Does Sam Altman Have a Telling Habit When He’s Lying?
404Sam Altman, the CEO of OpenAI, has often been accused of stretching the truth, especially during public talks and interviews. One of his critics, AI researcher Gary Marcus, suggests there’s a tell that reveals when Altman isn’t being completely honest. Marcus believes that whenever Altman looks up and away from the camera while speaking, he might be hiding something or bluffing.
Back in February, Altman was caught on video during a visit to Japan. He was talking about GPT-5, the next big thing in AI at the time. He made bold claims about its capabilities, suggesting it would be smarter than previous models like GPT-4. But as it turned out, GPT-5 didn’t meet those lofty expectations. Instead, it was seen as underwhelming, with many users saying it was actually worse than GPT-4. Shortly after, OpenAI announced that GPT-4 would be coming back for paying customers, showing how much the new model had fallen short.
During the February interview, Altman was very confident as he described the progress from GPT-3 to GPT-4, and then to GPT-5. But Marcus points out that every time Altman spoke about the leap to GPT-5, he looked up and away. This was a consistent “tell,” according to Marcus, that Altman might have been bluffing. When he talked about the models’ intelligence, Altman again looked away, almost as if he was trying to hide the truth or avoid directly answering.
Marcus’s point is that Altman wasn’t necessarily lying out of malice. Instead, he might have been selling a vision of the future to investors. OpenAI was aiming for a huge valuation—around $500 billion—and needed strong confidence from stakeholders. Altman’s bold claims about GPT-5 and the goal of achieving artificial general intelligence (AGI)—a human-level AI—were part of that vision. He talked about combining different models to create a single, unified AI that could do everything, which would be a major breakthrough.
But in reality, GPT-5 didn’t turn out to be the all-encompassing, super-smart AI Altman described. It has obvious flaws: it doesn’t recognize past presidents, it switches errors between models due to how OpenAI combined previous versions, and it’s fairly easy to bypass its restrictions. The hype didn’t match the actual product.
Despite the gap between promises and reality, Marcus makes a convincing argument that Altman’s confident posture and the way he looks away when making certain statements suggest he might have been bluffing about the AI’s true capabilities. It’s a good reminder that even leaders of big tech companies can sometimes stretch the truth when talking about the future.
Altman’s confidence isn’t just about deception; some see it as a way to inspire investors and build excitement around AI advancements. Still, the disconnect between the bold claims and what GPT-5 actually can do raises questions about how much of the hype is real versus strategic storytelling. As OpenAI continues to develop its models, only time will tell if the vision of AGI becomes reality or remains a lofty dream built on bold promises.
What do you think?
It is nice to know your opinion. Leave a comment.