Sam Altman says AI will achieve novel insights by 2025, signaling a leap in model understanding, creativity, and real-world reasoning capabilities
OpenAI CEO Sam Altman recently published a new essay titled “The Gentle Singularity,” he shared his most recent perspective on how AI will transform the human experience over the next 15 years.
The essay is a classic example of Altman’s futurism, as it simultaneously downplays the arrival of AGI while hyping up its promise and arguing that his company is near the feat.
Essays of this nature are frequently published by the CEO of OpenAI, who concisely outlines a future in which AGI disrupts our contemporary understanding of energy, labor, and the social contract. However, Altman’s essays frequently indicate the upcoming projects OpenAI is developing.
Altman asserted in the essay that the world will “likely see the arrival of [AI] systems that can figure out novel insights” in 2026, the following year.
Although this statement is ambiguous, executives at OpenAI have recently stated that the company is currently dedicated to enabling AI models to generate novel and intriguing concepts regarding the world.
In April, Greg Brockman, co-founder and President of OpenAI, declared that the o3 and o4-mini AI reasoning models were the first models that scientists had employed to generate new, beneficial ideas.
Altman’s blog post implies that OpenAI may intensify its endeavors to create AI to produce innovative insights in the upcoming year.
OpenAI is not the sole company that is dedicating resources to this endeavor; in fact, several of OpenAI’s competitors have redirected their attention to the development of AI models that can assist scientists in formulating new hypotheses, which in turn leads to innovative discoveries about the world.
Google published a paper in May regarding AlphaEvolve, an AI coding agent that the company asserts has developed innovative solutions to intricate mathematical problems.
FutureHouse, an additional startup that has received funding from former Google CEO Eric Schmidt, asserts that its AI agent tool has the potential to generate an authentic scientific discovery. Anthropic-initiated scientific research
These companies can potentially disrupt major industries, including drug discovery, material science, and other fields fundamentally rooted in science, if they effectively automate a critical component of the scientific process.
This would not be the first instance in which Altman has disclosed his intentions regarding OpenAI in a blog. Altman published an additional blog post in January proposing that 2025 would be the year of agents. Subsequently, his organization discontinued its initial three AI agents: Codex, Deep Research, and Operator.
However, it may be more challenging to induce AI systems to produce innovative insights than to render them agentic. The scientific community continues to doubt AI’s capacity to make genuine, original insights.
Thomas Wolf, the Chief Science Officer of Hugging Face, authored an essay earlier this year in which he contended that contemporary AI systems are incapable of posing profound queries, a critical component of any significant scientific advancement.
Kenneth Stanley, a former OpenAI research director, has previously stated to TechCrunch that the AI models of today are incapable of generating novel hypotheses.
Stanley is currently in the process of establishing a team at Lila Sciences. This startup has raised $200 million to establish an AI-powered laboratory that enables AI models to generate more accurate hypotheses.
According to Stanley, this issue is challenging because it necessitates providing AI models with an understanding of what is intriguing and innovative.
It remains to be determined whether OpenAI can develop an AI model capable of generating innovative insights. However, Altman’s essay may contain a familiar element, which may foreshadow the direction in which OpenAI is likely to proceed.
