In a new essay released Tuesday called “The Gentle Singularity,” Openai CEO Sam Altman shared his latest vision of how AI will change the human experience over the next 15 years.
The essay is a classic example of Altman’s futurism. He hyped the promises of AGI, claiming that his company is very close to the feat, while simultaneously disregarding the arrival. Openai CEO frequently publishes essays of this nature, neatly laying out the future where AGI disrupts our modern work, energy and social contracts. But in many cases, Altman’s essays contain hints on what Openai is working on next.
At one point in the essay, Altman argued that next year, in 2026, the world will “see the arrival of (AI) systems that can grasp new insights.” While this is somewhat vague, Openai executives have recently shown that the company has focused on coming up with new and interesting ideas about the world in AI models.
When unveiling Openai’s O3 and O4-Mini AI Reasoning models in April, co-founder and president Greg Brockman said these were the first models scientists have used to generate new and useful ideas.
Altman’s blog post suggests that next year, Openai itself could bolster its efforts to develop AI that can generate new insights. Openai is not the only company that focuses on this initiative. Several Openai’s competitors have shifted their focus to training AI models that will help scientists come up with new hypotheses, and therefore have made new discoveries about the world.
In May, Google released a paper on Alphaevolve. This is an AI coding agent that the company claims to have created a new approach to complex mathematics problems. Another startup, backed by former Google CEO Eric Schmidt’s Futurehouse, claims that AI agent tools can make real scientific discoveries. In May, humanity launched a program to support scientific research.
If successful, these companies could automate critical parts of the scientific process and potentially invade large industries such as drug discovery, materials science and other fields.
This is not the first time Altman has tilted his hat on Openai’s plans on his blog. In January, Altman wrote another blog post suggesting that 2025 will be the year of agents. His company then began dropping the first three AI agents, operators, deep search and Codex.
However, getting an AI system and generating new insights can be more difficult than making it an agent. The broader scientific community is somewhat skeptical of AI’s ability to generate truly original insights.
Earlier this year, Hag Face Chief Science Officer Thomas Wolf wrote an essay claiming that modern AI systems can’t ask great questions. Kenneth Stanley, a former Openai research leader, also previously told TechCrunch that today’s AI models cannot generate new hypotheses.
Stanley is currently building a team at Lila Sciences. This is a startup that raised $200 million to create an AI-driven lab that focuses specifically on creating better hypotheses for AI models. According to Stanley, this is a difficult problem. Because it involves giving AI models a sense of creative and interesting.
It remains to be seen whether Openai will really create AI models that can truly generate new insights. Still, Altman’s essays may feature familiar ones. This is a preview of where Openai is likely heading next.