Overview

• Geoffrey Hinton, the 2024 Nobel laureate in physics, revealed in a CBS interview that he places too much trust in OpenAI's GPT-4 despite its flaws.
• Hinton demonstrated a simple riddle that GPT-4 answered incorrectly, underscoring the model's occasional factual errors.
• He described GPT-4 as broadly capable but not consistently reliable, calling it 'an expert at everything, but not a very good expert at everything.'
• Social media users reported that newer versions, such as GPT-4o and GPT-4.1, correctly solved the riddle, reflecting ongoing improvements in AI performance.
• Hinton expressed confidence that future models, including the anticipated GPT-5, will address such limitations and deliver more accurate results.