Large Language Model = Other People's Content

Who’s down with LLM? Everybody (these days)! But let’s not call it AI; it’s just Other People’s Content. This is even more evident from a monetization perspective: WSJ just published “New Sites are Getting Crushed by Google’s New AI Tools”. This is nothing new, Google has also been stealing other people’s content (and hence traffic) to monetize for itself, from embedding Yelp Reviews to Wikipedia articles directly on its SERP. Now with LLM, Google (and OpenAI and others) are calling reassembled statistically clustered documents “AI” - with the guise that they are producing “original AI generated” answers and hence do not have to pay for it. It took DoJ 20 years to rule Google Search a monopoly and anti-competitive; expect DoJ to take another 20 years to rule Gemini AI (and OpenAI, Meta AI, Anthropic, etc.) results anti-trust as well. By then it will be too late, all its competitors will be gone, and landscape will have shifted again - but it will be just a variation of the same theme.

UPDATE: One day after I write this, Disney and NBCU sue Midjourney over copyright infringement. Let the Hollywood Studio lawsuits begin.

UPDATE 2: Cloudflare CEO Matthew Prince says Google's ratio of pages crawled per visitor sent to a publisher fell from 2:1 10 years ago to 18:1; for OpenAI, it's 1,500:1 - Publishers facing existential threat from AI, Cloudflare CEO says.