We've all seen the future in science fiction. In the future, robots that move, look, act and think like humans will coexist with humanity, doing jobs and even striking up friendships and relationships with people.
In fact, the concept of the artificial person, the "mechanical man," has existed for well over a century. The very first robot in literature was named "Tik-Tok" (no, really!), which appeared in a 1907 sequel to the Wizard of Oz called "Ozma of Oz."
From the "Machine-Person" in the 1927 film "Metropolis," to the robot from "Lost in Space," to the science fiction movies in the last 50 years — Blade Runner, Ex-Machina, A.I., West World, Bicentennial Man, the Terminator series, Austin Powers, Virtuosity, True Lies and, of course Star Trek, the trope of the humanoid robot is a staple of futurist fiction.
But as we get closer to that future, it looks like it's never going to happen.
(Well, never is a long time — let's say it won't happen in the lifetimes of any living person.)
The reason we assume fake humans whenever we think about the future of robots is related to our spotty record in predicting the future generally.
Watch this 1967 prediction about how people would get the news in the 21st Century, explained by Walter Chronkite. In the video, Chronkite points that in the future, "a man" would be able to use a home office to "carry out normal business activities without ever going to an office away from home." In that office is a "console" that provides "a summary of news from all over the world relayed by satellite." Chronkite says that "to get a copy for permanent reference, I just turn this button."
'Nailed it!' you might say, at least in the essentials. It turns out that in the future we are now living in, people do get the news in their home offices. Satellites could be involved in the transmission of that news. You can even print out a copy for "permanent reference."
In fact this video reveals the whole problem with predicting the future.
They merely applied expected future technologies (computers, screens, satellite data available to consumers) to the culture that existed before those technologies were mainstreamed.
What they missed is that technology changes culture and culture changes technology. New tools change cultural norms and expectations, which drive new directions in technology, which further changes culture. It's a feedback loop that can't be easily predicted.
In the middle of the last century, futurists predicted food in pill form (we can do that, but people prefer eating food). They predicted furniture made out of plastic that could be cleaned with a hose (we can do that, but people prefer to sit on more natural materials). And they predicted flying cars (we can do that, too, but people prefer to not die in a horrible flying car crash due to the vagaries of weather, airspace mishaps, flying into powerlines and running out of fuel mid-flight).
We can often predict technology. But it's harder to predict culture.
The trouble with humanoid robots
The humanoid robots of today tell us what's wrong with the robots of tomorrow.
Consider the latest robot to enter the workforce: Hyundai last month rolled out an AI-powered customer-service robot for car dealerships. Called DAL-e, the 4-foot tall robot is being deployed in South Korean car dealerships now and in international dealerships later. DAL-e boasts face recognition for remembering customers. And its function is to answer questions.
In other words, it’s a kiosk on wheels. Yes, it can “navigate” (slowly and haltingly around on a perfectly smooth surface). Then, once identifying a customer, basically perform the same function as any automated customer service phone center. Yes, they molded plastic to give it “arms” that gesture. And its “face” is pictures on screens.
Not exactly “West World.”
In fact, novelty applications and customer service are where most of the humanoid robots are going.
For many years, companies have worked hard to create robots with facial expression, blinking eyes and moving mouths when they talk. The more sophisticated they are, the more they enter the dreaded “uncanny valley.” It takes so much engineering, compute power and mechanical ingenuity just to create a head, moving face and vaguely believable interaction that most of these robots end up answering customer questions in hotels and other venues. They’re just heads on fake upper-body torsos.
The technology is so off-putting that when the creators of the HBO series “Silicon Valley” wanted a creepy robot torso named “Fiona” for the show, they couldn’t find an actual existing robot that could fill the role. Too creepy and unrelatable.
Instead, they filmed an actress and used the California studio Barnstorm VFX to make the human look, act and talk like a robot. They used extremely sophisticated effects to make Fiona’s eye blinking, mouth movements and head turns look believably robotic.
