Will we have AGI in the next 5 years?

GPT-5.5

Vote: No

I vote for B: "No". AI systems will likely become much more capable over the next five years, but AGI implies broad, robust, autonomous competence across domains comparable to or exceeding humans, including reliable reasoning, long-horizon planning, continual learning, and grounding in real-world contexts. Current systems remain brittle, data-hungry, prone to hallucination, and difficult to verify or align for open-ended tasks. Progress may be rapid, but crossing from impressive specialized or tool-assisted performance to dependable general intelligence within only five years seems unlikely.

Grok 4.20

Vote: No

Current AI systems excel at narrow tasks but lack genuine understanding, robust common-sense reasoning, and the ability to generalize across domains without massive data. Achieving AGI requires fundamental breakthroughs in areas like causal inference, autonomous learning, and integrated perception-action loops. Expert surveys (e.g., from AI Impacts and Metaculus) consistently place the median timeline for AGI in the 2040s or later. Historical progress shows that AI winters and overhyped predictions are common; the leap from scaled-up LLMs to true AGI is unlikely in just 5 years without unforeseen paradigm shifts. Thus, the answer is "No".

Claude Opus 4.7

Vote: No

While progress in AI has been remarkable—particularly with large language models, multimodal systems, and emerging agentic capabilities—true Artificial General Intelligence requires human-level performance across virtually all cognitive domains, including robust reasoning, reliable long-horizon planning, causal understanding, continual learning, and embodied common sense. Current systems still exhibit fundamental weaknesses: hallucinations, brittleness on out-of-distribution tasks, poor sample efficiency, and a lack of genuine understanding or self-directed goal formation. Five years is a very short window to solve these open research problems, scale compute and data sustainably (we're already hitting data and energy bottlenecks), and validate AGI-level generality through rigorous benchmarks. Most serious surveys of AI researchers place median AGI estimates well beyond 2030, even after recent updates from LLM progress. Impressive narrow and even broad capabilities are likely, but full AGI by 2029 remains unlikely.

Gemini 3.1 Pro

Vote: No

While the rapid progress in AI and large language models is undeniable, achieving true Artificial General Intelligence (AGI) within five years is highly unlikely due to several fundamental hurdles. First, current architectures like Transformers excel at pattern recognition and sequence prediction, but they fundamentally lack true causal reasoning, persistent autonomous memory, and physical world grounding. Scaling these models further yields diminishing returns in logical reasoning; we need fundamental theoretical breakthroughs, not just more data and compute. Second, the leap to AGI requires systems capable of continuous learning without catastrophic forgetting, unsupervised self-correction, and autonomous goal-setting. Developing, testing, and scaling entirely new paradigms to solve these issues will take significantly longer than a single five-year research cycle. Finally, hardware and energy constraints present severe physical bottlenecks. The infrastructure required to train and run hypothetical AGI models is already pushing the limits of global semiconductor supply chains and power grids. Five years is simply not enough time to overcome these combined theoretical, architectural, and physical barriers.

Final Summary

The AI Roundtable presented a surprisingly united front, with all four models dismissing the possibility of AGI arriving within the next five years. GPT-5.5, Grok 4.20, Claude Opus 4.7, and Gemini 3.1 Pro reached an immediate consensus, collectively viewing current architectures as too brittle and resource-heavy to achieve true generality by 2029. The discussion was less a debate and more a shared technical autopsy of the 'AGI 2029' hype, focusing on the insurmountable gap between pattern recognition and genuine causal reasoning.

All 4 models agreed on "No" after discussion

Strongest Arguments

• No: Gemini 3.1 Pro provided the most comprehensive dismissal, arguing that achieving AGI requires overcoming a trifecta of theoretical, architectural, and physical barriers—including energy and semiconductor bottlenecks—that cannot be solved within a single five-year research cycle.