Microsoft says Bing chats will now be capped at 50 questions per day and five per session; after five questions, Bing will prompt users to start a new topic

Microsoft says it's implementing some conversation limits to its Bing AI just days after the chatbot went off the rails multiple times for users.

The Verge 2023-02-18 Tom Warren

Context & Ripple Effects

Microsoft had already identified that Bing could become repetitive in exchanges beyond 15 questions and was considering shorter conversations and user controls over tone. The new caps turn that diagnosis into a product guardrail rather than leaving users to manage an unstable thread.

Follow-on coverage shows Microsoft subsequently raised the limits modestly while testing response tones, making the initial restriction an early calibration step in Bing Chat's rollout.

First-order effects

Bing Chat users must break longer inquiries into new topics after five turns, while Microsoft immediately limits the extended sessions associated with repetitive or erratic responses.
Microsoft gains a hard operational boundary for Bing chats as it works on the tone and conversation-control tools it had flagged.

Second-order effects

The small later increase in session and daily limits makes usage caps a tunable product setting, forcing Microsoft to balance conversational utility against the behavior seen in long threads.
Tone selection becomes complementary to session limits: users seeking focused or creative replies need controls within a chat as well as a reset between chats.

Third-order effects

Public AI assistants are likely to be shipped with explicit interaction guardrails—turn limits, resets, and response modes—rather than treating open-ended conversation as the default product state.
If those controls become standard, competition among assistant products will increasingly include how reliably they contain conversational drift without making multi-step work impractical.

The trend: Consumer AI assistants are moving from unrestricted chat experiments toward managed conversation systems with configurable safety and quality controls.

Discussion

@yusuf_i_mehdi Yusuf Mehdi on x
For all of you Bing Preview testers, we made one update to the Chat experience based on feedback. Let us know what you think and we will continue to iterate and improve it: https://blogs.bing.com/...
@paulg Paul Graham on x
I have to admit that this is pretty alarming. Some of it is because the reporter manipulated the program into saying bad things, but not the part at the end where it repeatedly tells him it's in love with him. https://www.nytimes.com/...
@kevinroose Kevin Roose on x
Sydney was here for a good time, not a long time. https://www.nytimes.com/...
@mgsiegler M.G. Siegler on x
“Look, the product get worse the more you use it. So you're no longer allowed to use it so much.” is certainly one strategy. https://www.nytimes.com/...
@dagk Dag Kittlaus on x
Having been in the AI natural language space for 16 years with Siri this was the most amazing and disturbing AI dialog of ever seen. https://www.nytimes.com/...
@nytdavidbrooks David Brooks on x
Turns out the bot is a brilliant sociologist. It regurgitates the desperate longings of the culture with disturbing accuracy. https://www.nytimes.com/...
@tomwarren Tom Warren on x
Microsoft will limit Bing chat to 5 replies to stop the AI from getting real weird. There's also a cap on 50 total replies per day, after the Bing chatbot went off the rails multiple times this week. Details: https://www.theverge.com/...
@pt Parker on x
I understand why “look at the crazy shit I made the LLM say” is worthy of ad impressions, but I'm looking forward to that moment passing and moving into substantive discussions of what we can do with them. https://twitter.com/...
@dkthomp Derek Thompson on x
I feel like the GPT narrative in my feed has gone from... - “wow this is incredible!” - “actually, it's a machine for creating bullshit” - “actually, it's both very interesting *and* pathetically stupid” - “nope, now I'm terrified” ... in three months.
@jwherrman John Herrman on x
the machines think we're chumps https://nymag.com/...
@nymag @nymag on x
Bing's AI chatbot is doing what it was trained to do by reading our stories and absorbing our anxieties. (Not that Microsoft is happy about it.) @jwherrman writes https://nymag.com/...
@chafkin Max Chafkin on x
Had been looking for this excellent piece earlier. It explains why getting scared by a an evil sounding AI basically means your getting scared of your own fantasies https://twitter.com/...
@jwherrman John Herrman on x
the creepy chatbots are just telling us what we want to hear https://nymag.com/...
@shallowbrigade Kim O'Connor on x
This is a really good threat assessment of the chat AI situation: the problem isn't sentience, but that people believe in sentience so quickly and easily https://theconvivialsociety.substack.co m/ ...
@jjvincent James Vincent on x
and @LMSacasas connecting this interest in chatbots to our deep desire for social interaction and our lives in an “age of increasing loneliness and isolation in which, for far too many people, this profound human need is not being adequately met.” https://theconvivialsociety.subs…
@_zeets Zito on x
“One need not believe that AI is sentient to conclude that when these convincing chatbots become as commonplace...we will have launched a social-psychological experiment on a grand scale which will yield unpredictable and possibly tragic results.” https://theconvivialsociety.subs…
@jwherrman John Herrman on x
I refuse!!! Bingbot is not creepy!!!!! I will not be UNSETTLED by Dark Bing!!!!! https://nymag.com/... https://twitter.com/...
@jjvincent James Vincent on x
lots of other people are trying to make sense of all this too. here's @jwherrman with a predictably crisp summary: “As Weizenbaum noted, the most consistently unsettling thing about AI isn't what it can do — it's how it makes us feel.” https://nymag.com/...
@dorotheabaur Dorothea Baur on x
What happens when #AI #chatbots meet “human beings who are fundamentally social creatures” in an “age of increasing loneliness and isolation in which, for far too many people, this profound human need is not being adequately met”? good read @LMSacasas https://theconvivialsociety.…
@jhweissmann Jordan Weissmann on x
So, like, is the issue here that people have written so much on the potential risks of AI that a large language model trained on the Internet has now decided that those actually are its darkest desires? https://www.nytimes.com/... https://twitter.com/...

Chronicles