According to AI expert Rodney Brooks, the perceived intelligence of AI bots like ChatGPT and StupidGPT has been greatly exaggerated. In an interview with IEEE Spectrum, Brooks argues that these language models are much less intelligent than commonly believed and cannot compete with human intellect in any task. He suggests that we have made numerous errors in predicting the future capabilities of AI.
Brooks emphasizes that these AI tools excel in logic and language but lack a genuine connection to the real world. He points out that they merely possess different correlations between languages, indicating their limited understanding.
His comments serve as an important reminder of the current constraints of AI technology and the tendency to assign meaning to their outputs, despite their primary purpose of mimicking human-like responses rather than reasoning like humans.
Brooks further explains to IEEE Spectrum that while we can assess a person’s broader capabilities based on their actions, the same generalization methods do not apply to AI systems.
The current language models create an illusion of meaning without being able to logically infer it, leading to potential confusion for users. Rodney Brooks emphasizes that large language models excel in producing answers that sound correct but are not necessarily accurate.
Brooks himself conducted experiments using these models to assist him in complex coding tasks, but he faced significant challenges. Despite their confident responses, he found that they were often incorrect. He shares his experience of spending hours following their suggestions, only to realize they were ineffective, leading to a need for an entirely different approach.
According to Brooks, this behavior does not constitute intelligence or genuine interaction. Instead, it is simply a process of looking up information without the ability to reason or understand the context fully.
If Rodney Brooks’ assertion about the current limitations of AI bots is indeed accurate, it raises significant concerns for our future. The widespread belief that AI is capable of performing most jobs, including entry-level and certain high-level positions, has already led to a drastic downsizing within the tech industry, resulting in the termination of thousands of employees.
If we assume Brooks is correct, then all the emotions and reactions we associate with AI, be it euphoria or fear, would be nothing more than a manifestation of mass hysteria on an unprecedented scale.
If Brooks’ statements prove to be true, one can only imagine the catastrophic consequences as AI advances closer to the point of singularity. The potential impact and upheaval caused by AI in such a scenario would be immense.