When Rodney Brooks talks about robotics and artificial intelligence, it’s worth paying attention. As the Panasonic Professor of Robotics Emeritus at MIT and a co-founder of influential companies such as Rethink Robotics, iRobot, and Robust.ai, Brooks has a wealth of experience and insight. He also led the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) for a decade, starting in 1997.

Brooks frequently makes predictions about AI’s future and even keeps a scorecard on his blog to track his accuracy. Despite the current excitement surrounding generative AI, Brooks suggests it may be time to temper expectations. He acknowledges the technology’s impressive capabilities but warns that it isn’t as all-encompassing as some believe.

Generative AI: Impressive but Limited

“I’m not saying LLMs (large language models) are not important, but we have to be careful with how we evaluate them,” Brooks told TechCrunch. He explained that while generative AI excels in specific tasks, it can’t match the breadth of human abilities. “When a human sees an AI system perform a task, they immediately generalize it to things that are similar and make an estimate of the competence of the AI system; not just the performance on that, but the competence around that,” he said. This often leads to overestimating AI’s capabilities.

Practical Applications and Limitations

Brooks shared an example from his latest company, Robust.ai, which develops warehouse robotics systems. Someone suggested using an LLM to direct the robots, but Brooks found this impractical. “When you have 10,000 orders that just came in that you have to ship in two hours, you have to optimize for that. Language is not gonna help; it’s just going to slow things down,” he explained. Instead, connecting robots to data streams from warehouse management software is far more efficient.

Focus on Solvable Problems

Brooks emphasizes the importance of targeting solvable problems where robots can be easily integrated. He highlights the success of Robust.ai in warehouses, environments that are relatively controlled and predictable. “We need to automate in places where things have already been cleaned up. Warehouses are pretty constrained. The lighting doesn’t change, there’s no clutter on the floor, and people are generally not malicious towards the robots,” he said.

Human-Robot Collaboration

Rather than building human-like robots, Brooks’s company designs practical robots for warehouse operations. “These look like shopping carts with a handle. If there’s a problem with the robot, a person can grab the handle and do what they wish with it,” he explained. This approach makes the technology accessible and purpose-built, focusing on ease of use and deployment at scale.

The Challenge of AI Deployment

Brooks acknowledges that there will always be challenging outlier cases in AI deployment. “Without carefully boxing in how an AI system is deployed, there is always a long tail of special cases that take decades to discover and fix. Paradoxically, all those fixes are AI complete themselves,” he said.

The Myth of Exponential Growth

Brooks also cautions against the belief in perpetual exponential growth in technology, inspired by Moore’s law. He uses the example of the iPod, which saw rapid increases in storage capacity early on but eventually plateaued because consumer demand didn’t require more. Similarly, he suggests that the trajectory of AI advancements may not continue to accelerate indefinitely.

Future Prospects

Despite his cautious stance, Brooks sees potential for LLMs in specific applications, such as domestic robots for eldercare. “People say, ‘Oh, the large language models are gonna make robots be able to do things they couldn’t do.’ That’s not where the problem is. The problem with being able to do stuff is about control theory and all sorts of other hardcore math optimization,” he said. He envisions future robots with useful language interfaces for assisting people in care situations, although this comes with its own challenges.

Brooks’s insights urge a balanced perspective on AI, emphasizing practical applications and realistic expectations while recognizing the ongoing potential for innovation.

By Impact Lab