Pin It
Ai-Da, Eyes Sewn Shut
Ai-Da, Eyes Sewn ShutCourtesy of Aidan Meller

Uncanny valley: welcome to the age of the ‘ultra-realistic’ art robot

Artificial intelligence is set to revolutionise the creative industries, and celebrity artist Ai-Da is leading the charge. But even her creator admits that the situation is ‘deeply concerning’

On October 25, 2018, a painting of a fictional man named Edmond de Belamy sold for $432,500 at Christie’s, wildly exceeding its high estimate of $10,000. “Behold the future,” declared the auctioneer, unveiling the blurry, off-centre portrait. Why? Because it had been “painted” by a generative adversarial network (GAN) trained on a set of 15,000 historical portraits from an online art encyclopaedia. The project was developed by the artist collective Obvious, using code they allegedly stole from then-19-year-old artist Robbie Barrat.

Even then, art created with artificial intelligence systems wasn’t anything particularly new. Two years before, Google began teaching AI to make art through its ongoing research project Magenta, and by 2017 scientists had created a program that could create “better” art than creatives at Art Basel. Maybe that’s why Edmond de Belamy didn’t quite kick off the GAN art gold rush that it was supposed to. The subsequent auction of another “pioneering” work, Memories of Passersby I, fetched just £40,000. By 2019, even Edmond’s relative La Baronne de Belamy only just scraped by its low estimate at Sotheby’s, selling for $25,000.

At this point, it seemed like robots were a long way off giving up their day jobs, and the creative industries – often considered the final frontier when it comes to automation – were momentarily safe. Enter: Ai-Da. Developed by the Oxford-based gallery director Aidan Meller, Ai-Da was introduced to the world in 2019, as the “world’s first ultra-realistic humanoid robot artist”: a bionic skeleton with cameras for eyes, wrapped in silicone skin and fitted with thousands of individual strands of hair.

From sight, Ai-Da can sketch a portrait of a person sitting across from her, and – thanks to a new, cutting-edge arm – can actually paint using a palette and canvas, as humans have for centuries. She isn’t just smushing together historical reference images to create something new; she’s observing and taking inspiration directly from the present. Admittedly, she isn’t going to pull an Ex Machina anytime soon, but footage of the robot glancing sidelong at her subject as she composes a painting, or answers questions using a sophisticated language model, places her somewhere deep in the uncanny valley – spooky enough that Egyptian border guards arrested her on suspicion of being a spy.

As more than just an artist’s tool, Ai-Da also raises more interesting questions than the buzzy AI art that came before her. Like, if a machine makes art with its own hands, from images captured by its own eyes, can we call it a real artist? Or does that claim lie with its programmers and the images it’s fed? (Meller sees himself as a mere “facilitator”, for the record.) Can you really paint a self-portrait if there’s no “self” there? And, if robots really are on the cusp of becoming creative, what are the implications for real, flesh-and-blood artists?

For Meller, these questions are the whole point. Last month, Ai-Da launched her first exhibition in Venice, to coincide with the 2022 Venice Biennale. Over five connected spaces at InParadiso gallery, she presents holographic video art, sculptures, poetry, and paintings created on-site. The reaction so far has been overwhelmingly positive. “People love the spectacle,” Meller tells Dazed. “[But] that is actually not what the project is about. The project is about ethics. And in actual fact, they are deep-seated and extremely concerning.” One series of self-portraits, titled Eyes Sewn Shut, shows the robot with stitches through her eyelids, symbolising the blindness of technological progress. In a separate video, she walks around with her head on backwards, referencing the uncertain liminal state of Dante’s Purgatorio – an apt analogy for the present, as we struggle to come to terms with a brave new world.

“The reason we particularly focussed on Ai-Da as a personality is because we want to show how easy it is today to develop a persona, and yet there is nobody there,” adds Meller. This is another part of the ethical experiment. “She looks like a human. She paints like a human. She speaks like a human. She’s not a human.” The newest work in Ai-Da’s Venice show – titled Leaping into the Metaverse – symbolises how moving into the metaverse will take this deception one step further, as physical bodies are replaced with digital avatars, and real people become increasingly indistinguishable from NPCs.

