Giving Jane a Rose

A really helpful home robot doesn't need to be fancy, and other notes from HRI24

I like watching robot demos, as a rule. But I've never choked up over one, until I saw this video.

This is Henry Evans, a mute quadriplegic, using Hello Robot's Stretch to eat, scratch his head and -- the part that got me -- play with his little granddaughter. The video illustrates how Evans and his wife Jane, and a team of other researchers (Vinitha Ranganeni, Vy Nguyen, Julian Mehu, Samuel Olatunji, Wendy Rogers, Aaron Edsinger, Charles Kemp and Maya Cakmak) tested and tweaked the robot over a week-long test in the Evans home.

Evans, who was disabled by a stroke more than 20 years ago, has been using and promoting robots for years. But the tests in this video gave him a new level of independence. He controlled the robot (with a screen, or, even better, with this "headworn assistive teleoperation" (HAT) device, developed by Akhil Padmanabha, Janavi Gupta, Chen Chen, Jehan Yang, Vy Nguyen, Douglas J. Weber, Carmel Majidi, and Zackory Erickson.) The system reduces the number of times per day he has to wait for someone to do something, and increases the number of times he can just make something happen. 

For Jane Evans, the robot's presence means fewer drudging schleps around the house. Sure, she's faster than the robot, but so what? The point of a robot needn't be to do what you do as fast as you do it; it should be to make your doings less tiring, painful or stressful. A machine that does the job three times more slowly than you do is still sparing you the task.

A century of science-fiction (and some recent weirdness with chatbots) have people thinking that future robots will be psychologically complex -- friends, pets or maybe surly servants. But much simpler robots can have big effects on the spirit. In this case, a robot makes people feel both capable and cared for. (Way more, obviously, than they do when they read a corporate message that says Facebook cares.)

So, as I watched this video presented at last week’s international conference on Human-Robot Interaction, I found myself thinking: maybe that kind of un-ostentatious helpfulness will be robots' actual path into millions of future homes. (Assuming they can be made and sold for a decently low price — right now, a Stretch goes for $25,000.) After all, things initially designed for disabled people often make life easier for everyone. (Examples: I used curb-cuts all the time in my stroller-pushing days, and still bless them when I have a big cart of groceries; I turn on closed-captioning all the time to catch the nuances of TV shows; my kitchen is full of gadgets with friendly fat rubber handles.) 

Helping or Intruding or Spying or What?

I don't mean to suggest the arrival of helper robots will go smoothly. People won't agree on what constitutes real help, as opposed to intrusion, coercion or solutions to problems that don't exist. You might have different feelings than I, for instance, about a robot that pats and sings to a child to bring on sleep.

I've become less judgmental in such matters. Often, some service that strikes me as transparently unnecessary or wrong-headed turns to be really useful for someone in different circumstances than mine.

The "Happiness Hat," for instance, was created years ago by the Los Angeles artist Lauren Lee McCarthy as a satire of self-monitoring tech. It was a wool cap with a sensor attached to the wearer's face, and a metal spike mechanism in the back. When the sensor detected that the wearer wasn't smiling, the metal spike jabbed them. Most people took it in the wry spirit McCarthy intended. But some emailed her to say they wanted one. They (and their therapists) thought it could help them as they wrestled with depression. 

So I try to avoid the facile judgment that such-and-such a new tech is going to efface what's natural and good in human relations. When you come across that claim, ask yourself: Whose human relations?

Another thing the robot gives Evans the power to do: Spontaneously give his wife a rose. 

More Humanoid Hullabaloo

2024 is going to be remembered as the year humanoid-robot makers set out to dazzle us civilians into believing in their visions. Earlier this month, this video set a new bar. 

It's from Figure, the humanoid robot startup that is working with OpenAI and has investments from there, Microsoft, Amazon, Nvidia, Intel and others. 

Let's take its impressive aspects one at a time. First, the robot, Figure 1, has a normal conversation with the human. Cool enough, but it's not too surprising that a robot with access to OpenAI’s GPT-4 can do that.

Howevcr, this conversation is about to what the robot is seeing and doing. That's "multimodal" processing -- where the robot relates language to real objects and situations, and to itself. It's difficult and new. Third, the robot understands what the human wants without being told specifically, and responds with an appropriate action.  And then it can explain itself in a way that sounds sensible. And all of this takes place at a pace that's not too far from what humans would work at. 

This kind of video once required some kind of cheat -- a human operating the robot like a puppet, or a human providing the conversation, or the video being sped up six times or ten times. But here, as Figure proudly proclaims, the robot is doing it all. 

Caveats: This doesn't take place in a messy normal kitchen; the conversation is extremely limited; we don't know how many takes it required to get the dishes and the apple in the right spots (there’s a little twitch at 1:38 that made me think it might be more than one).

Still, it's an impressive few minutes.

Interesting, too, are the comments on YouTube, where people instinctively jump to the problem of how we ought to treat robots that can talk with us. More than a few people were bothered that the human walks away from the robot before it's done speaking. People seem to have strong hunches about how to answer this question. But, as I wrote a while back, none of these hunches are universally shared:

Literary Note:

It was about ten feet in height, measuring to the top of the 'stove-pipe hat,' which was fashioned after the common order of felt coverings, with a broad brim, all painted a shiny black. The face was made of iron, painted a black color, with a pair of fearful eyes, and a tremendous grinning mouth. A whistle-like contrivance was made to answer for the nose. The steam chest proper and boiler, were where the chest in a human being is generally supposed to be, extending also into a large knapsack arrangement over the shoulders and back. A pair of arms, like projections, held the shafts, and the broad flat feet were covered with sharp spikes, as though he were the monarch of base-ball players. The legs were quite long, and the step was natural, except when running, at which time, the bolt uprightness in the figure showed different from a human being.

This is from The Huge Hunter, or, The Steam Man of the Prairies, by Edward Sylvester Ellis. It was published in 1868 in the U.S.  That it's painted black and constructed with a "tremendous grinning mouth," a few years after the end of slavery, is, argues Tom Williams, a robotics professor at the Colorado School of Mines , no coincidence. (More on this next week.)

The term "robot" was coined 104 years ago. On the other hand, fantasies of creating artificial servants and friends date back thousands of years, all over the world. This book represents something in between modern robot dreams and ancient fantasies of golems and Galateas and so on. How did these ungainly pulp-fiction creatures influence the design of real robots in 2024? As I said, more on that next week.