Every invention in history — fire, the wheel, the internet — was created by the human mind. AGI, Artificial General Intelligence, could be the last invention we ever need to make. Because after it, the inventor will be able to invent everything on its own.
What Is AGI?
The artificial intelligence we know today — ChatGPT, Siri, Alexa — is “narrow” (Narrow AI): exceptional at one task, but helpless beyond its design. AGI (Artificial General Intelligence) is fundamentally different: an intelligence that equals or surpasses humans across every cognitive domain. It can think, learn, plan, create — and transfer knowledge between fields without reprogramming.
DeepMind's 5 Levels of AGI
In November 2023, Google DeepMind researchers proposed a framework classifying AGI into 5 performance levels:
🌱 Emerging
Equal to an unskilled human. Today's LLMs like ChatGPT and LLaMA are at this level.
📊 Competent
Outperforms 50% of skilled adults across a wide range of non-physical tasks.
🎯 Expert
Outperforms 90% of experts across many fields — medicine, law, engineering.
🏆 Virtuoso
Outperforms 99% of humans. Only the world's top specialists can compete.
⚡ Superhuman
Outperforms 100% of humans in every domain. This is Artificial Superintelligence (ASI).
How Close Are We to AGI?
Progress in recent years has been staggering. In March 2023, Microsoft researchers stated that GPT-4 could be viewed as an “early but incomplete version of AGI.” GPT-4 outperformed 99% of humans on the Torrance Tests of Creative Thinking. In March 2025, a study showed GPT-4.5 passed the Turing Test — judged as “human” 73% of the time, surpassing even real humans (67%).
Sam Altman Declares: “We Built AGIs”
In December 2025, OpenAI CEO Sam Altman conceded that “we built AGIs” and that “AGI kinda went whooshing by” — proposing the world now needs to define what “superintelligence” means.
Tests for Human-Level AGI
The Benefits
🏥 Medicine
Faster diagnosis, personalized treatment, record-time drug discovery. AGI could simulate molecular interactions and identify treatments for cancer and Alzheimer's.
🔬 Science
Solving unsolved problems in physics, mathematics, biology. AGI could understand dark matter, quantum systems, and prove mathematical theorems.
📚 Education
Personalized learning for every student. Tutoring systems that adapt in real time. Education could become free and accessible to everyone.
🌍 Climate
Emissions reduction models, energy optimization, natural disaster prediction. AGI could even assist with climate engineering.
The Risks
That same power poses the threat. In May 2023, CEOs of Google DeepMind, OpenAI, and Anthropic signed a statement: "Mitigating the risk of extinction from AI should be a global priority alongside pandemics and nuclear war."
Mass Unemployment
OpenAI researchers estimate that 80% of the US workforce could see at least 10% of their tasks affected. Geoffrey Hinton advised the UK government to adopt Universal Basic Income (UBI). Stephen Hawking warned: “If the wealth is not redistributed, most people will end up miserably poor.”
Geoffrey Hinton, the “Godfather of AI,” stated: "The idea that this stuff could actually get smarter than people — a few people believed that. But I no longer think it's far off." He estimated systems smarter than humans could appear in 5-20 years.
Where We Stand Today
The debate has shifted. We no longer ask “if” we'll reach AGI, but “when” — or “have we already?” Researchers like Blaise Agüera y Arcas and Peter Norvig write: “Artificial General Intelligence Is Already Here.” Resistance to this recognition comes from skepticism, ideological commitment to alternative theories, or human exceptionalism.
Jensen Huang (CEO of Nvidia) stated in March 2024 that within 5 years, AI would pass every human test. Leopold Aschenbrenner considers AGI by 2027 “strikingly plausible.”
"If a superior alien civilization sent us a message saying 'We'll arrive in a few decades,' would we just reply, 'OK, call us when you get here'? Probably not — but this is more or less what is happening with AI."