On Thursday, OpenAI CEO Sam Altman and several executives engaged with the public during an ask-me-anything (AMA) session on Reddit. Participants were encouraged to pose questions regarding the company’s offerings, including ChatGPT, or to delve into broader topics around artificial intelligence (AI) and artificial general intelligence (AGI). Altman confirmed that GPT-5 will not debut this year but indicated that OpenAI will unveil “some very good releases” before the close of 2024.
OpenAI Staff Host AMA on Reddit
The AMA took place on the ChatGPT subreddit, which Altman described as “our Reddit launch.” Alongside him, OpenAI’s Chief Product Officer Kevin Weil, Senior Vice President of Research Mark Chen, Vice President of Engineering Srinivas Narayanan, and Chief Scientist Jakub Pachocki participated in the interactive event. The official OpenAI account on X (formerly Twitter) also shared information about the AMA session.
When questioned about the timeline for GPT-5 and its potential substitutes, Altman stated, “We have some very good releases coming later this year! Nothing that we are going to call GPT-5, though.” This aligns with speculation that has emerged in various reports, indicating that a new flagship model from OpenAI may not arrive until 2025.
Another user inquired about the advantages of the SearchGPT feature or the ChatGPT Search functionality. Altman responded by emphasizing that he finds it to be an effective and efficient means of accessing information, particularly for intricate research tasks. He expressed anticipation for a future scenario where a search query could dynamically generate a custom web page in response.
Weil addressed multiple inquiries from users, including one about the delays affecting Sora. He explained that the postponement was necessary for refining the model, ensuring safety from impersonation issues, and scaling computational power. However, he did not specify a launch date for the project.
Weil also mentioned that the ‘o’ series AI models, such as GPT-4o and o1-preview, will continue to be integral to OpenAI’s offerings, remaining relevant even after the introduction of GPT-5. Furthermore, he noted that the upcoming ChatGPT Advanced Voice Mode might include an option to incorporate a singing voice into the AI’s capabilities.
In response to a user suggestion for a hands-free method to end a voice conversation with ChatGPT, Weil expressed enthusiasm about the idea, stating, “I love this idea. Sharing with the team now!”
Mike Chen, OpenAI’s SVP of Research, tackled an important concern regarding AI hallucinations. He acknowledged that while models are improving, the issue is inherently complex because they are trained on human-written text, which can contain inaccuracies. He noted, “Our models are improving at citing, which grounds their answers in trusted sources, and we also believe that reinforcement learning (RL) will help with hallucinations as well – when we can programmatically check whether models hallucinate, we can reward them for not doing so.”
