Skip to main content

More Human Than Human makes the state of AI look ironically grim

More Human Than Human makes the state of AI look ironically grim

/

It’s meant to be about the field’s promise for the future, but it mostly suggests that future is far away

Share this story

If you buy something from a Verge link, Vox Media may earn a commission. See our ethics statement.

Submarine

Welcome to Cheat Sheet, our brief breakdown-style reviews of festival films, VR previews, and other special event releases. This review comes from the 2018 SXSW Interactive Festival.

For the documentary More Human Than Human, which premiered at SXSW 2018, directors Tommy Pallotta and Femke Wolting started with a hooky premise: to explore the current state of artificial intelligence, Pallotta (producer of Richard Linklater’s Waking Life) was going to try to replace himself with a robot. The directors set a robotics lab to work on the project, building a “camerabot” that was meant to scan faces, recognize emotions, train its camera on its subjects, generate questions, and interview them. As the dev team works on that project, Pallotta — a frequent onscreen presence, as interviewer and the intended final interviewee — fills the time by talking to other programmers, roboticists, and futurists about their AI projects or research, trying to build a sense of the state of AI art. “Are we witnessing the birth of a new species?” the film asks. “What will this tell us about intelligent machines… and about ourselves?”

As it turns out, More Human Than Human does say a lot about the state of AI and humanity, but ironically, while it angles toward presenting the field as just on the edge of major breakthroughs that could produce wondrous things, it winds up as more of a portrait of humanity’s tendency toward wishful thinking and anthropomorphizing. It’s a telling, fascinating documentary, but apparently not in the way Wolting and Pallotta openly wanted it to be.

What’s the genre?

Story-doc, the kind of documentary ostensibly built around building up to one specific moment, and following the path to that moment. But the film spends surprisingly little time on the details of the director-replacement project — possibly since by the time they were editing it, Pallotta and Wolting knew the outcome would be anticlimactic. Instead, they bring in movie clips from I, Robot, The Terminator, and 1984 to illustrate the possible dystopias waiting for us, and zip around the world for short interviews.

What’s it about?

The philosophical question of whether robots could really replace people, and whether there are qualities that might be unique to humanity, or unique to machine intelligence.

What’s it really about?

The text is “AI is fascinating, and we’re really close!” but the subtext to many of the projects is that we’re a long way off from creating intelligent robots, but not particularly far from deluding ourselves into thinking we’ve cracked the machine-intelligence code. Pallotta looks in on a number of experiments, including Eugenia Kuyda’s online chatbot constructed from a dead friend’s text messages, and Will Jackson’s RoboThespian, an acting robot. He interviews authors of books about AI or robots, most often returning to Daniel H. Wilson, the outspoken, thoughtful author of the novel Robopocalypse. He visits Oxford’s Future of Humanity Institute and Carnegie Mellon’s Studio for Creative Inquiry. He talks to the designers of Hanson Robotics’ Sophia, a robot granted Saudi citizenship in 2017.

And over and over, he finds the same thing: people hopefully talking to machines as if they can understand, taking whatever vague, gnomic answers they get in return as a positive sign, and either attempting to bridge the gap between regurgitated programmed language and actual understanding, or just acting as if there’s no gap at all.

Submarine

Is it good?

It’s consistently a bit off-putting, because so much of it seems either disingenuous or delusional. Pallotta brings up the case of computer scientist Joseph Weizenbaum, who designed ELIZA, a 1960s chatbot meant to simulate a doctor by asking patients questions about their state of mind, then reflecting their answers back at them. Weizenbaum was disturbed to find that patients took ELIZA’s questions seriously, and forged emotional connections with the program, in spite of its rote, predictable responses. Similarly, a more recent experiment sat senior citizens down with a blank-eyed, plastic-bodied child robot named Alice, which asks questions like “Are you lonely?” And the designers found it left some of the subjects feeling more connected, and like they had someone to talk to.

Oddly, the film presents these examples of one-way emotional investment without acknowledging the imbalance, or examining the ramifications — that some of the scientists and programmers working in AI may be making the same hopeful emotional connections with their projects, even if there’s nothing meaningful there to connect to. It’s particularly strange watching a project team treat Sophia like a functioning person who makes perfect sense, rather than a jerky-faced human simulation asking awkwardly phrased questions that don’t link together. It takes heavy human interpretation to make the onscreen conversation with her flow smoothly, but the film doesn’t address or analyze her limitations, which only makes them feel more prominent.

More Human Than Human could certainly stand to spend more time either exploring the actual mechanics of AI development, or exploring the contradictions created by working in AI. Either way, Pallotta should be asking tougher, more insightful questions. The filmmakers highlight some fascinating projects, but they whisk past most of them in a way that suggests they’re only interested in easily digestible descriptions that fit their planned narrative. There’s nothing, for instance, distinguishing between how Alice, Sophia, or the Pallotta-replacement camerabot are programmed, or what issues researchers are trying to solve. One terrific insight explains that the Alice robot was designed as a child because her appearance makes older people more likely to forgive any awkwardness or verbal errors she makes, but that smart observation — and a moment where Pallotta and crew consciously choose a female voice for their camerabot, because it’s more comforting — just raises more interesting and unaddressed questions about how human subjectivity factors into AI projects.

Submarine

To some degree, the doc just feels like a best-of montage of AI projects worth examining further. Alice, for instance, was the subject of a different 2015 documentary, Alice Cares. And the RoboThespian site makes a much better argument for the unconvincing light-up robots than their creator does in this film. Jackson is a colorful figure, laughing about how acting is the simplest and worst-paid job in the world — “Brad Pitt? How easy is that?” he chuckles — but as seen in the doc, his robots look like clunky animatronics that get all their performance nuance by playing familiar voice clips from actual human actors.

The film certainly has its highlights. One comes when the filmmakers interview the mother of a high-functioning but trivia-obsessed autistic boy who’s found a perfect companion in Siri, an endlessly patient source of answers to his questions. The sequence works particularly well because no one’s suggesting Siri is close to sentient — it’s just a telling example of one specific reason a person might connect with a machine to fulfill specific needs. Another has filmmaker Richard Linklater (who knows Pallotta through Waking Life) and actor Billy Crudup meeting the camerabot and analyzing their own reactions to it, how it feels aggressive and pushy to them. And Pallotta’s final interview with the camerabot is disappointing for him, but darkly hilarious for viewers.

What should it be rated?

G. None of the robots get naked, swear, or go on a wild rampage, slaughtering everyone named Sarah Connor they encounter.

How can I actually watch it?

More Human Than Human is currently seeking distribution.