Backflips are easy, stairs are hard: Robots still struggle with simple human movements, experts say

However, speakers at the Fortune Brainstorm AI conference on Tuesday cautioned against being too fascinated by stunts. A robot doing a backflip – something that is difficult for a human – looks impressive. But when you ask a robot to perform seemingly easy tasks, like climbing stairs or drinking a glass of water, many of today’s robots still struggle.

“What looks hard is easy, but what looks easy is really hard,” explained Stephanie Zahn, a partner at Sequoia Capital, paraphrasing a remark from computer scientist Hans Moravec. In the late 1980s, Moravec and other computer scientists noticed that it was easier for computers to perform well on intelligence tests, but they failed at tasks that even young children could do.

Deepak Pathak, CEO of robotics startup Skild AI, explained that robots, and computers in general, were good at doing complex tasks when working in a controlled environment. While showing a video of Skild’s robot jumping on the sidewalk, Pathak noted, “Except for the ground, the robot doesn’t interact with anything.”

However, for tasks like picking up a bottle or walking up the stairs, a person uses vision to “constantly correct what they are doing,” Pathak explains. “This interaction is the root of general human intelligence, and it’s something you don’t appreciate because almost every human has it.”

Zhan explained that viral videos of humanoid robots do not show how the product has been trained, nor whether it can operate in an unsupervised environment. “The challenge for you as a consumer of all these videos is to really discern what’s real and what’s not real,” she said.

The next step for robotics

However, both speakers were optimistic that advances in general intelligence will soon lead to more advanced and flexible robots.

“The robots were driven more by human intelligence. They would have been looked at by a very intelligent person [a task]“And…pre-program the robot mathematically to do this,” Pathak said.

He explained that the field of robotics is now shifting from “programming something to learning from experience.” This allows new robots that handle more complex tasks in uncontrollable environments, which can be easily adapted for other tasks without the cost of reprogramming and retooling.

Robotics companies today “are still limited to having robots designed for only specific things,” Zhan said. A robotics platform with more general intelligence could open up “possibilities that we would not otherwise have been able to achieve,” including tasks that currently pose a risk to human workers.

Consumers can benefit too. “You see all these home robots, but they can only do one thing,” Zhan said. “But if we succeed in building universal intelligent robots, you will finally have consumer robots that can handle the full range of household tasks you do now.” A similar point was made earlier at Brainstorm AI by Arm CEO Rene Haas, who said that the general adaptability of humanoid robots would make them better suited to factory jobs than the robotic arms used today.

There are social ramifications to the robot boom, displacing jobs that still have to be done by humans. However, Pathak was optimistic about the social benefits of deploying automation. The first is safety, as robots eliminate the need for humans to do dangerous or unhealthy work in the long term. Another benefit is filling massive labor shortages in blue collar and manufacturing jobs. (This deficit was an obstacle to the efforts made by the United States to re-support a large portion of its advanced industries from Asian economies.)

However, Pathak also envisioned a future in which robots liberate humans from the drudgery of daily work, even as he acknowledged that societies needed to figure out how to spread the gains from automation. “There is a scenario, a good scenario, where everyone is doing the things they love,” Pathak said. “Work is becoming more optional, and they are doing things they enjoy.”

revision: Due to an editing error, a previous version of this article misstated the name of the company led by Rene Haas. He is the CEO of Arm.