What I Read This Week... - Chamath Palihapitiya

The race to lead AI has become a hunt for new and growing amounts of training data. OpenAI, Google and Meta ignored corporate policies, altered their own rules and discussed skirting copyright law as they sought online information to train their newest AI models. In a violation of YouTube’s policies, an OpenAI team transcribed more than a million hours of videos to train its latest model GPT-4. Meta, which owns Instagram and Facebook, discussed using copyrighted data from the internet on the grounds that negotiating licenses with publishers would take too long. And last year, Google broadened its terms of service to allow the company to tap publicly available Google Docs, restaurant reviews on Google Maps and other internal information for training data. Elon Musk announced that Tesla will be unveiling its first Robotaxi on August 8th later this year. The announcement comes alongside another report that Tesla is cancelling its plans to build a more affordable $25,000 EV, instead direct

saved by

• Gaurisha Dewan