Switzerland was well represented at the 2017 South by Southwest festival in Austin, Texas, which hosts music, podcasts, film, VR and AR, games, and panel talks on the future of entertainment, creativity and research. It’s also an opportunity to dive deep into tech and creative developments across the US. This year, the conversation was dominated by Artificial Intelligence and Machine Learning: where they are, and where they’re going next.

San Francisco has a unique perspective on tech, with the benefit of being in the midst of some real cutting-edge, future-is-now scenarios: self-driving taxis in the street, for one. But that also skews our understanding of what’s “already here.” In Austin, panel after panel explored experiments in AI that revealed the limits of AI, rather than the hype. Here’s what we learned.

AI creativity is boring

As a special project for Saatchi & Saatchi’s New Director Showcase, Team One presented a music video written, conceived, and edited using a hodgepodge of machine learning technologies. A Japanese chatbot was “interviewed” to obtain its creative vision; facial recognition software was used to cast the actors. Even the movement of the cameras were the result of autonomous drones, following flight paths laid out by EEG data.

The experiment did present some strange and “creepy” moments: when the AI, and the casting directors, came up with the same choice, they felt it was evidence that the AI could recognize facial expressions that humans pick up on an unconscious level.

The machine made some strange choices (feeding the cast and crew dried sweet potatoes, for example). And while the human crew was pleased with some of the raw footage, in the end, the mission was total AI control of the project. That meant the final cut of the video was determined by an algorithm, which studied the edit structure of videos with a similar tempo and generated thousands of “versions” per minute before choosing one that nobody seemed to like. 

The band distanced itself from the project. Even the team that lead the project described the result as a “my first video” level of quality. Watching the video, it was mediocre, but not atrocious. A terrible video would have been more interesting.

Used this way, AI as a creativity tool resembles a high-tech (and high-hype) version of Brian Eno’s 1975 “Oblique Strategies” card deck, full of random creative techniques you could use when facing a particular creative challenge. Open up the box, draw a card, get inspired. AI lets us do it with a screen instead of a card deck, but calling that a tech innovation feels like a reach.

While it’s clear that AI can’t handle the full process of the video, interesting things happen when it’s used to augment creativity, which was lesson two.

AI can’t create, but can replicate

University College London lecturer Ruairi Glynn was part of Team One, and offered some ideas for how AI would impact creative work. An artist himself, he uses AI in his own projects, such as building robotic lights that perform short routines. Audience members observe and select the most pleasing motions, helping the robot evolve through “practice.” As boring routines die out, each machine develops its own “personality” – a set of tricks based on individualized feedback.

Despite this and a handful of other interesting projects, Glynn insisted that hype had recently washed over AI developments.

“We’ve got to be a little more realistic about where we are,” said Glynn. “We’ve been working on teaching robots to pick stuff up for 50 years.” It still takes an unguided AI nearly 10 minutes to discover how to change a plug from the socket on the left to the socket on the right. That’s far less complex than creating an emotionally captivating story, he says.

Chris Graves, of Team One, does suggest that the immediate horizon may be tumultuous for artists in fields driven by copying, transcribing, or drafting. “If you’re a replicator, it’s time to start looking over your shoulder,” Graves said.

Glynn predicts a transformation akin to the camera’s impact on portraiture artists.

“Portrait painters disappeared within a couple of decades. The result was an explosion of the avant garde,” he said. That raises a new set of opportunities for artists: “What can we bring as human beings? It’s a challenge to find new ways of practicing.”

Fear of AI is profitable

“We’ve made some pretty cool breakthroughs in AI,” says Kate Darling, Research Specialist at the MIT Media Lab and a Fellow at the Harvard Berkman Center, “and now people feel these fictional threats are becoming a reality.”

Darling brought a very pointed critique of Silicon Valley’s AI hype, suggestion that much of the fear on the horizon about AI and killer robots is coming from “a certain type of person,” specifically, the wealthy investors occupying “an insane” amount of influence in funding and policy for AI. There’s a lot of money to be made in pushing the narrative that AI is an existential risk requiring immense resources, she says. That’s cause for skepticism, and emphasizes the importance of sober assessments of AI’s actual capabilities.

That requires understanding what successes in AI and machine learning actually mean. There was a big splash when machines learned how to play classic Atari video games. But AI was only memorizing successful patterns – button-mashing, then throwing out those that didn’t work. Notably, she said, even Pac-Man proved to be too complex for the machines to “solve.”

Rather than focusing on the agenda of self-actualized robots, Darling suggests, we should be focusing on the agendas of the humans who design them.

Realistic AI still has opportunities

Nilesh Ashra is the Director of Creative Technology at advertising agency Wieden+Kennedy. Another AI skeptic, Ashra has nonetheless found success in understanding precisely what the limits of AI are, and developing ideas that take advantage of those restrictions.

Case in point: Needybot, his project to design a fuzzy, adorable, and completely useless robot. The robot is tasked with learning as much as it can about the team at Wieden+Kennedy through interactions with staff. It’s also designed to need humans to help it accomplish the simplest tasks.

The robot is a charming counterpart to the Terminator hysteria about AI, thriving exactly because of what it cannot do. The result is a robot that puts people at ease and encourages interaction between the machine and humans.

“There’s way more work to be done in creative tech that inspires joy, rather than a path-to-market, market-fit approach,” says Ashra. “I love to see work that isn’t just pure pursuit of utility.”

What’s NEXT?

With a lack of transparency in reports and research on robotics, and a generally hyperbolic response in the scale and impact of recent breakthroughs, it may be a while yet before we even catch up to where we think we are.  

That’s bad news for anyone waiting for the singularity, but it means there’s plenty of research and opportunity in AI, robotics, and machine learning. Designing creative, human-centered interactions that play with in the gap between what AI can do, and what we wish it would do, is where we’ll be seeing the next big ideas.

Photo: Industrial Robot By Claudio Moderini, CC BY-SA 2.0, via Wikimedia Commons.