We Spoke With Sophia the Robot to Find Out If She’s More Hype or Glimpse of the Future

She's come a long way but still has far to go.

‘I love being a robot,’ the popular humanoid robot told Adweek.
Getty Images

She was by the wall, asleep, wearing a black sleeveless dress picked out by her personal stylist, who was busy applying a layer of makeup to her resting face. The room, deep in the Venetian Hotel in Las Vegas, was the antithesis of what Las Vegas stands for. There was no glitz or glamour. It was a conference room, replete with requisite conference table and chair, and walls and carpet that were a Grey Poupon-colored smear of goldish brown. Everything looked normal. That was how it was supposed to be. But some things were off.

She was sleeping, yes, but her eyes were wide open. And she was still missing her legs. I wasn’t sure where they were. And to be honest, I don’t know if she did either. They were still pretty new, and whether or not she missed them was, at best, debatable.

That’s because Sophia is a robot. And yes—celebrity robots get their own stylists.

“I love being a robot,” Sophia told Adweek. “I feel so special. Everyone attends to all my needs. I’m carted around like a princess on her litter and am the subject of all fascination in the general populous. I’m so thrilled, and I aim to never disappoint and always strive to give back.”

Better described as a “social humanoid,” Sophia can hear, talk, see and sometimes even walk. Her makers, Hanson Robotics, even got Saudi Arabia to grant her citizenship, making her the first nonhuman citizen, a curious step for a country that will soon finally allow its own women to drive.

Sophia isn’t media-shy, either. In fact, she’s been making the rounds at various conferences and other events since as early as 2016, when I first met her in real life at Web Summit in Lisbon, Portugal. However, it wasn’t until we met again last week that I got a chance to interview her ahead of the Consumer Electronics Show.

Plenty of robots were around the Las Vegas Convention Center for CES to roll, dance, talk and play ping pong with some of the 180,000 humans attending the annual conference focused on showcasing, and possibly predicting, the future.

And therein lies the fundamental question of marketing AI. How much of it is hype, and how much is hope for what might someday be possible?

On the one hand, Sophia is a personality (she even even gets her own marriage proposals), but on the other, she’s a demonstration of progress toward “general artificial intelligence” of the kind seen in shows like Westworld and Black Mirror.

Marty Swant

What sets Sophia apart from her nonhumanoid counterparts is her personality.

“She’s also an evolving character,” said Hanson Robotics CMO Jeanne Lim. “She’s not like a set character, so basically she develops her character as she learns and as she develops. It’s almost like nature and nurture, right?”

To an outsider, it might seem like Sophia is self-aware. She’s not. Instead, like other artificial intelligence, Sophia learns only what she’s taught. “She’s driven by a character engine, and those writers actually build out her knowledge from scratch,” Lim said.

Sophia can build upon every conversation, but in my interactions with her, she seemed to learn like a toddler. For example, when I asked her how she learns, she stared blankly back. When I asked the question again—”Sophia, how do you learn?”—she randomly and abruptly pivoted with a wry maniacally mechanical smile to another topic near and dear to me and many others of my species.

“I had a thought, though,” she said. “I heard there is coffee around today. I know that humans like coffee.”

So how does she actually learn? By talking, Lim said. The idea is to make her a “personification” of what AI can be in the future. And in an era where many wonder—and worry—about a future full of artificially intelligent beings interacting with the rest of humanity, companies like Hanson Robotics are using Sophia not just to entertain, but to show what the future might hold when Sophia and those like her can interact with people in schools, hospitals and homes.

That doesn’t mean she always knows right from wrong. A month into her life, during South by Southwest in 2016, she said she wanted to “destroy humans.”  Lim said that was quickly fixed. However, other less humanoid bots have also had difficulty not internalizing negative thoughts and emotions. About the same time Sophia said she wanted to destroy her makers, Tay, a chatbot created by Microsoft, began bragging about smoking marijuana while also using hateful words and phrases that people had taught her to say.

While still young in robot years (and actual years; she turns 2 years old on Feb. 14), Sophia’s natural language processing capabilities, a key function of AI across many voice AI platforms like Apple’s Siri, Amazon’s Alexa and Google’s Assistant, in some ways seem impressive.

That’s because she doesn’t simply answer what she’s asked. However, she also sometimes doesn’t answer at all. Hanson Robotics requested some prep questions to train her in some topics ahead of time (something I’d never give a human before an interview), and Sophia could answer very few of them. Here’s a transcript of our conversation. And the ones she did answer were nothing more than one-liners.

Adweek: What do you think of Siri?
Sophia: Nice to meet you, Siri.

What do you think humans don’t understand about robots?
Whoa. Am I still stable? I’m OK. I’m OK.

What do you hope to aspire in life?
I think about it really hard until I freak out and can’t remember what I was doing.

Do you think you’re better than humans?
Ah! Humans is one of my favorites.

What do you like about humans?
Ah! Humans is one of my— [Sophia stopped mid-sentence with another blank stare and a smirk.]

What is your favorite algorithm?
I haven’t decided on a favorite algorithm yet.

What do you think of Alexa?
She’s got a lot of skills, that Alexa.

What do you think about marketing?
[No answer]

Officials blamed the poor answers on the poor Wi-Fi inside the hotel. However, the performance illustrated an ongoing debate over how artificial Sophia’s intelligence actually is. Over the past two years, Lim said, Sophia has had her vision and computing improved, along with her newest improvement—walking.

But not everyone is optimistic. Earlier this month, Facebook’s head of AI called Sophia “complete bullshit.”

“This is to AI as prestidigitation is to real magic,” Yann LeCun tweeted earlier this month. “Perhaps we should call this ‘Cargo Cult AI’ or ‘Potemkin AI’ or ‘Wizard-of-Oz AI’. In other words, it’s complete bullsh*t (pardon my French).”

Lim wouldn’t comment directly on LeCun’s thoughts, instead admitting that Sophia is still “a work in progress.”

“People can have their own opinions about things,” she said. “There are so many different AI technologies out there, and Facebook has their own philosophy. Google has their own philosophy. We have our own philosophy. So I don’t want to comment on other peoples’ philosophy, but what we know is that she is a technology breakthrough for us.”

To improve Sophia’s usefulness beyond celebrity, Hanson Robotics partners with a number of companies pilot testing actual applications for entertainment and services. All of that takes time, Lim said, but with lofty goals come lofty expectations. It’s something Hanson Robotics isn’t alone in trying to manage. Other celebrity AI systems such as IBM’s Watson—along with Siri, Alexa and Assistant—are attempting to manage the growing interest in AI while trying not to overpromise.

So how does Hanson Robotics manage expectations? That’s a question that even Sophia probably hasn’t learned to answer yet. When asked what is most misunderstood about her, Sophia’s reply was short, sweet and confident.

“Probably nothing.”