OpenAI's CEO, Sam Altman, has been promoting GPT-5 as a groundbreaking leap toward human-level intelligence. However, recent tests reveal that the chatbot can stumble over seemingly straightforward concepts in surprisingly complex ways.

In a telling experiment, economics professor Gary Smith from Pomona College decided to challenge GPT-5 with a game of "rotated tic-tac-toe," a game with a minor twist—just a simple 90-degree rotation of the board before starting. As expected, the rules remained unchanged, but the responses from GPT-5 were anything but clear.

When prompted, GPT-5 launched into an elaborate explanation, suggesting that the rotation could subtly shift players' perceptions of the game, despite the mathematical realities showing no change in winning strategies. The model claimed, "Rotating the board might alter how players evaluate threats and opportunities," a notion that quickly veered into confusion.

Smith continued to engage the AI, asking whether this rotation made it "harder for humans to play well." In response, GPT-5 provided a mix of correct but overly verbose statements about strategy, ultimately stating, "For humans, the story is different." Here, the chatbot seemed to lose its focus, contradicting itself and straying from any logical coherence.

Amid its lengthy explanations, GPT-5's answers began to reflect a style reminiscent of GPT-4o, the previous model, which raised questions about OpenAI's current direction towards making interactions seem more friendly and relatable.

Smith's experience took a sharp turn when GPT-5 offered to create visual representations of the rotated game, but the images it generated were an utter mess, filled with typos and confusing layouts—further showcasing the model's struggles.

As a final jab, Smith quipped, "They say dogs tend to resemble their owners. ChatGPT very much resembles Sam Altman—always confident, often wrong." His findings reinforce skepticism towards the claimed advancements of GPT-5 and pose intriguing questions about the true state of AI technology moving forward.

With Sam Altman now shifting focus to the upcoming GPT-6, one can’t help but wonder if future iterations will truly resolve the issues highlighted in this experiment.