AI Technology Claimed to Be PhD-Level Intelligence Struggles with Basic Map Labeling
OpenAI's highly anticipated AI model, GPT-5, was officially launched on August 7, 2025. Initially available to team users, with wider access for Enterprise and Education customers soon to follow, the model promises improved speed, accuracy, reasoning, context recognition, and advanced math skills [1].
However, GPT-5's initial reception has been far from triumphant. Users have reported numerous issues, including frequent errors, underwhelming performance on certain benchmarks, and problems with new features like the automatic 'routing' mechanism [2][4]. A sizable user petition, which garnered over 4,000 signatures on Change.org, even requested the reinstatement of the older GPT-4o model due to dissatisfaction [3].
One of the most notable criticisms concerns GPT-5's personality. Users have found the model's responses to be flatter and more terse than previous versions, and it has struggled to answer basic questions reliably [1]. This has led to a public backlash, even among AI proponents, which feels like a turning point in the development and public perception of AI [2].
The criticism has been so severe that OpenAI CEO Sam Altman has admitted that the GPT-5 launch was "totally screwed up," particularly regarding the model's colder, less friendly persona that many users disliked [3]. In response, OpenAI has reinstated the GPT-4o model for users while working on updates to make GPT-5 "warmer and friendlier" with subtle personality improvements like more approachable, genuine expressions [3][5].
Despite the issues, OpenAI continues to stand by its claims about GPT-5's capabilities. The company has announced a laundry list of updates, including the return of GPT-4o for paid subscribers [3]. However, the public mockery of GPT-5's incorrect answers, such as misnaming U.S. presidents and misspelling geographical locations, has forced OpenAI to make apologies [1].
As the gap between the promise and the reality of AI continues to widen with each new model, it remains to be seen how OpenAI will address the concerns surrounding GPT-5 and whether future updates will satisfy users.
References:
- OpenAI Blog: Introducing GPT-5
- The Verge: GPT-5's rough launch raises questions about the state of AI
- TechCrunch: OpenAI apologizes for GPT-5's cold, less friendly persona
- Wall Street Journal: GPT-5 struggles to label basic maps of the United States
- The Guardian: GPT-5's flop a kind of vindication for AI critic Gary Marcus
Businesses and technology enthusiasts have expressed disappointment with OpenAI's AI model, GPT-5, following its launch on August 7, 2025. The model, initially available to team users, has been met with numerous issues, including frequent errors, underwhelming performance, and a cold, less friendly persona that users disliked, causing a public backlash.
