OpenAI is latest “strategic partnership” The licensing agreement with the Financial Times (FT) follows similar agreements between the US technology company and publishers equivalent to Associated Press, German media giant Axel Springer and French newspaper Le Monde.
OpenAI will license the FT's content to make use of as training data for its products, including successors to its AI chatbot ChatGPT. The AI systems developed by OpenAI are exposed to this data to enhance their performance by way of language use, context and accuracy. The FT receives one undisclosed payment as a part of the deal.
This comes amid global legal challenges from media firms alleging copyright infringement in the usage of their content to coach AI products. The most famous of those is a New York Times case against OpenAI. There can also be a fear amongst technology firms that as they develop more advanced products, the Internet will do the identical now not have enough high-quality data to coach these AI tools.
So what’s going to this deal mean for the Financial Times? There continues to be an absence of detail on partnerships like this, aside from the indisputable fact that the FT is paid for its content. However, there may be evidence of other potential advantages.
In an announcement, FT Group chief executive John Ridding stressed that the newspaper was committed to “human journalism”. But he also acknowledged that the news business cannot stand still: “We are keen to explore the sensible outcomes by way of news sources and AI through this partnership… We value the chance to be involved in the event cycle as people engage content in latest ways discover.” .”
The FT has said previously It would “responsibly” experiment with AI tools and train journalists to make use of generative AI for “story discovery.”
OpenAI is probably going keen to announce this partnership since it hopes it is going to help solve probably the most acute problems facing its flagship products. The first is that these generative AI tools sometimes invent things, a Phenomenon generally known as hallucination. Using reliable content from the FT and other trusted sources should help.
The second problem is that it could help offset the legal scrutiny OpenAI faces. Signing official contracts with news sources offers the tech company some control over its reputational damage because it shows it’s attempting to make amends on the earth of journalism. It may additionally offer greater legal certainty in the long run.
The FT's licensed content – and other media sources – could provide ChatGPT and the upcoming GPT-5 with more specific, referenced answers for users. Gemini, Google's ChatGPT competitor, is already attempting to do that by providing Google searches that back up the claims it makes. When OpenAI gets results directly from the source, it has more reliable evidence to look and train.
This appears to be following the trend “On-demand prolonged generation” (RAG), which is becoming increasingly popular within the AI world. RAG is a method where a big language model (the technology behind AI chatbots like ChatGPT) may be equipped with a knowledge base that may be searched to support what the chatbot already knows. It's a bit like taking an exam with a textbook open in front of you.
This helps reduce the chance of a hallucination, during which the AI authoritatively produces a response that appears real but is definitely made up. Access to a database of trustworthy journalism helps offset the issue Reliability issues with AI products because they were trained on the open web.
Partnership program
There is a subtext to this global media partnership program that will not be about law or ethics. OpenAI needs an increasing number of data over time to proceed making big improvements through upgrades to its AI products. But these products are running out of high-quality training data from the open Internet.
This is at the very least partly because there may be one now Distributing content created by AI on the Internet. This potentially undermines OpenAI's ongoing have to prove to its partners, governments and investors that it may deliver major improvements to its flagship products.
The New York Times lawsuit claims that products like ChatGPT threaten the business of media firms. Whatever the consequence of this case, it’s in OpenAI's interest to maintain its sources of coaching data, including media firms, productive and commercially viable. ChatGPT's success, at the very least for now, depends heavily on the success of the people and organizations that produce the info that make it useful.
PR from the AI industry has done much to advertise the thought of inevitability: that AI, in the shape of products like ChatGPT, will transform industries – and other people's lives typically. But the technology keeps failing. The FT deal highlights the dynamic tension that exists between AI and the industries it’s transforming. ChatGPT now needs the trustworthy journalism that its own generative capabilities and training methods have undermined.
The concept that Generative AI has poisoned the web is nothing latest. Some AI researchers have compared the proliferation of AI-generated trash online to how radioactive contamination of metals forced steelmakers within the Nineteen Fifties to scavenge for steel from wrecked ships manufactured before the atomic age. This prenuclear steel was needed for certain applications, equivalent to particle accelerators and Geiger counters.
Similarly, it doesn’t seem like a viable path for OpenAI and similar firms to coach their products on data “waste.”