"This is not intelligence": a new, more honest term proposed for AI
Renowned mathematician Terence Tao has proposed abandoning the term «artificial general intelligence» (AGI) in favor of what he calls a more honest term: «artificial general cleverness» (AGC).
According to Tao, what we call «AI» today is better described as the ability to solve complex problems using partially heuristic methods.
These solutions may rely on brute force, randomness, massive computational power, or training data, but do not necessarily indicate the presence of «intelligence» in the human sense.
Post by @tao@mathstodon.xyzView on Mastodon
He notes that this kind of «cleverness» can be highly effective—especially when results are rigorously verified and erroneous answers are filtered out. However, this does not make AI truly intelligent.
«This leads to a somewhat paradoxical situation: the technology can be simultaneously very useful and impressive—yet fundamentally unsatisfactory and disappointing,» Tao wrote.
The mathematician emphasizes that in humans, intelligence and cleverness are closely connected, whereas in AI they are effectively separated. Machines can demonstrate outstanding results on specific tasks without possessing understanding, intentions, or general reasoning abilities.
Tao does not take an anti-AI stance. He has repeatedly said that AI-based tools have significantly accelerated his own scientific work. His criticism is directed at inflated expectations and the imprecise language used to describe current technologies.
In Tao’s view, replacing «AGI» with «AGC» would help provide a more sober assessment of AI’s current state and prevent conflating computational efficiency with genuine intelligence.
Читать на dev.by