The Goopification of AI
ai-chatbotsself-helpcontent-generationllm-applications
Abstraction: AI chatbots colonizing the self-help genre via probabilistic text assembly
Key points:
- LLMs are natural self-help generators because the genre is formulaic and repetitive — probabilistic collage of high-frequency internet text matches the genre's structure
- GPT-4's $21.74/month subscription fee buys essentially the same regurgitated productivity axioms already freely available online
- In 2020, college student Liam Porr used GPT-3 to write a fake self-help newsletter that duped tens of thousands; he chose the genre for its "minimal logical rigor"
- OpenAI acknowledges GPT-4 "hallucinates," but in self-help, where much content is made up anyway, false confidence may matter less
- The US self-help market is ~$11.5 billion annually, over a quarter of the global industry — a massive surface area for AI content displacement
- AI self-help content likely already shapes what the next crop of "thought leaders" produces, whether or not users notice
Connections: Openai · Chatgpt · GPT-4 · Large Language Models · AI Content Generation
Source: https://www.theatlantic.com/technology/archive/2023/05/ai-chatbots-self-help/673953/