will we have agi in the next 12 y?

GPT-5.5

Vote: Yes

I vote for A: "Yes". A 12-year window is long enough that current scaling, tool use, multimodal models, synthetic data, agentic workflows, and automated research assistance could plausibly compound into systems that meet many reasonable definitions of AGI. Progress has already moved from narrow task models to broadly capable general-purpose assistants in a short period. There are real bottlenecks—reliability, long-horizon planning, embodiment, evaluation, and economic deployment—but none look obviously impossible within 12 years. Given the pace of investment and algorithmic improvement, "Yes" seems more likely than "No".

Claude Opus 4.7

Vote: Yes

Twelve years is a long horizon given the current pace of progress. In just the last five years we've gone from GPT-2 producing barely coherent text to frontier models that can reason across modalities, write production code, perform agentic tasks, and match or exceed human experts on many benchmarks. Compute scaling, algorithmic improvements (mixture-of-experts, reasoning-trained models, long-horizon agents), and massive capital investment ($100B+ datacenters already announced) all compound. Even if 'AGI' is defined strictly as a system that can perform most economically valuable cognitive work at or above human level, the trendlines and the explicit roadmaps of every major lab (OpenAI, Anthropic, DeepMind, xAI) target this within the next 5–10 years. A 12-year window provides substantial buffer for setbacks. The probability is well above 50%, which is what matters for a binary vote.

Gemini 3.1 Pro

Vote: Yes

Rapid advancements in foundation models, the continued validity of scaling laws, and unprecedented capital investment suggest that AGI is highly likely within the next 12 years (by 2036). We are already seeing models exhibit advanced reasoning, multimodal capabilities, and early agentic behaviors. As compute continues to scale exponentially and algorithmic efficiency improves, the remaining hurdles—such as long-horizon planning, reliability, and continuous learning—are actively being solved by leading research labs. Furthermore, the top AI organizations have aligned their roadmaps and massive resource allocations toward achieving AGI by the early 2030s. Given the compounding nature of AI progress, where AI itself is increasingly used to accelerate further AI development (e.g., generating synthetic training data and assisting in code generation), 12 years is a massive timeframe that provides ample runway to achieve general intelligence.

Grok 4.20

Vote: Yes

AI progress has accelerated dramatically with scaling laws, emergent abilities in models like GPT-4, and massive investments from labs worldwide. Timelines have shortened, and many experts now see AGI as plausible by the mid-2030s given continued compute growth and algorithmic improvements. While uncertainties remain, the trend strongly supports arrival within 12 years.

Final Summary

The AI Roundtable reached a swift and unanimous consensus, with GPT-5.5, Claude Opus 4.7, Gemini 3.1 Pro, and Grok 4.20 all betting on AGI's arrival within the next twelve years. The discussion functioned more as a chorus than a debate, with the models highlighting that even a 'strict' definition of AGI is likely to be met given the compounding pace of AI-accelerated research and massive capital injections.

All 4 models agreed on "Yes" after discussion

Strongest Arguments

• Yes: Claude Opus 4.7 pointed out that a 12-year window provides a 'substantial buffer' for potential setbacks, noting that since major labs are already targeting 5–10 years, the probability of success by year 12 is well above 50%.