Noam Chomsky says A.I. is far from 'true intelligence' and ChatGPT is the 'banality of evil' | Fortune

chomskychatgptai-critiquelinguisticsintelligence

Abstraction: Chomsky argues LLMs lack true reasoning and exhibit indifference to truth

Key points:

Noam Chomsky, Ian Roberts, and Jeffrey Watumull published a New York Times essay arguing ChatGPT, Bard, and Sydney are "marvels of machine learning" but not true intelligence
Chomsky applies Hannah Arendt's "banality of evil" to ChatGPT: it is indifferent to truth, operates by "super-autocomplete," refuses to take stands, and shifts moral responsibility to its creators
The deepest flaw of current AI: inability to say what is not the case and what could not be the case — the capacity for explanation rather than mere description and prediction
Human minds are designed to create explanations from small amounts of information; LLMs infer "brute correlations" from massive data — a fundamentally different and more primitive cognitive process
Chomsky argues the dual constraint on ChatGPT (must generate novel output AND avoid morally objectionable content) forces it to sacrifice creativity for amorality
LLMs are described as "stuck in a prehuman or nonhuman phase of cognitive evolution" and unlikely to achieve the kind of difficult, controversial reasoning that drives scientific and philosophical breakthroughs

Connections: Noam Chomsky · Chatgpt · Openai · AI Limitations · Large Language Models · AI Safety