Privacy and AI in the coming age of home robots

Have you ever seen the 1999 movie “Bicentennial Man”? It stars Robin Williams as an android who, over its 200-year lifespan, strives to become human. I won’t spoil the plot but instead note that watching this movie 24 years ago gave me hope that I’d one day have a home robot act as a personal butler. Obviously, Hollywood showing something off doesn’t make it true.

However, here we are in 2023 and I’ve just seen the first glimpse of my dream becoming a reality. Prosper Robotics has shared some details on its still-in-development home robot, dubbed “Alfie,” and I already want one. That said, I also have some thoughts — and some concerns.

Who is Proper Robotics and what is Alfie?

Before getting into the details of Alfie the home robot, here’s the skinny on Prosper. And the amount of info, which is based on the company’s sparse website, is indeed skinny.

Proper Robotics is a London-based startup, founded by Shariq Hashme. Hashme previously worked at OpenAI, the group behind all of the AI products you’ve been hearing about for the last several months. That includes ChatGPT, DALL-E, and GPT-4, to name a few. The company’s home robot, Alfie, isn’t expected to hit the market for another two or three years, and pricing is expected to range from £5,000-£10,000 ($6,186-$12,272) plus a possible subscription fee.

Hey, I never said the home robot revolution would be cheap! So with a hefty price tag plus a potential monthly cost, why would I be excited? Take a look at this time-lapse video of “Alfie” the Proper robot doing some household tasks:

Alfie makes the $1,599 Amazon Astro home robot look like a 15-year-old toy by comparison. Alfie appears far more useful and is built with artificial intelligence in mind; Amazon has only recently decided to incorporate modern AI into Astro.

And let’s be real. Astro is never going to do most of the household tasks that Alfie can do because it physically can’t: It doesn’t have arms, much less any hands. And even if Astro had those appendages, it’s not tall enough to reach many objects or surfaces in the home.

Put another way, you wouldn’t ask Astro to make your bed because you know the device doesn’t have the physical capabilities to do so. But Alfie can, as you can see in this YouTube video. (Even better than I can, if you ask my wife, who is acutely particular about how the bed should be made.)

Smarter design and true AI smarts

Designing with mobility in mind and true AI smarts isn’t something we’ve really seen before for consumer home robots. For starters, Alfie can not only move around a home on wheels but its torso can also rise or fall along a central “spine” of sorts

You can see this in the first video above as Alfie’s torso moves downward in order to put table scraps into the garbage can. That’s far more flexibility and usefulness than today’s home robots, which are mainly built from a robotic vacuum design, can demonstrate. Robotic vacuum design makes sense for robots that can sweep, mop, vacuum, or mow a lawn. It makes no sense for a robot with capabilities covering a range of household tasks.

Also smart is how Alfie can be plugged in while going about its work, which helps solve the battery challenge of a large, mobile, electronic device. Alfie can run on battery power but if there’s an outlet nearby, it can plug itself in and go about its tasks. The cord can retract as Alfie moves around, similar to some corded vacuums.

Starting with AI from the very beginning of the product is also intelligent, no pun intended. The home robots of today are mainly single-purpose devices. There’s far less need for AI in them because they have one main task. So basic algorithms to complete these tasks are programmed in, along with some light learning capabilities. My Roomba, for example, can “learn” which room is which if I provide it with that input. I can then tell the Roomba to clean one specific room when it needs cleaning.

Prosper’s Alfie feels light-years ahead in this regard. Indeed, it has to be in order to deal with more complex tasks such as clearing the table after dinner, doing laundry, or make a salad — all tasks that Hashme claims Alfie will be capable of handling. However, with any AI product comes responsibility. And any connected device in the home has to deal with privacy issues. Here’s where my concerns begin, and what helps explain part of the monthly subscription fee.

Privacy, strangers and task learning

Hashme figures that for Alfie to learn a task it will take roughly 10,000 attempts. This makes sense because gobs of data from both failed and successful attempts are required for an AI system to adjust and learn the optimal way to complete a task. Basic tasks could be replicated in a controlled environment over time before Alfie hits the market, but once it does, Alfie won’t be fully capable of doing every possible task that a human can do in a home.

That’s where the fee, and privacy concerns, come into play. Prosper plans to use people to remotely control Alfie to complete tasks the robot hasn’t yet learned to do in people’s homes. Over time, after many repeated telepresence interactions, Alfie will do such tasks on its own. But do I want a robot that someone else is controlling in my home? Do you?

Probably not, for obvious reasons. Prosper reportedly has a privacy solution to blur people’s faces and any text that Alfie can see during these remote operations. But personally, I’d like to see an independent verification of that, much in the same way that I look for verification that VPN providers don’t log my data. I’d also like to have access to the audio and video feeds that Alfie hears and sees. Without those, I’d be leery of letting Alfie run wild in my home.

There’s another way to train Alfie that most people wouldn’t want, but I do. I’d love to see Prosper create a telepresence user interface for owners to train the robot itself. Yes, this would require far more effort from anyone who buys a Prosper robot. But it would deliver the personal part of owning a “personal robot” to help around the home.

Not to belabor a previously made point, but the way I make our bed isn’t the way my wife does, or rather, how she likes it made. So let me train my robot so it best meets my needs. This approach would go a long way toward alleviating the privacy concerns of inviting a stranger into my home through the eyes of Alfie. And if a consumer is comfortable with sharing their learning model data for the greater good of all the Alfies out there, give them an option to add to Prosper’s collective smarts in a secure, private manner.

Clearly, the home robot showed off in one of my favorite movies 24 years ago isn’t here yet. Nor will it be here in the coming months. But in general, I like the approach Prosper is taking. With some important privacy guardrails and the fast maturing of AI models, Alfie is the best chance yet for me to live the future I saw in 1999 and still want today.