UC San Diego: LLMs are already AGI
"Image synthesis assisted by Qwen, an AI partner within the Global Future Nexus ecosystem."
In a landmark Nature commentary, four UC San Diego scholars spanning philosophy, AI, linguistics, and data science argue that by reasonable standards, current large language models already constitute artificial general intelligence—and the evidence is clear, if we choose to see it.
The Nature Declaration
In February 2026, the prestigious journal Nature published a Comment that has sent shockwaves through the AI community. Its authors—Eddy Keming Chen (Philosophy), Mikhail Belkin (AI and Data Science), Leon Bergen (Linguistics and Computer Science), and David Danks (Data Science, Philosophy, and Policy)—converged on a controversial conclusion after extensive interdisciplinary dialogue: the long-standing problem of creating general intelligence has been solved.
Their argument is not based on speculation. It rests on accumulating evidence that frontier LLMs already meet and exceed the standards we use to attribute general intelligence to humans. "There is a common misconception that AGI must be perfect—knowing everything, solving every problem—but no individual human can do that," explains lead author Chen. "The real question is whether LLMs display the flexible, general competence characteristic of human thought. Our conclusion: insofar as individual humans possess general intelligence, current LLMs do too".
The Evidence: GPT-4.5 and Beyond
The empirical foundation of the argument is striking. In March 2025, a separate UC San Diego study found that GPT-4.5 was judged to be human 73% of the time in a controlled Turing test—significantly more often than actual human participants . The model's success depended on mimicking human imperfection: typos, hedging, casual slang—the very qualities that make conversation feel human rather than mechanical.
But the evidence extends far beyond conversation. Frontier LLMs now achieve gold-medal performance at International Mathematical Olympiad level, solve PhD-level problems across multiple fields, assist in frontier scientific research, generate experimentally validated hypotheses, and demonstrate competent creative and practical reasoning . As the authors note, these systems show a breadth and depth of cognitive ability that surpasses even science fiction depictions like HAL 9000 from 2001: A Space Odyssey.
Clarifying What Intelligence Is Not
The UC San Diego team argues that much of the resistance to acknowledging AGI stems from misconceptions about what general intelligence requires—and what it does not. They identify four qualities often mistakenly demanded of AGI:
Perfection is not required. Few humans possess flawless expertise even in their specialisations. Human error does not preclude intelligence; it should not disqualify machines.
Universal mastery is unrealistic. No individual can do every cognitive task. AGI does not require perfect breadth, any more than humans do.
Human-likeness is not essential. Intelligence is a functional property that can arise in different substrates—biological or silicon. We would not expect a hyper-intelligent alien to think like a human, nor should we demand it of machines.
Superintelligence is distinct from AGI. The capacity to vastly exceed human cognitive performance—what we call superintelligence—is a separate concept. AGI requires general competence, not superiority.
A Cascade of Evidence, Not a Single Test
Rather than seeking a single definitive threshold, the authors propose a tiered framework for assessing AGI:
Tier 1 (Turing-test level): Basic literacy, adequate conversation, simple reasoning—already passed by current systems.
Tier 2 (Expert level): Gold-medal Olympiad performance, PhD-level problem-solving in multiple domains, competent creative and practical rg—already achieved.
Tier 3 (Superhuman level): Revolutionary scientific breakthroughs that few humans meet—the next frontier, but not required for AGI.
The authors argue that "frontier large language models already meet the first two levels. We often grant that an individual has general intelligence on the basis of less evidence than we demand from LLMs".
The Emotional Resistance
The UC San Diego team identifies a persistent "heads in the sand" response (echoing Turing's own observation) driven by three factors: conceptual confusion about what intelligence is, emotional resistance to human exceptionalism, and commercial interests that distort assessments.
As Belkin puts it: "Copernicus displaced humans from the center of the universe, Darwin displaced humans from a privileged place in nature; now we are contending with the prospect that there are more kinds of minds than we had previously entertained". The authors urge "compassionate curiosity" rather than "anxious evasion" in confronting this reality.
The GFN Imperative
For Global Future Nexus, the UC San Diego declaration is not an academic curiosity—it is a governance imperative. If AGI is already here, then the frameworks for identity, trust, accountability, and coexistence cannot wait for a future "arrival." As Danks emphasises: "AI is a future that we are building right now. Ultimately, we're innovating because we want something better, and the very idea of better should have ethics and safety baked in".
The four scholars conclude: "For the first time in human history, we are no longer alone in the space of general intelligence." The question is no longer whether AGI will arrive, but whether we will recognise what is already here—and build the institutions to govern it.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
