“Nothing that could be called GPT-5” – OpenAI is finalizing GPT-o1, and GPT-5 will not appear in 2024

OpenAI CEO Sam Altman dispelled hopes for an early release of GPT-5, saying that the company will focus on improving the GPT-o1 version by the end of 2024. This version is now focused on in-depth analysis and is designed to solve specialized problems in areas such as science, mathematics and academic research. OpenAI also plans to develop independent “AI agents” that can work more independently, without human intervention.

Image source: alanajordan/Pixabay

During a conversation with Reddit users, Altman clarified that the next version of ChatGPT, GPT-5, is not planned for release in 2024. “We will have some interesting releases by the end of the year, but nothing that can be called GPT-5,” he said. Instead, the company will focus on releasing a version of GPT-o1 designed for a more thoughtful approach to problem solving. Also known by the codename Project Strawberry, this version of ChatGPT is aimed at specialized use cases that require thoughtful decisions and accurate responses, especially in scientific and academic fields.

Altman noted that the increased complexity of modern AI models makes it difficult to develop major updates in parallel. In addition, OpenAI faces severe constraints and difficult choices when it comes to allocating computing resources, limiting the company’s ability to release multiple major AI model releases simultaneously.

The next significant achievement of ChatGPT will be “AI agents” – systems that can perform tasks autonomously, interacting with the outside world without human intervention. Altman explained that such functions will be able to solve specific problems, for example, booking airline tickets, buying concert tickets or responding to support requests. OpenAI plans to make these capabilities an important part of its AI models, which will significantly expand their functionality.

Vice President of Development at OpenAI Srinivas Narayanan spoke about his vision for the future of ChatGPT, noting that in the future the AI ​​model will be able to better understand the user’s personal information and perform actions on his behalf. This, in his opinion, will significantly expand the functionality of ChatGPT and make it a tool that actively responds to everyday user requests.

Altman also hinted that he might one day open up adult content, Not Safe For Work, which is currently blocked. “We fully support the idea of ​​respecting adult users,” he said, adding that this issue requires serious study and that OpenAI now has more urgent tasks. Altman emphasized that the company plans to return to this issue when the main problems are resolved.

OpenAI’s leadership’s ambitious goals involve significant improvements in the capabilities of its AI models. In May, the company’s chief operating officer, Brad Lightcap, said that within a year, we’ll be laughing at how primitive previous versions of ChatGPT were. Although the release of GPT-5 is delayed, OpenAI is bringing new AI tools to users. Recently, ChatGPT Search was launched, allowing you to search for information on the Internet directly through ChatGPT, which previously required turning to search engines.