They said it in The documentary ‘The dilemma of social networks’: If you are not paying for the product, you are the product. Surely you have heard the phrase more than once in reference to the apps and services that we do not pay with money, but with another type of currency. Social networks, browsers, GPS apps … almost all They collect our information staff and navigation data to do business. Artificial intelligence apps are not different, in fact, there is a career to attract more and more users and forge alliances to get data with which to continue training their chatbots.
Improve experience. “When you share your content with us, you help our models more precise since it can better solve your specific problems,” this phrase is taken from the Chatgpt Privacy Policy And in the one of Gemini There are other similar ones. Anthropic was the only one who held the conversations with Claude privately, but Yesterday they announced that they changed their policy. Our personal information, use data and especially conversations serve to continue training and improve. We can disable this option if we want, but we have to be we who do it. By default it is activated.
Data shortage. AI needs many data to be trained, many. With the first language models all kinds of content were used, including Content protected by copyright as books or images of works of art. But the data is not infinite. At the end of 2021 there was already a talk of a shortage problem already early this year, Elon Musk said that AI had already consumed all human knowledge. This is a problem for the advance of AI and would be responsible for The rhythm has slowed down.
Solutions. IA companies have had to look for life to get new data. OpenAI transcribed a million youtube hours To train GPT-4, Google decided that it would use Any information published on the Internet to improve your AI and Musk believes that the future is in the synthetic data generated by AI itself. But there is something that can also be used to train AI and they are our conversations with it. With the first models they did not have so many users, but currently the volume of data that users generate is much greater. Only Chatgpt has 800 million usersit sounds like a booty too juicy not to use it.
Giving away. The conversations we have with ChatgPT can be useful, but it is even more useful when they are data from more specific groups. They count on Rest of World That to achieve this, IA companies are forging alliances with other companies and organizations that allow them to access data that cannot be achieved Web scraping. Openai has been associated with Shopee and offers its Plus plan for VIP users in Indonesia, Vietnam and Thailand. Google offers its free Gemini Pro plan for a year to students in India and Perplexity Pro is available free through operators such as Movistar or Airtel in India … These alliances increase their user base and provide real data of consumption of specific groups, which allows them to train their models more precisely.
The Chinese case. China is an example of how to access specific data can give advantage when developing effective solutions. Pharmaceutical research companies that use AI have access to the national health system data, which covers more than 600 million people. This gives them a competitive advantage and has made Chinese companies sign Millionaire agreements with large pharmaceuticals.
Worry. Experts such as Sameer Patil, director of the observer Research Foundation, calls to establish a clearer regulation, especially in sensitive sectors such as health or finance: “participating companies will have to ensure that data sets are not personalized and anonymous,” he says in statements to Rest of World.
Image | Chatgpt
GIPHY App Key not set. Please check settings