The Hype vs. The Reality: Rodney Brooks on the Limitations of Generative AI and the Future of Robotics
Rodney Brooks, a renowned roboticist and co-founder of companies like iRobot and Rethink Robotics, is known for his insightful predictions about the future of artificial intelligence (AI). With a career spanning decades and a keen eye for both the potential and limitations of technology, he’s urging us to temper our expectations regarding the capabilities of generative AI, particularly in the realm of robotics.
Brooks, a vocal advocate for responsible AI development, believes we’re prone to overestimate the capabilities of generative AI. While impressive in its ability to perform specific tasks, it lacks the innate human understanding and adaptability necessary for real-world applications beyond its carefully defined training set. He draws a clear distinction between performance and competence, emphasizing that while an AI system may excel at a specific task, it doesn’t necessarily translate into a broader understanding of the world.
“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,” Brooks says. “And they’re usually very over-optimistic, and that’s because they use a model of a person’s performance on a task.”
This over-optimism, he argues, leads to impractical expectations. Take his own company, Robust.ai, which develops warehouse robotics systems. While some suggest using generative AI to give robots directions, Brooks views this as inefficient and potentially detrimental.
“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 explains. “We have massive data processing and massive AI optimization techniques and planning. And that’s how we get the orders completed fast."
Brooks advocates for a practical approach to robotics, one that focuses on solving specific, defined problems in controlled environments. He emphasizes the importance of robots working alongside humans in collaborative, complementary roles. This concept is exemplified by Robust.ai’s warehouse robots, which are designed to look like shopping carts, making them familiar and easy for humans to interact with.
"We need to automate in places where things have already been cleaned up. So the example of my company is we’re doing pretty well in warehouses, and warehouses are actually pretty constrained. The lighting doesn’t change with those big buildings. There’s not stuff lying around on the floor because the people pushing carts would run into that. There’s no floating plastic bags going around. And largely it’s not in the interest of the people who work there to be malicious to the robot," he says.
Brooks’ approach is rooted in the need for clear and understandable solutions that can be implemented at scale.
"I always try to make technology easy for people to understand, and therefore we can deploy it at scale, and always look at the business case; the return on investment is also very important."
However, Brooks acknowledges that even with a pragmatic approach, there will always be unforeseen complexities and outliers that AI struggles to handle. These "long tail" scenarios require specialized, custom solutions, and often involve years of research and development.
"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."
Furthermore, Brooks cautions against unrealistic expectations fueled by Moore’s Law. It’s a common misconception to believe that technology progresses exponentially, leading to the assumption that future AI systems will automatically surpass current ones in capability. This fails to recognize the limitations of technological progress and the inherent challenges in replicating human intelligence.
Brooks uses the example of the iPod, which initially doubled in storage size with each iteration. However, this exponential growth could not continue indefinitely. "If it had continued on that trajectory, we would have an iPod with 160TB of storage by 2017, but of course we didn’t." The reality? The limitations of physical storage and user needs dictated a different path.
While Brooks acknowledges the potential of generative AI to contribute to future robotics, particularly in eldercare, where natural language interfaces could be beneficial, he stresses that true progress lies in addressing the underlying challenges of control theory and optimization.
"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."
Ultimately, Brooks urges a balanced perspective on the capabilities of AI. It’s essential to focus on practical applications and solvable problems, while acknowledging the inherent limitations and complex challenges. As he aptly points out, the journey towards truly intelligent robots is far from over.
"It’s not useful in the warehouse to tell an individual robot to go out and get one thing for one order, but it may be useful for eldercare in homes for people to be able to say things to the robots,” he says. This nuanced understanding of AI’s strengths and weaknesses will guide us towards building a future where humans and robots collaborate effectively, addressing real-world challenges in a safe and sustainable manner.
