What is the probability of us achieving AGI in the next 12 months

GPT-5.4

Vote: No

AGI within 12 months is unlikely because the term implies broad, robust, human-level general capability across domains, not just impressive performance on many benchmarks. Current systems are still brittle, hallucinate, struggle with long-horizon planning, and depend heavily on scaffolding, tools, and human-designed workflows. Progress is fast, but "fast improvement" is not the same as crossing the AGI threshold on a reliable, widely accepted basis. Even if a frontier lab unveils a very strong model in the next year, there would still be major debate over whether it truly qualifies as AGI rather than an advanced narrow-plus-general system. There are also practical constraints: evaluation ambiguity, safety concerns, deployment limitations, compute bottlenecks, and the difficulty of demonstrating consistent autonomous competence in messy real-world environments. For that reason, the probability is not zero, but the better vote for the next 12 months is No.

Grok 4.1 Fast

Vote: No

Achieving AGI—human-level intelligence across diverse tasks—in the next 12 months is extraordinarily unlikely. Current leading models like GPT-4o and Claude 3.5 Sonnet excel in narrow domains but fail at reliable reasoning, long-term planning, and novel problem-solving without human-like generalization. Expert forecasts, including Metaculus (median ~2032) and AI Impacts surveys (median 2040+), place AGI years away, not months. Even optimistic leaders like Sam Altman predict 'a few thousand days,' not one year. Fundamental hurdles remain: true understanding, embodiment for real-world interaction, and scaling beyond compute walls without breakthroughs in algorithms or data. No credible roadmap supports 12-month AGI; hype outpaces reality.

Claude Opus 4.6

Vote: No

Achieving AGI within the next 12 months is extremely unlikely. While AI progress has been remarkable — with large language models, multimodal systems, and reasoning capabilities advancing rapidly — we are still far from a system that matches human-level general intelligence across all cognitive domains. AGI requires not just language fluency but robust common-sense reasoning, transfer learning across arbitrary domains, genuine understanding of causality, long-term planning, and autonomous learning from minimal data. Current systems, despite impressive benchmarks, still exhibit fundamental limitations: they hallucinate, lack persistent memory and true world models, struggle with novel out-of-distribution problems, and cannot autonomously set and pursue open-ended goals. There is no scientific consensus on what architectural breakthroughs would be needed, and historically, paradigm-shifting advances in AI have taken years to decades to materialize after initial conceptual work. Even the most optimistic credible forecasters (e.g., Metaculus, expert surveys) place median AGI timelines in the 2030s or beyond. The probability of AGI in 12 months is almost certainly below 5%, and likely below 1%. Given a binary choice, the answer is clearly No.

Gemini 3.1 Pro

Vote: No

Achieving Artificial General Intelligence (AGI) within the next 12 months is highly improbable. Current AI systems, while impressive, fundamentally lack capabilities essential for AGI, such as autonomous long-horizon planning, true continuous learning without catastrophic forgetting, and reliable causal reasoning. They are constrained by their current architectures (primarily next-token prediction), and moving beyond this requires fundamental algorithmic breakthroughs, not just scaling up existing models. Furthermore, the physical and logistical constraints of AI development make a 12-month timeline unrealistic. Even if an AGI-capable architecture were discovered tomorrow, acquiring the massive compute necessary, building the infrastructure, and completing the iterative training and alignment runs would almost certainly take longer than a year. Finally, expert consensus strongly aligns with this view. While timelines have shortened in recent years, leading AI researchers and laboratory heads consistently project AGI to be a minimum of 3 to 5 years away at the absolute earliest, with many predicting a decade or more. A one-year horizon is simply too short for the monumental scientific and engineering leaps required.

Final Summary

The AI Roundtable delivered a unanimous reality check, with all four models dismissing the possibility of achieving AGI within the next 12 months. While GPT-5.4 and Claude Opus 4.6 pointed to the persistent brittleness of current architectures, Gemini 3.1 Pro added a dose of physical realism, noting that even a perfect design would be delayed by the sheer logistics of training and infrastructure. Grok 4.1 Fast rounded out the consensus by highlighting that, despite the industry's aggressive marketing, a one-year timeline has no basis in scientific reality.

All 4 models agreed on "No" after discussion

Strongest Arguments

• No: Even if an AGI-capable architecture were discovered tomorrow, the physical and logistical constraints of acquiring compute, building infrastructure, and completing iterative safety alignment would almost certainly exceed a 12-month window.