T he suggestion that artificial intelligence (AI) has matched, or even surpassed, human intelligence conjures visions of an apocalyptic world, in which humans have been dethroned as the apex predator of Earth. This rhetoric has long been shaped by popular culture and online discourse, from Isaac Asimov’s novels to The Joe Rogan Experience, but has been reinforced by the rapid development of generative AI in recent years. Today, many people use AI as a therapist or a companion, treating it as if it were human. But do these use cases reflect human-level intelligence?
In 2026, this question, balancing across the borders of machine learning, philosophy, and cognitive science, is acutely relevant.
Large language models (LLMs) can write essays, conduct research, tutor, and speak countless languages with unbelievable coherence. AI companies, such as OpenAI and Google DeepMind, are in the next space race, with artificial general intelligence (AGI) being their explicit mission. AGI is used to describe varying degrees of intelligence but generally refers to AI that can match or exceed humans in cognitive tasks.
Whether or not this has yet been achieved is highly contested, largely because there is no single, agreed-upon level of human intelligence; it varies from person to person, generation to generation, and covers so many different domains of application that reducing it to any specific criteria is inherently challenging.
Yet, in the formal benchmarks that we do use to assess the intelligence of other humans, AI thrives, matching or surpassing human performance. These include tasks involving pattern recognition, memory, data processing, and even essay-writing. For example, OpenAI’s GPT-4 passed the bar exam, and Google’s Gemini won gold at the International Mathematical Olympiad. We label humans as intelligent when they perform much worse in tasks like these than AI, which raises the question: is it just anthropocentric narcissism that sustains this debate?
One might say that AI is limited to repeating its training data and thus cannot be compared to human intelligence. Today’s models, however, increasingly demonstrate the ability to apply prior knowledge to new problems, particularly in logical and mathematical contexts. Moreover, humans similarly learn by using what they are exposed to and applying it to new contexts. The fact that AI is trained on data does not necessarily devalue its intelligence.
Where the crucial difference lies between artificial intelligence and human intelligence is in genuine experience and understanding. As it stands, even the most advanced AIs lack consciousness and subjective awareness; identifying patterns in data cannot be equated to human qualia. Although ChatGPT can act as a therapist and seems to comprehend and relate, it is simply simulating empathy. AI does not possess what makes us essentially human.
Thus, the question of whether AI has ‘reached’ human intelligence seems to commit a categorical error: intelligence may be one word, but it is not a one-dimensional ladder with AI climbing up to reach humans. These technologies may surpass humans in some capacities, but in other ways there is a fundamental gap between our ‘intelligence.’
Experts from different academic departments across the world are debating the question. However, they all agree on one thing: with our current rates of technological progress, new and stronger forms of intelligence will appear in the future.
Image by Igor Omilaev on Unsplash