The Real Breakthrough in Robotics
While flashy videos of robots performing backflips and parkour continue to captivate social media users, industry leaders say these physical feats are not the most significant milestones in robotics. Instead, experts argue that the true revolution lies in robots developing the ability to think for themselves.
At the Fortune Brainstorm AI conference in San Francisco, held in December, executives highlighted that robots are now moving beyond pre-programmed tasks. They are starting to learn from data and experiences, much like how large language models such as ChatGPT operate.
From Code to Learning
For decades, robotics relied heavily on human-engineered code. Experts like Stephanie Zhan, a partner at Sequoia Capital, and Deepak Pathak, CEO of Skild AI, say that paradigm is now outdated. Instead of coding every movement, developers are training robots using data, allowing them to adapt to new environments and tasks.
“The change is that things in robotics used to be driven more by human intelligence,” said Pathak. “What has now changed is that these models or these robots can now learn from data.”
Pathak’s background is a testament to this shift. Growing up in a small town in India, he taught himself programming by writing code by hand and using limited internet access at a local café. He later earned a Ph.D. in artificial intelligence at Berkeley and worked at Facebook AI Research before co-founding Skild AI.
Why Opening Doors Is Harder Than Backflips
It might seem counterintuitive, but teaching a robot to open a door is much harder than programming it to flip through the air. This is due to a concept known as Moravec’s paradox, which suggests that tasks humans find simple are often the most difficult for robots to perform.
“It’s actually a lot easier to program a robot to do a backflip than it is to get them to climb stairs,” said journalist Allie Garfinkle during the panel discussion. Both Zhan and Pathak agreed, explaining that physical stunts are physics problems, while real-world tasks require constant sensory input and decision-making.
Interacting with unpredictable environments—like picking up a fragile glass or navigating a cluttered hallway—demands what Pathak calls “sensory motor common sense.” This is a form of intelligence that humans take for granted but is incredibly hard to replicate in machines.
Building the Brain for Every Robot
Companies like Skild AI are working toward developing general-purpose intelligence that can function as the “brain” for any kind of robot. This software would enable different robotic hardware to operate intelligently in a variety of environments, reducing costs and increasing scalability.
Zhan compared this to the rise of generative AI, noting that just as OpenAI opened up new markets for digital knowledge work, robotics companies are now targeting the vast market of physical labor. The vision is to create intelligent systems capable of handling tasks in semi-structured environments like hospitals and hotels before eventually reaching the complexities of private homes.
Data Challenges and Market Opportunity
One major hurdle facing robotics is the lack of large-scale training data. While language models can be trained on the vast content of the internet, there is no equivalent for physical interaction data. Pathak believes that the first company to deploy effective robots in the field will gain a “data flywheel” advantage—robots will generate the data needed to improve their own performance.
“Deploying robots early creates a feedback loop where they get smarter over time,” he explained.
A Future of Safer, Smarter Work
Far from simply replacing human jobs, robotics could help address critical workforce gaps. Zhan and Pathak emphasized the “Three S’s” guiding the future of robotics: Safety, Shortages, and Social evolution.
Robots can take on dangerous or physically taxing roles, reducing human exposure to risk. With millions of jobs going unfilled due to labor shortages, intelligent machines could fill essential roles in logistics, healthcare, and manufacturing. Ultimately, the goal is to make mundane or dangerous work optional, allowing humans to pursue more fulfilling careers.
“We see this as an opportunity to enhance human life, not replace it,” said Zhan.
Looking Ahead
Although robots that fold laundry or cook dinner in homes may still be years away, the groundwork is being laid now. Industrial and service environments will likely see the first wave of intelligent robotics, setting the stage for broader adoption in the future.
As companies continue to merge AI and robotics, the future where machines can think, learn, and adapt is closer than ever. The next era of innovation isn’t just about robots moving—it’s about them understanding.
This article is inspired by content from Original Source. It has been rephrased for originality. Images are credited to the original source.