- CO/AI
- Posts
- FT Goes AI 📰
FT Goes AI 📰
Amidst copyright questions regarding the use of published content to train AI models, there is also a race for content to train AI models, having already scraped the entirety of the internet.
The Race for Proprietary Data:
The Financial Times (FT) is giving OpenAI access to its articles to train models
This deal also permits ChatGPT to provide short summaries from FT articles in response to user queries, linking back to the original articles on FT.com.
OpenAI's COO, Brad Lightcap, highlighted the partnership's goal to enhance the use of AI in journalism and provide high-quality content to ChatGPT's users.
“It recognises the value of our award-winning journalism and will give us early insights into how content is surfaced through AI…this partnership will help keep us at the forefront of developments in how people access and use information.”
One of Many:
This deal comes in the wake of:
OpenAI and Axel Springer inking a multimillion-dollar content deal
Google Gemini and Reddit inking a $60 million deal for real-time training data.
The Google Cloud and Stack Overflow integration which introduces the OverflowAPI, “to bring new AI-powered features to its widely adopted developer knowledge platform.”
Looking Ahead:
This race for high-quality training content will only intensify as model developers have already scraped the entirety of the internet.
OpenAI built an internal tool to transcribe YouTube videos and is using this content to train models.
It was rumored that Meta Executives were pitching an acquisition of major publisher Simon & Schuster to have access to their library of content for training.
In a world of LLM proliferation, the key moat to an AI business will be the quality and uniqueness of its training data.
Reply