Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster
Apr 12, 2001
70,313
41,930


OpenAI today announced the worldwide rollout of group chats in ChatGPT, allowing multiple people to collaborate with ChatGPT in a shared conversation. OpenAI started testing group chats with a small number of people last week, but the feedback was positive so it is now expanding the feature to everyone.

openai-chatgpt-group-chats.jpg

With group chats, ChatGPT users can create a shared conversation space and invite friends, family, or coworkers. Group chats are separate from private conversations, and OpenAI says that the group creator's personal ChatGPT memory is never shared with anyone in the chat.

According to OpenAI, group chats can be useful for planning a weekend trip with friends, designing a backyard garden, collaborating on design ideas, finding a restaurant that suits everyone's tastes, or settling debates. Group chats can also be used for research for school or work projects, incorporating shared articles, notes, and questions.

Creating a group chat can be done by tapping on the people icon in the top right corner of any new or existing chat. People can be invited by sharing a link, and anyone with the link can invite more people, with up to 20 supported. Group chats are aggregated in a clearly-labeled section of the sidebar, separate from personal chats.

OpenAI says ChatGPT was taught new social behaviors for group chats. It follows the flow of conversation and decides when to respond and when to stay quiet based on the context of the conversation. Mentioning ChatGPT in a message will prompt it to respond, and ChatGPT can react to messages with emojis.

Group chats are available to all logged-in users on free, Go, Plus, and Pro ChatGPT plans. OpenAI says that it will continue to refine the feature as more people start using it.

Article Link: ChatGPT Group Chats Now Available Worldwide
 
I’d like to try this new way of interacting with ChatGPT without involving anyone else… maybe I can set a second account on my iPad to test drive this, and see if it reveals anything from what I’ve shared with GPT in previous chats, or how does it react when two people disagree or have an argument in front of it.
 
  • Like
Reactions: Frantisekj
I know too many people that are insufferable in the way they use ChatGPT and literally can't seem to make a decision/opinion without it and refuse to recognize that how they ask and how they've interacted with it in the past has influence in how it responds because, in the end, it is programmed to please the user. How is it going to handle 4 people with diverse opinions, subtexts, preferences and styles?

This thing will have the power to ruin relationships. lol.

Does it take on the settings/profile of the person that creates the chat? Because holy moly does that seem dangerous. Someone could start an account, acclimate it to whatever point of view they want and then invite 20 unsuspecting people into their sycophantic gaslighting chamber.
 
Siri, the new 'Egg Freckles'

They need to train these other AI models to trash Siri without mercy. Only then will this be right! :D
 
I know too many people that are insufferable in the way they use ChatGPT
It's becoming quite annoying. Especially those people who just put anything they see into AI and paste the answer into social chats as a truth, without any thinking.

I wish this bubble would pop soon, both to stop unwanted AI being rammed into everything, and also to return a little thinking to humanity.
 
It's becoming quite annoying. Especially those people who just put anything they see into AI and paste the answer into social chats as a truth, without any thinking.

I wish this bubble would pop soon, both to stop unwanted AI being rammed into everything, and also to return a little thinking to humanity.
Not only that, but the information those on that chat could be wrong and it would affect 5 people or how many are in it. Not a good idea, it will fizzle out quick, I hope.
 
The next person to inevitably write that LLMs are not "AI", "intelligent", "can't think", "just pattern matching", "word predictions", I challenge you to a public duel:

You vs Gemini 3 Pro in an intelligence contest. Maybe graduate college level math test?
 
Last edited:
The next person to inevitably write that LLMs are not "AI", "intelligent", "can't think", "just pattern matching", "word predictions", I challenge you to a public duel:

You vs Gemini 3 Pro in an intelligence contest. Maybe graduate college level math test?
re: math test - is basically just proving the point youre fighting against. Thats like having a duel with an encyclopedia, it doesn't prove anything.
put a real world problem, with subtleties, situational and uncommon edge cases, which requires actual reasoning and problem solving, not just filling in the already known blanks and things become very clear
 
It's becoming quite annoying. Especially those people who just put anything they see into AI and paste the answer into social chats as a truth, without any thinking.
Yep, I’ve been seeing this lately. It’s like they want people to know they use ChatGPT. “Hey, look how smart I am! I know how to ask a bot about this!”.

This also reveals how dumb people is, instead of asking ChatGPT and mixing that with a quick search on Google or Wikipedia, to elaborate a “smart” explanation, and look smarter, they instead just copy the outcome of ChatGPT. It’s like… lowest effort possible.
 
  • Like
Reactions: attila and ikir
Group rats ,think tanks aw yes got it!
Study the humanoids for conversation and later install it in the llm`s.
Freaky stuff.
They should just patch into all the TV Stations and listen to that.
Wouldnt that be enough human interactions to add to their collections for LLM`s.
 
  • Like
Reactions: attila and ikir
re: math test - is basically just proving the point youre fighting against. Thats like having a duel with an encyclopedia, it doesn't prove anything.
put a real world problem, with subtleties, situational and uncommon edge cases, which requires actual reasoning and problem solving, not just filling in the already known blanks and things become very clear
Ok, give us an example problem.
 
Also might add chatgpt learns from all the groupies then later in turn charges us all for being their test subjects.
We are the product becomes even more a scary phrase.
Charge me for MY information WHAT?> weapon to my head.
 
  • Like
Reactions: attila
The next person to inevitably write that LLMs are not "AI", "intelligent", "can't think", "just pattern matching", "word predictions", I challenge you to a public duel:

You vs Gemini 3 Pro in an intelligence contest. Maybe graduate college level math test?

Gemini doesn't do calculations. It looks up answers.

You don't need a degree for that.
 
I’m not exactly happy about all this AI interference in our personal lives. I mean, I’m using ChatGPT or Copilot to help me with my research… but we are going too far.
 
  • Like
Reactions: wanha
Gemini doesn't do calculations. It looks up answers.

You don't need a degree for that.
No it does do calculations. In fact, you can even ask it to show you step by step how it arrived at the answer. Heck, you can even ask it to write a program that does the calculation.

This is how out of touch commentators are on Macrumors about state of AI.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.