how software can help machines become more capable of reacting to unexpected scenarios, which is essential in caregiving situations.

Looking ahead, what types of assistive robots will we see in homes?

I think they will go two directions. One is machines that do a specific thing, like a robot that’s just there to help someone get into a chair or off a toilet seat. That robot will be permanently installed in a location and have just one job. By that definition, you might even call a washing machine a robot, but you could also imagine something that sorts and loads clothing. These types of devices could be useful to people within the next five to ten years. We already have robotic vacuum cleaners in homes, but I’m talking about more widespread adoption of different types of machines.

But ultimately, having a different robot in your house for every different task is not going to be practical – it would be very expensive. What people would like is a general-purpose robot. There’s a lot of interest and investment in humanoid robots right now, and they’re attractive because they could potentially do many of the same kinds of tasks humans can do. But they are extremely complex, so my feeling here is we’re talking more like twenty to twenty-five years before they are in homes taking care of people. We’ve seen this with self-driving cars: the technology evolved over a couple of decades and it’s only today that we’re seeing them deployed on roads in some cities.

Will robot carers still require some amount of human oversight?

There’s still a lot of work to be done in both soft robotics and AI to get to where we want to be. We will need more focus on – and investment in – making soft materials cheap and durable, as well as in understanding how to balance the need for softness with the need for strength. That’s a key area of research. And on the intelligence side, this new generation of AI can still make unforeseen mistakes. That may not matter for creating an initial draft of a letter, but the consequences could be very serious if AI is used to direct a robot helping someone out of a bathtub. There are roboticists who are building foundation models specifically for physical actions that can be used to program robots, but it’s still early days and we’ll need to assess what level of errors is going to be acceptable and what safeguards will be required.

Robots can be teleoperated today. For example, a surgical robot can be controlled from a distance by a surgeon. But for robots to assist people at the scale we’re talking about in the future, you can’t have this one-to-one relationship. You could envision scenarios where a human has a more supervisory role over a group of robots. To scale, though, we’re going to ultimately need robots with enough intelligence to operate autonomously.

How are we going to pay for these machines? They’re likely to be expensive, aren’t they?

Well, let’s start by talking about how expensive care is when it’s delivered by humans. A lot of families wind up putting their grandparents or parents into nursing homes because that’s more affordable than hiring an individual carer. So the question is going to be: when is the cost of assistive robots going to become less than the expense of paying for human-delivered care? The cost of human carers is likely to keep increasing while the cost of assistive robots will come down. The crossover point is factored into my earlier prediction that it will take a couple of decades or so before we have general-purpose robots in the home. Hardware-wise, we really need some big steps forward in things like how long batteries can hold charge for and the amount of power a lightweight motor can output, and we’ll need to keep taking advantage of commoditization of components – like highly capable miniaturized cameras, whose costs have fallen dramatically thanks to their integration in smartphones.

Social interactions are really important for humans. What are the implications of using robots rather than people as in-home carers?

We found during the COVID-19 pandemic that people managed to have interactions through video calls, but the lack of in-person contact was socially and emotionally damaging. The question – and I think it’s a very important one – is to what extent can interactions with machines replace social interactions with humans? We’ve seen chatbots become popular in the digital realm. Now if you bring these into the physical domain with a robot in the home, there are a lot of concerns about how this will affect people who are being cared for.

On the flip side, if you want someone to stay in their home as they age, that might also inherently make them lonely. If a robot can empower and enable them, that can be beneficial to their mental health. They might feel a sense of abandonment, but on the other hand, they might feel a sense of independence. I don’t think we yet fully understand what these tradeoffs are going to be.