Artificial intelligence (AI) reached unprecedented levels of excitement in 2023 with the emergence of ChatGPT-3, making AI technology accessible to the general public. However, there may be a potential downturn in the near future, according to Rodney Brooks, a renowned expert in robotics and AI.
Brooks, a former director of MIT’s Computer Science and Artificial Intelligence Laboratory, has been making predictions about various technologies since 2018. In his latest scorecard, he suggests that 2024 won’t be a golden age for AI, cautioning that the current hype is following a familiar pattern in the history of AI.
Despite his expertise in the field, Brooks is not pessimistic but rather pragmatic. He has witnessed the cycle of hype, letdowns, and setbacks in AI throughout his career. His previous predictions have proven accurate, lending credibility to his current skepticism.
Brooks specifically refers to Large Language Models (LLMs) like ChatGPT when discussing AI. While he acknowledges their impressive capabilities, he argues that they lack the potential to achieve true Artificial General Intelligence. These systems may lack genuine imagination and substantial understanding.
In an interview, Brooks criticizes advanced LLMs for their propensity to make mistakes when faced with simple coding tasks. He highlights that their confident answers may be inaccurate, leading to wasted time and effort. According to him, LLMs are adept at mimicking what an answer should sound like rather than comprehending what it should be.
Brooks concludes that LLMs, and potentially future iterations like GPT-5 and GPT-6, are still far from becoming fully-fledged Artificial General Intelligence. They lack a model of the world and a connection to reality – characteristics essential for genuine intelligence.
While AI may not be on the cusp of a technological revolution, Brooks encourages utilizing LLMs for positive purposes. However, he emphasizes that there is much more to life than relying solely on these language models.
In light of Brooks’ insights, the AI community should temper its expectations and recognize that the path to Artificial General Intelligence is complex and multifaceted.