The Machines Against Infinity: Wolfgang Sperlich on AI
Scott Hamilton
One day in 2023 Noam Chomsky, the most influential linguist of all time and the most famous anarchist alive, picked up the phone and called a small village on the edge of the Aotearoan rain forest. Chomsky was calling his friend and collaborator Wolfgang Sperlich, who has lived for years in Huia, between the Waitakere Ranges and the turbulent entrance to the Manukau harbour.
Like Chomsky, Sperlich is both a linguist and an anarchist. In the seventies he was part of Wilderlands, an experimental community in the Coromandel. In the eighties he lived in Vanuatu, the world's most linguistically diverse nation. Sperlich studied previously undocumented languages, and also got a close-up view of newly independent Vanuatu's experiments with what its Prime Minister Walter Lini called 'Melanesian socialism'. He was excited by the tongues he studied, but became critical of Lini's policies.
In 2006 Sperlich published a study of Chomsky as part of Reaktion Books' Critical Lives series. His book deals with both Chomsky's politics and ideas about language, without conflating them or judging one in terms of the other. Sperlich's remains the best introduction to Chomsky.
At the same time that he was writing his book on Chomsky, Sperlich was looking hard at the ramifications of new technology for linguistics. Having worked with small Pacific languages that were threatened by what Vaughan Rapatahana has called 'the hydra of English', Sperlich was interested in whether the internet and the communities it made possible could help preserve linguistic diversity. His 2005 paper 'Will cyberforums save endangered languages? A Niuean case study' looked at OKA-KOA, an Australian-based website that included fora where Niueans could communicate in their native tongue. The piece is notable for its level-headedness. At a time when the internet was still attracting utopian expectations, both inside and outside academia, Sperlich emphasises its negative as well as positive effects.
More recently Sperlich has been thinking about Artificial Intelligence. Linguists do not seem to be playing a central role in debates raging over AI. Computer engineers, neuroscientists, philosophers, and self-appointed futurologists have all been much more prominent. But linguists should have much to offer to discussions about AI. Their training gives them a sense of the limitations of semiotics and the difference between a lexicon and a body of knowledge. They could offer an antidote to the triumphalism of the tech bros who believe that Large Language Models will inevitably become sentient.
Sperlich sent the following comment to Reef Shark after reading the post Notes on the Endangered Species of Humanity.
I found his insistence on the difference between human and AI languages fascinating, and also inspiring. He links human language to human creativity, and AI language to an inevitable lack of creativity. The common deep structure of human languages means, he suggests, that our syntax can create an 'infinite' number of expressions. It is this faith in human creativity that grounds both Sperlich's linguistic work and his radical politics.
Wolfgang Sperlich writes:
Having just read Dario Amodei's silly essay on how to mitigate AI risks, I find your evaluation to the point. Maybe a point that needs more focus is that Amodei, like a modern-day Oppenheimer, is ready to deploy AI against the Chinese Communist Party, evil incarnated, just as you portray fundmentalist Christians doing.
Only when the CCP is defeated can we look forward to the great benefits of AI like curing all diseases, and most importantly ensuring 'economic growth' that guarantees 'abundance' and prosperity of all of humanity. What abject nonsense!
Another important point you raise is that AI destroys creativity by virtue of not being able to be creative. Here I would note that Cartesian linguistics a la Chomsky is the antithesis of AI and Large Language Models in that syntax can generate an infinite amount of expressions, some of which are the creative ones that are necessary for human survival.
AI via LLM can only match or recombine existing data and as such will forever lag behind the creative human mind that is capable of producing brand new ideas and art forms.
As a sideline missive, one might suggest that your creative 'on-line' essay and my 'on-line' comments are fodder for the Large Language Model data centers, hence hopefully we are somewhat skewing the dreadful 'language' produced by the likes of Amodei, Luxon and Co.
Comments