Episode #183 ... Is ChatGPT really intelligent?

Stephen West delves into the recent advancements in artificial intelligence (AI), particularly with large language models like ChatGPT. He reflects on how the launch of ChatGPT in November 2022 has made AI more accessible and discussed widely, even capturing the imaginations of those who engage with it. He sets the stage for deeper philosophical questions about whether these AI models truly exhibit intelligence in a way comparable to human beings.

West explores the historical context of machine intelligence, focusing on the Turing Test, proposed by Alan Turing, which evaluates whether machines can exhibit intelligent behavior indistinguishable from that of a human. He cites philosopher John Searle, who challenges the validity of the Turing Test by differentiating between syntax (how computers process information) and semantics (how humans understand meaning). This section questions the assumptions we make when asserting that AI demonstrates real intelligence.

To illustrate his critique, West explains Searle's Chinese Room thought experiment, which posits that a person following scripted rules to respond to Chinese symbols could appear to understand the language without actually comprehending it. This argument underlines the distinction between mere pattern recognition in AI and true understanding that characterizes human intelligence. The conversation shifts to whether machines can truly possess minds or if their processing capabilities will always remain superficial.

Stephen introduces Noam Chomsky’s viewpoint on the false promises of ChatGPT and other current AI technologies, emphasizing a significant discrepancy between the capabilities of AI and the complexity of human thought processes. Chomsky argues that while AI can be an impressive engineering feat, it lacks the scientific underpinning and genuine understanding that humans possess. He raises concerns about the limitations of language models in generating meaningful theories and insights.

The podcast concludes with a discussion on the societal implications of misunderstanding AI, highlighting a threefold danger: corporate hype surrounding AI technologies for funding, a futurist mindset that sees AI as a savior, and the risk of inaction regarding actual major threats facing humanity like climate change and nuclear war. It prompts listeners to be wary of attributing too much intelligence or authority to AI systems based on misconceptions.