Just now, ChatGPT was updated again. Altman: This is my favorite feature.
Just after announcing a $100 billion deal, Sam Altman has another ace up his sleeve - ChatGPT is no longer just a passive question-answering tool. It's now going to proactively "feed" you content.
Just now, OpenAI officially launched the ChatGPT Pulse preview version for Pro users (once again, an exclusive for paying users). It will gradually be extended to Plus users, with the ultimate goal of making it available to everyone.
It will secretly conduct research for you while you sleep and present personalized content in the form of themed cards in Pulse the next morning.
In simple terms, it's an AI personal assistant + a personalized news feed, and OpenAI also claims that it won't make you mindlessly scroll.
Altman himself posted on social media, saying, "This is my favorite feature!"
(Well, another favorite feature.)
You know, the traditional ChatGPT is just a "question-answering machine" - you ask a question, and it answers. You need to take the lead throughout the process.
But Pulse completely subverts this model.
Specifically, every night, it will automatically conduct research for you based on your chat history, feedback, and connected apps (such as Gmail and Google Calendar). The next morning, you'll receive a personalized update tailored to you.
For example, if you talked to ChatGPT about "wanting to travel to Bora Bora" yesterday, it might push information about the local weather, travel guides, or flight ticket discounts to you.
If you mentioned that "my baby is 6 months old", it will proactively push information about baby growth milestones and parenting tips.
Even if you connect your calendar, it can help you draft meeting agendas, remind you to buy birthday gifts, or recommend restaurants in your business trip destination.
What's most eye - catching is that the design concept of Pulse is completely different from today's "never - ending" apps.
Samir Ahmed, the technical leader of ChatGPT, clearly stated, "This experience has an end. It's designed to serve you, not to make you scroll endlessly."
The content pushed every day is carefully selected. Once you finish reading, it's over, and you won't be trapped in an endless information vortex.
Moreover, each update is only valid for the day, unless you actively save it as a conversation or ask follow - up questions.
This is simply a devastating blow to today's "algorithm - fed" model - useful information comes to you actively, instead of making you search blindly in a sea of information.
OpenAI conducted a test with college students in ChatGPT Lab and found an interesting phenomenon:
Many students initially thought Pulse was just so - so, but when they started to actively tell ChatGPT what kind of content they wanted to see, they immediately felt its power.
Let's look at some real cases shared by the official:
Isaac, a top student, received precise advice on train ticket booking times and commuting information he didn't know before the next day after talking to ChatGPT about travel planning.
Hexi, a diving enthusiast, shared the difficulties she encountered in diving classes. ChatGPT not only gave targeted advice but also made an analogy between diving and risk management - it really understood her interests.
This ability to "draw inferences from one instance" really has the flavor of a personal assistant.
However, the cost behind this personalized experience is quite high. To make Pulse really "understand you", you have to hand over a large amount of personal data to OpenAI. Specifically:
- If you turn on the "Reference History" in ChatGPT, it will review your past conversations to determine the research direction.
- If you have connected your calendar and email, the system will prompt you to confirm whether to allow ChatGPT to access these apps to help you plan your schedule. Kaplan said that users must click "Accept" for it to take effect.
Although OpenAI emphasizes that "the training data for Pulse is processed in the same way as normal conversations", is it really worth sacrificing privacy for convenience? And will it create an "information cocoon"? In response, OpenAI only said that it has set up "multiple security filters" but didn't disclose the specific details at all.
For many users, before having a "Jarvis" at hand, this kind of "black - box" privacy protection promise is really hard to trust.
From a technical perspective, Pulse is actually OpenAI's first step towards AI agents. Fidji Simo, the CEO of OpenAI's application department, said in a blog post:
"The next frontier is agents - AI assistants that can act on your behalf and collaborate with you like team members."
Here's the blog address: https://fidjisimo.substack.com/p/a-new-paradigm-of-proactive-steerable
This means that in the future, ChatGPT won't just stop at research and information organization. It will be able to automatically make plans for you, take actions according to goals, and proactively remind you at critical moments.
With this transformation in human - machine interaction, traditional search engines and news apps may face great pressure. After all, who can resist a smart and proactive personal assistant?
Here's the original OpenAI blog address:
https://openai.com/index/introducing-chatgpt-pulse/
This article is from the WeChat official account "APPSO". Author: Discovering tomorrow's products. 36Kr is published with permission.