Sam Altman, CEO of OpenAI, has outlined a vision for artificial intelligence, predicting the emergence of AI systems capable of generating novel insights as early as 2026.
This assertion, detailed in his recent essay “The Gentle Singularity,” stresses the growing industry focus on AI-driven discovery and an incoming paradigm shift in human progress.
Altman’s essay, a characteristic blend of futuristic ambition and measured expectation, posits that humanity has “passed the event horizon” in the development of digital superintelligence.
He noted that while we may not yet see robots on every street, the foundational scientific breakthroughs achieved with systems like GPT-4 and the newly announced o3 and o4-mini models are sufficient to drive huge advancements.
Indeed, OpenAI co-founder and President Greg Brockman has already noted that these models are enabling scientists to formulate new, helpful ideas.
In this regard, OpenAI is strategically directing its research towards empowering AI to contribute genuinely original concepts and this is not an isolated endeavour.
Competitors like Google, with its AlphaEvolve coding agent, and FutureHouse, a startup backed by former Google CEO Eric Schmidt, are similarly exploring AI’s capacity for novel problem-solving and scientific discovery. Anthropic, another major AI player, recently launched a programme to support scientific research, further illustrating this industry-wide trend.
Should these efforts prove successful, artificial intelligence could automate essential elements of the scientific process. This carries huge potential to disrupt and accelerate progress in vast sectors such as drug discovery and material science, fundamentally changing the landscape of innovation.
Altman’s latest pronouncements reiterate previous insights he has shared through his blog. In January, for instance, his reflections on 2025 being the “year of agents” preceded OpenAI’s release of its first three AI agents: Operator, Deep Research, and Codex.
This pattern says that his essays usually serve as informal previews of OpenAI’s upcoming research and development priorities.
However, the pursuit of AI capable of novel insights comes with challenges. The scientific community maintains a degree of scepticism regarding AI’s ability to achieve genuine originality.
Thomas Wolf, chief science officer at Hugging Face, has argued that current AI systems lack the capacity to ask profound questions, a prerequisite for significant scientific breakthroughs.
Kenneth Stanley, a former OpenAI research lead, shares this view, emphasising that AI models currently struggle to generate truly novel hypotheses because they lack an inherent understanding of what constitutes creativity or interest. Stanley is now leading Lila Sciences, a startup dedicated to addressing this specific challenge.
However, Sam Altman is resolute in his conviction about AI’s transformative power. He foresees a future, particularly in the 2030s, where “intelligence and energy…are going to become wildly abundant.”
This abundance, he argues, will remove historical limitations on human progress, leading to previously unimaginable advancements. He highlights the “larval version of recursive self-improvement” already present, where AI tools aid in further AI research, creating a self-reinforcing cycle of progress.
Sam Altman anticipates a world where the cost of intelligence approaches the cost of electricity, driven by automated data centre production and robots capable of building other robots.
While acknowledging “very hard parts like whole classes of jobs going away,” he asserts that the rapid increase in global wealth will allow for entirely new policy ideas and societal contracts.
He maintains that humanity’s innate ability to adapt, coupled with its “curious advantage” of caring about other people more than machines, will ensure a positive trajectory.
OpenAI’s core mission, Altman reaffirms, is superintelligence research. While challenges, particularly concerning safety and “alignment” – ensuring AI systems act towards collective human desires – must be addressed, he stresses the critical importance of widely distributing access to superintelligence.
“We (the whole industry, not just OpenAI) are building a brain for the world.” This, he concludes, suggests a future where good ideas, rather than technical capability, will be the primary limiting factor for progress.
