AI Insights: Geoffrey Hinton’s Trust in GPT-4 Revealed

Geoffrey Hinton, known as the “Godfather of AI,” has openly acknowledged his excessive trust in artificial intelligence, particularly OpenAI’s GPT-4, despite its shortcomings. During a recent interview with CBS, Hinton, who received the Nobel Prize in Physics in 2024 for his pioneering work in machine learning, candidly discussed his dependence on the chatbot for day-to-day activities.

Rethinking AI Dependence

Hinton expressed, “There is a tendency to trust what the AI says, even though skepticism might be warranted,” emphasising the subtle dangers of over-relying on AI systems. In the interview, Hinton challenged GPT-4 with a simple riddle: “Sally has three brothers. Each of her brothers has two sisters. How many sisters does Sally have?” GPT-4 incorrectly responded with two, while the correct answer is one: Sally herself.

Spotting Limitations in AI

“It astonishes me. I find it surprising that it still makes such mistakes,” Hinton remarked, underscoring the persistent limitations of advanced AI models. He described GPT-4 as “an expert at everything, but not an exceptionally good expert,” acknowledging its extensive abilities alongside its occasional inaccuracies.

Optimism for Future AI Developments

Nonetheless, Hinton remains hopeful about the advancements in AI, predicting that future iterations will show significant improvements. “I suspect,” he responded when queried if GPT-5 would answer the riddle correctly.

Rapid Advances in AI Technology

Interestingly, after the segment aired, users on social media noted that newer versions of OpenAI’s technology, such as GPT-4o and GPT-4.1, successfully solved the riddle, demonstrating swift progress in the field.

OpenAI’s GPT Evolution

OpenAI unveiled GPT-4 in 2023, rapidly establishing it as a standard for advanced reasoning and language tasks. Subsequently, it introduced GPT-4o, which features enhanced speed and versatility, while continuing to develop models like GPT-4.5 and GPT-4.1. In the competitive landscape, Google’s Gemini 2.5-Pro currently leads the independent Chatbot Arena leaderboard, although OpenAI’s models closely trail behind.