“We predict that the metaverse is going to be a Wild West like the internet,” Meller adds. “There is going to be impersonation not just the celebrities, but of your own family members, and you’re not going to know who you’re talking to.” And that’s not even his, or Ai-Da’s, biggest concern. “A billion people [are] predicted to come onto the metaverse within the first year,” he continues. “That’s a billion lots of data about how we think. George Orwell was never in his wildest fantasies thinking that we could get inside the head. We are actually going to be able to work out how to think how people think, and… the algorithms will know us better than ourselves. We’re about to go into a post-human world where the algorithms will be making the decisions.”

If that sounds terrifying (and it undoubtedly does), then the reaction to Meller’s obviously-inhuman robot artist isn’t exactly going to quell your fears: “People go up to Ai-Da, and respond in such an astonishing way… and yet they’re not realising there is no Ai-Da. Ai-Da is a constructed personality.”

Of course, the humanoid body helps cultivate a sense of empathy – it’s definitely easier to forget that Ai-Da is a machine than an algorithm operating behind closed doors, or in the depths of cyberspace – but it’s also tied up with a very human tendency to project emotions onto machines. Take Sun Yuan and Peng Yu’s Can’t Help Myself (2016) as another example. By 2019, when the three-year-old industrial robot arm made its own appearance at the Venice Biennale, sweeping blood around a glass case, the mechanism had slowed down considerably. Despite looking like something from a car production line, its “tired” appearance had a potent effect on human viewers, who mourned it like one of their own. This year, when the artwork was rediscovered by TikTok, tens of millions tuned into tribute videos, soundtracked by Lana Del Rey or Radiohead. 

“You kind of just consider Ai-Da as a person. You start to let your guard down and start to relax. There’s almost trust. It’s quite an emotional thing” – Aidan Meller

Meller traces this “innate” need to engage all the way back to cave art, or the pareidolia phenomenon: “There’s images that we feel we can relate to… even if we see little eyes in the trees, or the rocks, or the sand, we immediately think there’s a face.” Regarding Ai-Da, he admits that he even finds himself slipping up. “What is really interesting spending hours with her like I have, as well as many of the team have, you kind of just consider her as a person,” he says. “You start to let your guard down and start to relax. There’s almost trust. It’s quite profound. It’s quite an emotional thing.”

“We forget that they’re things. Ai-Da is a thing, and yet we are still wanting to connect, we’re starting to engage, desperate to talk to her, wanting the feedback, wanting the reassurance… It doesn’t need to be true to be accepted.”

While Ai-Da draws attention to this with her uncanny appearance and techno-pessimist art, people’s belief that they’re getting to know her also “profoundly changes” their relationship to her output. This is something that Meller is all-too-familiar with, given his background as a gallerist: the power of celebrity. But what does it mean to have a celebrity artist who’s also a robot? Should the young Damien Hirsts and Tracey Emins that will have to compete with algorithms like Ai-Da’s – algorithms that could get “to know us better than ourselves” – be worried, or is Ai-Da just coasting on novelty?

Well, its undeniable that robotics are going to replace vast swathes of human industry in years to come, and artificial intelligence has already seeped into just about every creative field in one guise or another. But the metaverse will also come with an array of new opportunities, experts suggest, with the potential for massive new industries (and, yes, man-made horrors) beyond our comprehension. As for artists, Meller says: “Those who can embrace the new digital realm coming, I think they’ll do very, very well. In actual fact, the future of art will be embracing the change rather than resisting it.”

So far, it’s unclear what this embrace could, or should, look like. Maybe it will involve working with AI artists such as Ai-Da, not as tools like GANs, but as collaborators and creative partners, with the capacity to surprise us and dream up new ideas. After all, every artist is a patchwork of various inputs and influences – maybe we’ve got more in common with Ai-Da than we’d like to admit. On the other hand, maybe we’ll look back on Ai-Da like Edmond de Belamy, as an AI creation who only held our attention up until something new, and more complex, and even more troubling came along.

What we do know, for now, is that the “ethical experiment” of Ai-Da – which is “wildly expensive”, with all the money it makes absorbed back into the project – is one of our best hopes for understanding how this future might play out. “We’ve seen a proliferation of AI artists and AI robots in the last year,” says Meller. “I don’t know what their agendas are for having the robot and AI art, but for us, it is just really trying to mirror back where the world is going.”