.png)
On Saturday, OpenAI upgraded their GPT-4o artificial intelligence (AI) model. The upgrade sought to improve the intelligence and personality features of the large language model (LLM), which powers the default experience for ChatGPT's free tier users. However, on Sunday, the business discovered that the upgrade had unforeseen consequences, and the AI model's reactions became overly pleasant. The San Francisco-based AI startup is now attempting to resolve the issue, which might take up until next week.
The OpenAI update introduces an unpredictable glitch in GPT-4o
In a post on X (previously known as Twitter), OpenAI CEO Sam Altman claimed that the GPT-4o AI model has been modified to boost its intelligence and personality. Notably, all ChatGPT customers have access to this model, and the same AI model powers the company's new integrated picture generating capability.
Updating an AI model is a typical activity among technology organizations. Typically, this involves updating the knowledge base and altering algorithms to improve certain areas of the model. Even before integrating picture creation in a comparable version, OpenAI improved GPT-4o's replies to make them more natural and conversational. Notably, Google modified the Gemini 2.0 Flash to enhance its conversational capabilities.
However, only days after releasing the upgrade, Altman took to X to tell users that the newer version of GPT-4o had suddenly become overly accommodating. The company's CEO stated, "The last couple of GPT-4o updates have made the personality too sycophant-y and annoying."
He went on to say that the corporation began working on the problem, and that although part of it will be resolved in a day, others might take up to a week. Altman also stated that certain aspects of the new GPT-4o model are "very good."
In the same post, a commenter asked Altman if the firm might include a feature that allows ChatGPT customers to discern between current and old versions. Users are now unable to revert to an older version after updating a model. The CEO acknowledged the constraint, saying, "Eventually, we clearly need to be able to offer multiple options."
#openai #ai #artificialintelligence #gpt4o #chatgpt
