GPT-6: Altman Hints at a More Personal AI

OpenAI’s CEO Sam Altman has shared his vision for GPT-6 during an interview on CNBC’s Money Movers. He said the next model will focus on long-term memory. Current versions of ChatGPT can already recall information, but GPT-6 is expected to “know” its user on a deeper level. It will adapt in subtle ways, almost like a true partner. Altman explained that OpenAI is working with psychologists to make this interaction more human and more supportive.

Shaped by the User

At launch, GPT-6 will remain neutral. The user will decide what kind of assistant it should become. This could include personality traits or even political leanings. Altman noted that the model could be set up as a conservative if the user prefers. The idea is simple: the AI should not only answer questions but also understand who is asking them.

Privacy as a Core Issue

The more data an AI collects, the greater the risk if that data is exposed. Altman admitted that privacy is one of the hardest challenges ahead. The team is considering encryption as a safeguard, although no timeline has been set.

Other developments at OpenAI:

• Work on GPT-6 has already started.
• The gap between GPT-6 and GPT-5 will be shorter than the one between GPT-4 and GPT-5.
• An experimental model can now think for up to 90 minutes.
• That same model recently won gold at the International Math Olympiad and the International Olympiad in Informatics.

OpenAI has not announced when this experimental system will be released to the public. For now, Altman wants GPT-6 to bring AI closer to being a true companion.