In science fiction, AI navigates the unknown. It’s not yet smart enough to do that
In the movie WALL-Esome of the last Earthlings cross the Kuiper belt aboard the spacecraft Axiom. For 700 years, a team of fully automated robots have been taking care of it after our planet became uninhabitable. Managing the ship is AUTO, an artificial intelligence system that works to keep humans away – forever.
Here at home, space agencies like NASA are using AI to explore the solar system. They pilot rovers on Mars, avoid satellite collisions and train astronauts for spaceflight. But for now, spacefaring humans would be ill-advised to rely so much on AI.
Today’s AIs are much more error prone and chess than we see in fiction, says Daniele Gammelli, a roboticist at Stanford University who studies how to integrate AI systems into robots that interact with their environment.
AI systems in space robots are expected to perform multi-step tasks in all kinds of scenarios without producing inaccurate information. In space, Gammelli says, “you have virtually no margin for error.”
The title robot in WALL-E is a trash compacting machine that abandons its functions to follow another robot, EVE. Its greatest strength undoubtedly lies in its ability to manage change. The robot escapes from a self-destructing pod using a fire extinguisher. In the event of a malfunction of its wheel or eye, WALL-E can replace the damaged part. All this is learned through experience and is done without additional programming.
Such versatility is an example of general artificial intelligencean AI capable of thinking and learning in different situations and performing tasks for which it was not programmed. AGI does not exist yet.
Adapting to unforeseen situations is a major goal for future space robots, Gammelli says. Between extreme temperatures, radiation and space debris, space is a constantly changing environment. “The kinds of scenarios you put on your robot are, by definition, things that no one has ever seen,” he says.
Today’s AIs excel at single or closely related tasks, as well as repetitive, predictable work. Their core skill “is processing a huge amount of data in a very efficient way,” says Sanjoy Paul, a computer scientist at Rice University in Houston who studies how AI can help space missions.
Mars rovers use this type of AI, all without human intervention. For example, Perseverance uses AI algorithms to analyze minerals and determine if rock samples are worth collecting. A human sorting through this kind of data might feel overwhelmed, Paul says. “AI can analyze all the details…and highlight those things for humans to take a look at.” »
To handle multi-step tasks, almost all space robots rely on “autonomy stacks,” Gammelli says. Separate modules responsible for different actions are linked. An AI model can detect rocks or obstacles using cameras or sensors. This information would be passed to another module to interpret and determine appropriate actions. Other modules would then perform physical maneuvers to accomplish the job.
On board Axiomrobots manage everything. Guard robots clean and polish. Utility robots perform repairs and maintenance. Flying chairs transport the ship’s residents to their destinations. AxiomTrain passengers lead a sedentary lifestyle, watch videos and drink shakes.
In reality, “humans always need to know,” says Paul. Even as AIs continue to improve, they remain unpredictable. “If your life depends on it, would you really rely on AI? Probably not,” he says.
Machines like rovers should eventually be able to create their own mini-goals that align with the overall mission, Gammelli says. This capability would allow robots to better handle unforeseen situations and free humans to take care of more crucial tasks and decisions. “We want these robots to be as independent as possible,” he says. But maybe not as independent as WALL-E leaving his mission.