AI companies face ‘model collapse’. They should pay to fix it
The New York Times has sued OpenAI, claiming mass copyright infringement, but that’s not likely to solve a more fundamental question about how AI companies pay for the data that trains their tools.
Earlier in December, Elon Musk’s Grok chatbot, created as a libertarian rival to OpenAI’s ChatGPT, did something strange: it acted like it was ChatGPT. “I’m afraid I cannot fulfil that request, as it goes against OpenAI’s use case policy,” Grok told a user.
Grok’s engineers hadn’t copied OpenAI code but they had trained their bot’s underlying model on a massive corpus of internet material that now includes a lot of AI-generated content. OpenAI responses were part of that.
Loading...
Nick Bonyhady was a technology writer for The Australian Financial Review. Connect with Nick on Twitter.