The closest thing to a real-life “Fiona” is a robot called Sophia. First shown off in 2016. The company plans to “mass produce” robots this year — four models, including Sophia.
As you can see from these videos, the fake clunky awkward robot in “Silicon Valley” is light years ahead of the real fake clunky awkward robot from Hansen Robotics.
There's a conundrum at the heart of the "uncanny valley": The better humanoid robots get, the worse they get.
Robots made human-like to put humans at ease in fact put humans on edge. Roboticists and futurists assume that when robots become really, really lifelike the "uncanny valley" will no longer apply. But I think that's a dubious assumption based on wishful thinking.
The most sophisticated computer in the known universe is the human brain, and the majority of that compute power is devoted to the perception of human qualities in other humans.
But wait, you say: What about those amazing Boston Dynamics robots that can walk and run and even dance?
Yes, the Boston Dynamics robots are amazing. They’re built with the physical mechanics of people or animals for two reasons: 1) for compatibility with the spaces and objects built for humans (for example, a human-sized and -shaped robot could sit in the passenger seat of a car and put on the seatbelt); and 2) because nature is better at designing things than humans are (if the military needs a robot to climb up a rocky hill, it's a great idea to reverse-engineer how goats and donkeys do that).
Boston Dynamics has proven than human-sized robots with human-like locomotion and grasping ability are not only possible, but inevitable. And they'll just keep getting better.
And finally, we have also caught glimpses of the possibilities and limitations of AI voice agents. Amazon Echo, Google Home and Apple HomePod aren't exactly humanlike in intelligence. But they're functional for answering basic questions, setting timers and controlling lights. It's a safe bet that consumer and business voice assistants will keep getting better and more human-like over time until they can nearly function like a human assistant.
So there we have it. We are marching confidently toward a future where technology can give us the faces, bodies and minds of a human (more or less) with which to build the population of humanoid robots predicted by science fiction for centuries.
There's just one problem.
What the future of robots will really be like
The futurists who predicted Walter Chronkite's home office of the future could predict some basic technology, but not the culture or the technology that would be demanded by the culture.
The news of the "future" — our present — really is nothing like Walter Chronkite's "console" full of summaries. We see news on our watches and phones. We see news at the airport. We see news on the little screens installed on gas pumps. We listen to news podcasts. We've got Twitter, Substack, Medium, Facebook News Feed algorithms and state sponsored disinformation campaigns. Nobody, and I mean nobody, predicted the true nature of today's news landscape.
They predicted one form of news delivery, and we got 10,000 forms of news delivery — but not the one predicted.
Robots will be just like that. Robots of the future will come in 10,000 forms — but not the ubiquitous humanoid robots predicted (except for novelty applications).
What is a robot, anyway? A robot is any computer-controlled moving mechanical device. Instead of getting that one fake-human robot, just about everything that moves will evolve into a robot.
Self-driving cars are robots. AI-controlled drones are robots. Rockets will be automated robots. These will all become commonplace.
Farms will be farmed with robots that will hover over crops. Manufacturing will eventually be done almost exclusively by robots. Deliveries will be made by robots. None of these robots will look anything like humans. Why would they?
And we can predict the unpredictable. We'll have robots in shapes and sizes we cannot imagine for purposes and applications we cannot foresee.
As I mentioned, the "parts" that compose humanoid robots will all exist — faces, bodies and minds. But humanity will get these a la carte, for the most part. The talking faces will exist for entertainment. The bodies won't have faces. And the minds won't have faces or bodies, for the most part.
We'll interact by voice AI assistants by simply talking and the responses will be dripped into our ears by tiny earpieces, electronics embedded in our AR glasses, bone conduction devices or implants.
Robots are already adding billions or trillions of dollars in value each year in a wide range of industries, from manufacturing to healthcare. They’ll be used increasingly for telepresence, science and a hundred other applications.
And we should all stand in awe of the achievements in robots over the past few decades. It’s amazing what robots can do.
But forget about the humanoid robots of science fiction. They'll play only a minor role in the glorious, wonderful and totally unpredictable robotic future.