Leading expert on robotics and artificial intelligence (AI), professor Rodney Brooks, calls for a realistic view of AI's potential, avoiding excessive hype.
Professor Rodney Brooks, a big name in the field of robotics and AI, is currently Professor Emeritus of Robotics at MIT and co-founded three prestigious technology companies: Rethink Robotics, iRobot and Robust.ai. In a recent interview with TechCrunch, he spoke candidly about the current state of artificial intelligence development, especially large language modeling (LLM).
According to the professor, although LLMs are very impressive, their capabilities have not yet reached the level that many people mistakenly believe. “The problem with generative AI (GenAI) is that while it can only perform certain tasks, it cannot do everything a human can do,” he said.
Professor Rodney Brooks - leading expert on AI
Using the example of Robust.ai, his logistics robotics company, Professor Brooks said that using LLM to control robots in a complex warehouse environment with thousands of orders that need to be processed quickly is impossible. and can even slow down processing. Instead, connecting the robot to warehouse management software will be much more effective.
According to Mr. Brooks, people often equate the capabilities of AI with humans, leading to the expectation that AI can do everything like humans. However, he asserted: “AI is not human and it is wrong to attribute human capabilities to it.”
Instead of trying to create human-like robots, Robust.ai focuses on designing robots with practical uses in logistics. Their robot is shaped like a shopping cart, making it easy for humans to interact and control when needed.
The professor also rejected the notion that technology, including AI, will always develop exponentially. He used the example of the iPod. Although the iPod's storage capacity doubled with each initial generation, this growth rate slowed down when it reached a level that was sufficient to meet user needs.
However, Mr. Brooks believes that LLM has potential applications in the field of household robotics, especially in the context of an aging population and health care workforce shortage. However, he also emphasized that this can come with many unique challenges. “The problem is not with large language models but with control theory and other complex mathematical optimization techniques,” he said.
Mr. Brooks concluded that the key to success in the field of AI is to create technologies that are easy to understand, easy to deploy on a large scale, and bring practical benefits to users. At the same time, it is necessary to acknowledge that there will always be exceptions in AI that are difficult to solve and will take years or even decades to fully resolve.