“Probably this summer. I do not know exactly when.” In this manner, Sam Altman tentatively set a timeline for OpenAI’s major release of the year. This new version of its popular AI, which the user community has been anticipating since last year, now appears, at last, to be on the verge of becoming a reality.
In any event, Altman is aware of the growing complexity in the naming of his company’s models. As he has explained, unlike before—when a new, large model was launched—current systems are more complex and can now be “post-trained continuously to improve them.” This evolution raises the question of whether updates to GPT-5 should continue to be called GPT-5, or if versions such as 5.1, 5.2, and so on, should be introduced. Altman admits that they do not yet have a clear answer, but they intend to find a better approach than the current system with GPT-4o.
This announcement from Altman occurred during the first episode of OpenAI’s new podcast, which, unsurprisingly, debuted with an interview featuring the company’s CEO. In this interview, Altman reflected on the current state of ChatGPT and the other tools they manage, while also looking toward the future of artificial intelligence and sharing some striking personal insights about his own experience as a user.
The definition of Artificial General Intelligence (AGI) was a central topic. Altman notes that the definition of AGI has been “far surpassed” by the current capabilities of the models, which “are already intelligent” and will continue to improve.
“If I or anyone else had been asked to propose a definition of general AI five years ago based on the cognitive abilities of software, I believe the definition many would have given back then… has now been far exceeded. These models are intelligent now and will continue to become even more intelligent. They will continue to improve. I believe that, with each passing year, more people will think that we have arrived at a general AI system. Although the definition will continue to expand and become more ambitious, more people will accept it. But we already have systems that truly enhance individual productivity and are capable of performing valuable economic work. Perhaps a better question would be: what is required for something I would call superintelligence?”
Altman introduces the concept of “superintelligence” as a system “capable of making autonomous discoveries in new science or greatly increasing the ability of people using the tool to discover new science.” He emphasizes that scientific progress is essential for improving people’s lives and considers the AI’s capacity to accelerate this process to be a “tremendously exciting milestone.” In this context, the use of AI for writing code, thereby making scientists more productive, serves as a clear example of this advancement.
Altman stated that OpenAI “has not yet created any advertising product.” Although he is not “entirely opposed” to advertisements as a whole, he believes it would be “very difficult to do it properly” in ChatGPT due to the “high degree of trust” that users place in the system. The idea of modifying the output of the LLM (Large Language Model) depending on who pays more would be “truly bad” and “destructive of trust” for him. He suggests that if there were to be advertising, it would need to be “extremely useful for users” and “very clearly not interfering with the output of the LLM.”
The ambitious “Project Stargate” is an effort to “finance and build an unprecedented amount of computing capacity.” Altman underscores the “enormous gap” between what OpenAI can currently offer and what might be achieved with ten or even one hundred times more computing power. Stargate aims to “bring together a large amount of capital, technology, and operational expertise to build the infrastructure for delivering the next generation of services” and to make intelligence “as abundant and affordable as possible.”
Altman confirmed that the project is “massive” and “global,” with the participation of partners such as the United Arab Emirates and other governments. Regarding Elon Musk’s accusations of attempting to derail the project, Altman regretted that Musk had “abused his governmental influence to compete unfairly,” though Altman expressed gratitude that the administration “did the right thing.”
Finally, Altman addressed OpenAI’s plans to develop hardware. While he did not disclose specific details or immediate timelines, he affirmed that it “will take time” to create something of “extremely high quality.” He explained that current hardware and software were devised for a world without AI, and that new devices will need to be “significantly more aware of their environment” and have “much more context” within the user’s life, enabling interaction quite different from simply typing and looking at a screen.
One of Altman’s most personal revelations was his reliance on ChatGPT in his newly adopted role as a “new parent.” Although he acknowledges that people have raised children without AI for a long time, Altman confesses that he does not know how he would have managed without the constant assistance of ChatGPT during the first weeks of parenthood. He now uses it more for questions concerning developmental stages, highlighting how AI has become a valuable resource for daily life.
Altman, who is in favor of everyone having “a lot of children” (literally), is optimistic about how AI will shape future generations. He believes today’s children will be exceptionally adept at using AI, without feeling “troubled by no longer being more intelligent than AI.” Citing an example of a child interacting with a magazine as though it were a broken iPad, Altman suggests that new generations will simply assume the presence of an “extremely intelligent” AI in the world.
Image: Flux Schnell
Your email address will not be published. Required fields are marked *
Δ