On Thursday, Openai announced ChatGPT PULSE, the new Push function, which generates personalized daily updates for users, without asking every time. Pre -viewing function now available for PRO subscribers on mobile devices, Marks Openai The last attempt To make ChatGPT proactive, and not reactive, with an artificial intelligence model conducting research research to provide morning updates based on the history of the user and connected applications. Openai calls this “personalized research and timely updates that regularly appear to keep you in the know.”
ChatGPT Pulse works by analyzing the history of the user chat, saved preferences and additional connections to Gmail and Google calendar every night. The next morning, users receive visual “cards” (small illustrated squares with a resume with those that can be expanded to obtain detailed information) containing updates on topics that determine the model, such as subsequent actions, dinners or recommendations on travel. Users can provide feedback through the ratings up or down and request specific topics through the “Kurator” button.
Openai says that instead of waiting for users to initiate conversations, ChatGPT is now trying to provide information using what Openai calls, “asynchronous research”, in fact, when the model generates requests and answers at night using traditional methods. Updates appear once a day and disappear after 24 hours, if users do not save them or ask subsequent questions that convert them into standard conversations in the chat.
An example of the ChatGPT pulse interface provided by Openai.
Credit: Openai
Users can also connect Gmail and Google Calendar to provide an additional context for the fact that Openai causes more appropriate sentences. When the calendar is connected, ChatGPT can make a sample of the agenda of the meeting, remind users to buy a birthday gift or the Surface Restaurant recommendations for the upcoming trip. These integrations are turned off by default and can be switched in the settings.
But, like most of modern artificial intelligence products, based on large language models that are based on templates absorbed with the help of educational data sets, there is a catch: its success varies greatly on the basis of a common topic. Openai says that he checked the new Pulse function with college students as part of his ChatGPT laboratory program, finding that users “began to feel their utility as soon as they began to tell Chatgpt that they wanted to see.”
