OpenAI is set to unveil an innovative feature called Pulse within ChatGPT, designed to generate personalized reports for users while they sleep.
This new addition promises to provide users with five to ten succinct briefs, helping them stay informed and up to speed first thing in the morning, akin to checking social media or news apps.
Pulse marks a significant shift in OpenAI’s consumer products, moving towards a more proactive approach.
Rather than merely responding to user inquiries, features like ChatGPT Agent and Codex are being developed to provide a supportive assistant experience tailored to users’ needs.
Fidji Simo, OpenAI’s new CEO of Applications, noted in a blog post, “We’re building AI that lets us take the level of support that only the wealthiest have been able to afford and make it available to everyone over time.
And ChatGPT Pulse is the first step in that direction — starting with Pro users today, but with the goal of rolling out this intelligence to all.”
Pulse will be available initially to subscribers of OpenAI’s $200-a-month Pro plan, appearing as a new tab within the ChatGPT app starting Thursday.
The company envisions expanding access to Pulse for all ChatGPT users in the future, with Plus subscribers expected to gain access soon, although there are plans to optimize the product first.
The reports generated by Pulse can include summaries of news articles on topics of interest, such as updates on favorite sports teams, along with personalized briefs tailored to individual contexts.
In a demonstration for TechCrunch, OpenAI product lead Adam Fry showcased several of Pulse’s outputs: news roundups for British soccer team Arsenal, Halloween costume ideas for his family, and a travel itinerary for a family trip to Sedona, Arizona designed to be suitable for toddlers.
These reports are presented as “cards” which feature both AI-generated images and text.
Users have the option to click on any card to view the complete report and can engage with ChatGPT about the contents.
Pulse is designed to proactively generate some reports, while also allowing users to solicit new automated reports or provide feedback on existing ones.
A distinct feature of Pulse is its design choice to conclude report generation with a message stating, “Great, that’s it for today.”
This is intended to differentiate the service from engagement-driven social media platforms, promoting a more streamlined user experience.
Furthermore, Pulse is compatible with ChatGPT’s Connectors, enabling users to integrate applications like Google Calendar and Gmail.
Once linked, Pulse can review emails overnight to highlight essential messages for users in the morning, and can reference the calendar to create upcoming event agendas.
If users have enabled ChatGPT’s memory features, Pulse can also draw from previous chats to enhance report personalization.
Christina Wadsworth Kaplan, OpenAI’s personalization lead, illustrated how Pulse utilized her passion for running to construct a London itinerary that included local running routes.
Describing Pulse as a “net-new functionality” for consumer products, she shared her positive experiences, noting how Pulse identifies dining options that comply with her pescatarian diet and makes dinner reservations on her calendar.
Despite Pulse’s innovative features, there are concerns about how it may engage with existing news platforms such as Apple News, traditional journalism entities, and paid newsletters.
Fry has emphasized that Pulse does not aim to replace these various news applications and will include sourced citations with links similar to those seen in ChatGPT Search.
The efficiency of Pulse remains a key topic, particularly regarding the computational resources it demands.
Fry mentioned that the service’s computational needs can fluctuate significantly based on the specific tasks involved.
Some tasks are completed efficiently, while others may require extensive web searches and document synthesis.
In the longer term, OpenAI aspires to enhance Pulse’s capabilities, eventually leading it to make restaurant reservations or draft emails for user approval.
However, achieving this level of agentic functionality could take time, as it would necessitate considerable improvements in OpenAI’s agentic models to build user trust in such automated decisions.
image source from：techcrunch
