AGI: Nature Journal Confirms AI Expert-Level Intelligence Exceeds Humans

[Para 1: The Lead] In a groundbreaking publication, The Nature journal confirms that artificial general intelligence (AGI), long considered a future concept, has already surpassed human-level intelligence. Published on February 4, 2026, the study by researchers from UC San Diego and other institutions reveals that AGI, defined by its ability to perform tasks at expert levels across multiple domains, is not a distant reality but an existing phenomenon. [Para 2-3: Supporting details & Context] The study challenges the traditional Turing Test as a measure of AGI, citing that OpenAI's GPT-4.5, released in March 2025, passed the test with 73% of participants unable to distinguish it from a human. This AI demonstrated superior capabilities in natural language understanding and expression, surpassing most humans in these areas. Despite this, 76% of top AI researchers still believe AGI has not been achieved, highlighting a cognitive dissonance between technological reality and human perception. The research outlines three tiers of AGI capability: - Turing Test level (already achieved) - Expert level (current reality, demonstrated in advanced AI systems' ability to solve complex mathematical problems, assist in scientific hypothesis generation, pass PhD qualifying exams, write complex code, and engage in multilingual literature creation) - Superhuman level (approaching, characterized by the ability to make revolutionary scientific discoveries) The study also addresses common misconceptions about AI, emphasizing that true AGI is defined by breadth and depth of capability, not perfection in all areas. It challenges the notion that AI must mimic human biology and emotional processes, arguing that intelligence is a function of effective problem-solving, regardless of physical form.