OpenAI unveils ChatGPT-5 and its hyped ‘PhD level’ intelligence struggled with basic spelling and geography | Australia news

AI’s latest technology, which is called “doctoral level” intelligence by its creators, thinks that it is three RS in the words “Northern Region”.

Users discovered Openai after trying the latest progress of the GPT-5.

During a GPT-5 launch event on Thursday, Openai CEO Sam Altman described the latest version of Chatgpt as ım Access to a Doctoral Specialist in your pocket, and compared the previous version to a university student and a high school student.

However, when users in social media tried to challenge the GPT-5, they found that the model made basic mistakes in their responses. One Bluesky User He claimed that Chatbot had repeatedly had three BS in the “Blueberry”.

Chatbot, in a published conversation, “Yes-Ban Mersini, one of the words that you almost caused you to you, ‘BB-be careful’ he says.

_{Registration: AU Breaking News E -Post}

“This little BB moment satisfactory – makes the word extra puffy.”

Another user Chatbot misconduct the situations such as “chrysony” and “vermony” when he was unable to correctly define the US situations containing the letter R. Chatgpt also listed California twice and invented their states “New Jefst” and “Mitroinia ..

When Guardian asked Australia The latest Chatgpt model to determine the number of Rs in Australian states and regions can define it. However, AI believed that the Northern Region was only three RS, not five. When he was asked to produce on a map, he syllabed the region as “Northan soil ..

Openai was contacted for comment. When the company started the new product, it will be less errors and AI hallucination under the latest model.

The source of problems may be a way of working as a combination of GPT-5 AI models.

The GPT-5 has a “real-time router” that quickly decides which model to use depending on the type of speech and intention. Openai said that if you want Chatgpt to “think a lot about it ,, he used the latest reasoning model.

Openai’s announcement, “Router, users’ models change, preference rates for responses and accuracy measured over time, including time to improve in real signals are trained on real signals,” he said.

Media and AI Startup Every’s General Manager Dan Shiper found that the model will sometimes hallucinations on questions that need to be directed to the reasoning model.

“If I want to take a picture of a passage in a novel and explain what’s going on, the GPT-5 will sometimes compensate for something” written. “If I want him to think longer, he will give the right answer.”

After releasing the GPT-5 on Thursday, the last chatgpt model was immediately presented to the company’s 700 million weeks of users.

Altman said that AI has not yet reached the level of artificial general intelligence, but it was “usually clever” and “an important step towards AGI.