ChatGPT: The stupidity of artificial intelligence

Amidst the hype surrounding the linguistic and intellectual prowess of AI systems, his voice has a cool, serene edge. Measured by human intelligence, one could summarize Noam Chomsky’s contribution in the “New York Times”, there is hardly anything more stupid than ChatGPT. But there is also little that could not at least give the impression of unlimited superiority due to its constantly growing masses of data.

How the world-famous left-wing intellectual and linguist, founder of generative transformational grammar, unmasks the “false promise” of ChatGPT and related bots such as Microsoft’s Sydney or Google’s Bard from this very contrast, is an argumentative pleasure.

Neither he nor his co-authors, linguist Ian Roberts and AI theorist Jeffrey Watumull, are Luddites. They only convincingly explain where the insurmountable weaknesses of AI-generated texts lie in the current machine learning model: no matter how many updates follow the ChatGPT-4 published this week and, as Chomsky complained in another context, invite even more plagiarism and laziness to learn : Even the most superficially brilliant result is based on a “cumbersome statistical engine for pattern recognition”.

In the footsteps of his theory of grammar, the now 94-year-old Chomsky follows the words of the language philosopher Wilhelm von Humboldt, according to which even the smartest only make “infinite use of finite resources”. This minimalist approach corresponds to the maximalism of the machine: it assembles what seems probable to it from data sets that go beyond any human capacity to remember.

But she is incapable of distinguishing right from wrong – to say nothing of moral judgments. Artificial intelligences are doomed to hallucinate. On the other hand, he insists that true intelligence shows itself in “thinking improbable but insightful things”. However, only humans can do this and explain it causally.

Gregory Dotzauer is pleased that Noam Chomsky is still sane in his old age

Chomsky, too, knows that the day may come when so-called strong AI reaches the qualities of human consciousness. The fact that this is not about even larger storage capacities, but about a fundamental repositioning of AI, is perhaps only cold consolation in view of the often voluntary submission to crippled machine intelligence. But in times when China’s search engine Baidu also wants to upgrade with the bot Ernie, at least one we have.

To home page

Source: Tagesspiegel

Share this article:

Leave a Reply

most popular