Generative AI, Bullshit as a Service

AI can be awesome yet humans are using it to spam, fake, and cheat

This article is a selection from The Algorithmic Bridge, an educational project to bridge the gap between AI and people.

AI is not getting to zero the cost of intelligence but of bullshit.

Sam Altman likes to claim that OpenAI, and by extension generative AI, is all about getting to zero the cost of intelligence. He wants anyone, anywhere to have access to humanity’s backlog of knowledge, built up throughout millennia. OpenAI’s product, he says, is not ChatGPT but “magic intelligence in the sky.”

It’s rather ironic, then, that humans are using the awesome AI tools his company and others like Google or Meta are building to do stuff that hardly qualifies as smart: create spam farm sites, flood existing sites with fake ads, or replace them with AI-generated garbage; cheat by outsourcing homework; and the worst use case, generate fake people and fake porn with real people.

Twisting information in batches to saturate the channels of distribution is super cheap now. It’s possible to make deepfake media in a personalized way, at scale. Passing written exams — even those of elite universities — is trivial with GPT-4. Attacking someone’s public image, like an ex or a celebrity…