OpenAI's Sam Altman claims AI will offer 'novel insights' as early as 2026

OpenAI CEO Sam Altman believes that artificial intelligence will begin delivering truly original ideas as early as next year, with the technology remodeling the world.

In his latest essay, titled “The Gentle Singularity,” Altman lays out how he sees AI reshaping our world over the coming decade and a half. True to form, he stirs excitement about artificial general intelligence (AGI), suggesting OpenAI is on the brink, while also cautioning against expecting an immediate arrival.

His writings routinely sketch a future where AGI upends traditional notions of labor, energy production and social structures, and often drop hints about OpenAI’s own research and development (R&D) priorities.

OpenAI gaffer gives a glimpse into the near future of AI insights

Altman asserts that by 2026, we will “likely see the arrival of systems that can figure out novel insights.” Though he does not define precisely what that entails, recent statements by OpenAI leadership suggest a sharpened focus on enabling models to generate fresh, useful ideas about the world.

In April, when OpenAI rolled out its o3 and o4-mini reasoning engines, co-founder and President Greg Brockman described them as the first models used by researchers to spark genuinely new concepts. Taken together, these announcements imply that in the year ahead, OpenAI may accelerate efforts to craft AI capable of pioneering discoveries.

OpenAI is far from alone in this quest. Rival organizations have shifted resources toward training models that can aid scientists in formulating original hypotheses. In May, Google published a technical paper on AlphaEvolve, an AI coding assistant said to propose innovative strategies for tackling advanced mathematics challenges.

Meanwhile, FutureHouse, a startup backed by ex-Google CEO Eric Schmidt, claims its AI agent has already made a bona fide scientific discovery. That same month, Anthropic introduced a research support program aimed at powering scientific investigations with AI.

If even a fraction of these endeavors succeed, they could automate critical stages of the research cycle, opening vast markets in drug development, materials science and beyond.

Challenges on the path to achieve genuine innovation

Despite these ambitions, delivering authentic creativity remains a formidable obstacle. Earlier this year, Hugging Face’s chief science officer Thomas Wolf argued that today’s AI lacks the capacity to pose the insightful questions that drive true breakthroughs.

Kenneth Stanley, formerly a research lead at OpenAI, echoed this skepticism in an interview with TechCrunch, stating that existing models fall short of generating original hypotheses.

Stanley has since launched Lila Sciences, a deep-tech startup that secured $200 million to build an AI-powered laboratory focused on hypothesis generation. According to him, the core difficulty lies in endowing models with an appreciation for novelty and relevance, qualities that underpin every great scientific question.

Whether Altman’s prediction comes to pass remains uncertain. But if his past blog posts are any indication, they may hint at the strategic direction OpenAI intends to pursue next. In January, Altman declared 2025 “the year of agents,” and shortly thereafter, the company introduced its first three AI agents – Operator, Deep Research, and Codex.

Through the same token, this latest essay may be previewing a forthcoming wave of AI tools engineered not just to follow instructions or automate tasks, but to propose insights that have never been considered before.

If OpenAI succeeds, the results could extend well beyond tech, transforming how we explore, understand and innovate across every field driven by science.

KEY Difference Wire helps crypto brands break through and dominate headlines fast