OpenAI founder Altman: GPT-5 and GPT-6 still cannot achieve AGI
Altman said that OpenAI is currently focused on improving GPT-4. The future focus of the model is to reduce costs and improve reliability. We look forward to applications such as AI medical advisors and personalized one-on-one tutoring.
Although the market has been envisioning the arrival of GPT-5 and even AGI, Sam Altman poured cold water on it without mercy.
On Monday, Sam Altman, the founder of OpenAI, stated at an event that GPT-5 and GPT-6 will provide higher reliability, personalization, and multimodal output, but there is still a significant gap compared to AGI (Artificial General Intelligence), indicating that AGI will not be achieved in these two versions.
AGI, or Artificial General Intelligence, refers to artificial intelligence that can perform all intellectual tasks of humans, including thinking, learning, planning, natural language communication, and adapting to new environments like humans.
OpenAI has stated that AGI refers to "highly autonomous systems that outperform humans at most economically valuable work," and the company's mission is to ensure that AGI is used for the benefit of all humanity:
We will strive to build safe and beneficial AGI directly, but we will also consider our mission fulfilled if our work aids others in achieving this outcome.
Altman also mentioned in his speech that the focus of future models is to reduce costs and improve reliability.
So far, the scaling laws have applied to large language models; it's a moment of transistors, similar to Moore's Law. Scalability and market forces will make models and the GPT API more cost-effective.
Regarding the applications of artificial intelligence, Altman also mentioned the potential of AI medical advisors and intelligent personalized one-on-one tutoring.
In another recent event, Altman stated that he sees a path to ChatGPT becoming a super cognitive assistant but did not disclose further details.
OpenAI filed a trademark application for GPT-5 in July this year, but Sam Altman has consistently stated in various interviews that GPT-5 is not yet ready for "debut", and OpenAI is still focused on improving GPT-4.
On September 21, OpenAI released the image generation model DALL-E 3 integrated into ChatGPT, and previously introduced new voice chat and image capabilities to ChatGPT.