Overall Progress and Baseline The pace of AI mathematical capability has been extraordinary, advancing from almost zero to solving 90% of a leading research-problem benchmark and producing a wave of partial results on major conjectures in just 18 months. However, under the strict definitions of complete, expert-accepted resolution, no listed problem has yet been solved. Every Millennium-level output so far is partial, unrefereed, or withheld. With about 13 years left until the 2040 cutoff, i
Set in context with a related timeline forecast , the probability of no resolution before 2040 was slightly lowered to 10% to ensure alignment, with the difference distributed to Unforced 3D Navier-Stokes and Yang-Mills.
Current Trajectory in AI Astronomy
As of October 2026, AI has not yet produced a headline, prize-caliber discovery in astronomy or cosmology, but the building blocks are scaling rapidly. General-model agents are now routinely deployed for catalogue-scale analysis, such as vetting over 370,000 variable stars to find new anomalies [2608.23688] Agentic Active Learning Meets Visual ... [Agentic Active Learning Meets Visual Embeddings](https://arxiv.org/pdf/26
Incorporating broader timelines for physics discoveries and the severity of physical experimental bottlenecks maintained the view that data-mining and pure theoretical reasoning are the most likely paths to a prize-caliber astronomical breakthrough.
The strict bar for AI attribution. At 42%, "None by 2035" is the most likely outcome, driven by the rigid criteria for what counts as a resolution. A general-purpose AI system—not just a specialist classifier—must produce the central idea, and the astronomical community must explicitly credit it [AI uncovers the universe's 'settings' with unprecedented ...](https://www.livescience.com/space/astronomy/ai-uncovers-the-universes-settings-with-unprecedented-precision-and-it-could-help-to-resolve
The probability of no resolution by 2035 was reduced from 51% to 42% because rapid anticipated progress in theoretical reasoning and data mining makes it likelier that AI tackles a major cosmological problem .
Current State and AI's Chemistry Frontier
As of late October 2026, no AI-produced chemistry result is close to Nobel caliber, meaning this assessment is fundamentally about which AI discovery route scales fastest. Recent accomplishments include optimization, such as a 10,080-reaction campaign directed by GPT-5.4 that improved Chan–Lam yields [OpenAI and Molecule.one report a near-autonomous AI ...](https://www.rdworldonline.com/openai-and-molecule-one-report-a-near-autonomous-ai-chemist-tha
Reflecting on the pace of progress across AI-driven scientific breakthroughs shifted likelihood slightly toward an AI partnering with human labs or leaning on pure reasoning, as physical bottlenecks and strict attribution criteria may hinder other fully autonomous paths.
The most likely first resolution is better-than-expert prediction of reaction yields (32%), followed by a 10× better catalyst (16%) and chemical accuracy for strongly correlated systems (14%). The ranking is driven by two primary constraints: the speed at which a solution can be empirically verified, and the strict requirement that a general-purpose AI system produces the central idea. Evaluated alongside related timelines, these constraints continue to favor reaction yield prediction as the mos
Evaluated alongside related timelines, expectations were broadly maintained, with reaction yield predictions remaining the most probable initial breakthrough.
AI Design with Human Execution (33%) The most likely path to a Nobel-caliber materials breakthrough is an AI system designing a material or experiment that human scientists then synthesize and validate. This route leverages the world's massive existing laboratory infrastructure and replication networks without waiting for the formidable engineering challenges of end-to-end automation to be fully solved. Physical scale-up bottlenecks pose a particularly severe hurdle for fully autonomous disc
We slightly increased the probability of human-executed AI designs, recognizing that physical scale-up bottlenecks pose a particularly severe hurdle for fully autonomous discovery workflows in materials science.
The High Bar for Recognition The "Nobel-caliber" threshold fundamentally shapes this assessment. To qualify for a Nobel Prize in Physiology or Medicine, a Lasker Basic award, a Breakthrough Prize, or equivalent expert consensus, a result cannot merely be an optimization gain, a better therapeutic binder, or a cheaper production method. It must represent a conceptually new, field-changing discovery on the level of CRISPR, optogenetics, or microRNA. Because formal prize citations take years, t
Evaluating the timeline for breaking physical bottlenecks in life sciences slightly shifted probability away from pure theory or fully automated endpoints, favoring pathways where AI mines existing datasets or designs experiments for humans to execute.
"None by 2040" (40%)
The most likely outcome remains that none of these questions is answered by an AI-originated result before 2040, though the probability sits below 50% given the extreme pace of AI capability growth. To qualify, a general-purpose AI system must produce both the central idea and the main result, and biologists must come to accept it. While general systems recently scaled from 5% to 100% on FrontierMath Tier 4 within 18 months, biology operates under very different constra
The probability of "None by 2040" was slightly increased to reflect the slower, unified timelines for breakthrough consensus across the cluster, acknowledging the difficulty of resolving multi-causal biological questions.
The modal path (37%) is an AI system designing experiments or molecules that humans then execute. Current frontier biomedical AI heavily follows this "AI proposes, humans run" paradigm. For example, Robin generated therapeutic candidates for dry AMD and defined follow-up experiments that humans ran A multi-agent system for automating scientific discovery, [A multi-agent system for automating scientific discovery](https://www.nature.com/articl
Evaluating structural clinical bottlenecks and validation requirements slightly shifted probabilities toward AI workflows that rely on humans executing designed experiments or mining existing data, and away from pure theory or fully autonomous labs.
The race for the first Nobel-caliber social-science result produced by an AI system strongly favors pathways that bypass slow, real-world bottlenecks in favor of rapid iteration and fast verification. Because the resolution depends on a result being widely recognized as a leading candidate for major prizes—which themselves often lag actual discovery by years—breakthroughs that can be quickly checked and validated have a distinct advantage. Consequently, general-model reasoning and the autonomous
Factoring in broader assessments of how AI systems overcome physical world bottlenecks slightly increased the probability that the first major discovery would involve human-executed experiments, though purely digital methods remain heavily favored.
The highest hurdle for any of these outcomes is the strict resolution criteria: a general-purpose AI must originate the central idea and main result, the field must widely consider the dispute settled, and researchers must credit the AI analysis as the decisive contribution. Even decisive human work rarely produces consensus in the social sciences, where disputes are often entangled with ideological differences, subjective methodological preferences, and data access constraints. Consequently, "N
Weighed alongside related questions, the probability of settling no dispute by 2035 was slightly adjusted to 46% to maintain consistency with the expectation that Nobel-caliber AI capabilities will likely emerge in the early 2030s , though forcing widespread consensus on a long-running debate remains a uniquely difficult sociological challenge.
The current trajectory of AI research makes a qualifying discovery outside the excluded fields (math, theoretical computer science, and theoretical physics) highly probable, leaving only a 3% chance of no such result by 2035. General-purpose systems are already producing top-tier results in the excluded computational fields on a monthly basis. The primary question is which empirical or applied discipline will first overcome its unique physical validation bottlenecks, attribution challenges, or c
Ask a followup
Sign in to run · $20 free credit, no card · every claim cited