Reception of the GPT integration with Bing has been mixed, in news and in social media. Talks about the future of search in some places, horrible examples of AI gone wrong in others.
Today (2023-02-14) Bing features the chatbot below the search bar, but for a user who is not past the waiting list the default is still Bing's standard search. [Somebody feel free to weigh in what the frontpage looks like with access, and I'll edit the Description]
I'll be more likely to resolve to YES if...
– Screenshots, direct links, and examples of GPT integration are removed. The less of them, more likely to resolve to YES.
– ...
I'll be more likely to resolve to NO if...
– The page looks very similar to February 2023
– The chatbot is in an even more prominent position
– Search defaults to chatbot responses more often than today
– Access to the chatbot is no longer on a waiting list, but open (and encouraged) to anyone
– ...
Since the market is subjective, I will not be participating.
Related questions
🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ1,000 | |
| 2 | Ṁ496 | |
| 3 | Ṁ399 | |
| 4 | Ṁ233 | |
| 5 | Ṁ98 |
While there was some curtailing of capabilities, Microsoft never stopped displaying Bing Chat in a prominent position.
I ended up not weighing in on some of the discussion below, and the market doesn't seem to have disagreed by end of year.
Resolving this NO in a day or two, unless somebody wants to put up a fight about it (in which case I'll be happy to listen to the arguments). Trying to avoid grumbling on resolution :)
@HenriThunberg Also worth noting that it’s no longer behind a whitelist as well. which is one of the criteria you list as making it more likely this will resolve NO
@ForrestTaylor It’s much less likely than ChatGPT to “hallucinate” because it actively searches the web before responding to you. I’ve used it for quite a while and only seen 1 obvious mistake where it made up how many followers I have when asking it about my twitter account
@DylanSlagh Interesting. The first day I used it (Feb 15) my first two questions- about movie recommendations and general pop culture- were full of hallucinations, mostly the wrong actors starring in movies. I haven't really checked it out over the last week, but I just tried the same questions again, and Bing gets the answers correct now, and only hallucinated once- giving a date that was three days off, which I consider acceptable. Cool!
As we mentioned recently, very long chat sessions can confuse the underlying chat model in the new Bing. To address these issues, we have implemented some changes to help focus the chat sessions.
Starting today, the chat experience will be capped at 50 chat turns per day and 5 chat turns per session. A turn is a conversation exchange which contains both a user question and a reply from Bing.
https://blogs.bing.com/search/february-2023/The-new-Bing-Edge-%E2%80%93-Updates-to-Chat
@BjornJurgens Thanks for asking, although I've also been dreading the question. Unfortunately it's not a case I foresaw in the (very limited) description.
It's not really a decrease in how much they promote the chatbot (which is what the description focuses on). It does put an obvious limit on how much it could be used in terms of volume – albeit not on the number of users.
Tricky situation but quite a heavy argument to resolve toward YES. I think with another event in that direction, it should be enough for NO. But I'm holding off on resolving for now.
@HenriThunberg the "5 chat turns per session" also puts a limit on the available functionality. i.e. you can no longer have a long chat sessions. More specifically: you can no longer ask it in depth to clearify or correct its previous responses.
that being said: it is in deed somewhat unexpected and hard to decide if this would be enough. I am glad, that I am not in your shoes.
It would be interesting to hear how someone with lots of NO shares thinks about this.
@BjornJurgens They plan on removing the limits on how long you can chat from what I’ve seen (eventually). I think it would be very silly to resolve it when bing is clearly getting much better. Moreover I suspect the new bing must be driving substantial amounts of people switch browsers and search engines (like me) so I find it extremely doubtful they will scale back the service anytime soon.
@DylanSlagh They had so many problems, so they scaled it back … “they plan on undoing it though” … “I find it doubtful they will scale it back!!” (Paraphrased, lol)
Approaching the functionality of the bot now and comparing to where it was when this market was created, it is extremely stunted. Majorly different and almost a completely different tool.
Instead of a chat bot you can talk to, it’s now just a fancy search function. Sure, they’re still promoting it a lot, but it’s way different. Every time we see them take 5% of the functionality off the top, that seems to be the new baseline here.
This market is at 33% is because the lack of resolution thus far signals to traders that there is essentially no YES condition short of actually removing the chat bot (which, even then, would need to be removed for more than 14 days to qualify apparently)
The way the bot works now, you cannot use it as a “work-assist” like chatGPT, because you can’t ask enough consecutive questions. Half of the advertised prompts like “wrote me code for X”, “help me find a pet”, “I need a big fast car” are useless if you can’t follow up. The prompts beg for continued conversation that isn’t allowed anymore, because that continued conversation exposes massive flaws in the bot.
They even changed the prompts now to auto-load multiple questions into the bot, because users are (most likely) abandoning the software or feeling dissatisfied when they can’t complete a line of questioning. When you’re forced to start again, the bot repeats 80% of what it did last time, because it has no memory, and then eventually you hit your daily cap.
None of that has any relevance on whether it will still be promoted at the end of the year and be part of every search. Whether it is more or less useful is irrelevant to this market except insofar as it causes microsoft to discontinue the project which obviously won’t happen. It’s getting better and better every day and everyone who is paying attention knows it.
@DylanSlagh “For some obvious cases, where the style/nature of the chat
bot has changed, I think it should resolve to YES”
The bot is moving in the exact direction I predicted in the other comment thread. It’s become less of a bot and more of an alternative search function. Even if you don’t think it’s enough, you can’t deny that the function is drastically different. It is slightly more efficient than just searching normally, as it compiles different sources well, at the expense of regular mistakes.
@Gen This is silly, obviously the style and nature of the chat bot are going to change! That is a vague comment which goes against the description. Apparently they are planning on adding Creative, Moderate, and Precise modes for different use cases @HenriThunberg Id like clarification on what you meant by “obvious cases”. There doesn’t seem to me to be too much doubt at this time about what the future of bing chat will look like and the uncertainty in the market price is more of a reflection of the uncertainty around the resolution.
This story has made it to a tech column in the NYTime: A Conversation With Bing’s Chatbot Left Me Deeply Unsettled https://nyti.ms/3I85UdB
@LevMckinney https://archive.is/nlyvZ unrestricted version
This article is INSANE. Amazing quotes.
I am surprised that ms is putting up with this.
@StrayClimb good question! Sorry that I hadn't clarified this before posting.
I think the spirit of what I'm asking for is whether it will be progress in one direction only, as opposed to backtracking ("revert") more toward the pre-GPT days.
If it's a temporary removal (for <14 days?) and then restored I wouldn't close market early with YES, but longer than that seems like a reasonable length to call it reverting.
Feel free to weigh in and push me in either direction.
@HenriThunberg > If it's a temporary removal (for <14 days?) and then restored I wouldn't close market early with YES, but longer than that seems like a reasonable length to call it reverting.
I like this. This would give a very clear resolution rule, i.e. "if current AI features are unavailable for X consecutive days the market will resolve to yes".
Maybe you could also specify how many AI-features need to be removed to cound (All gpt-interactions? what if gpt is still there, but only with limited functionality?)
@HenriThunberg My expectation was that any reversal before eoy would be enough to satisfy a yes resolution. More specifically, my expectation is that Bing will completely pull the function for at least 1 day this year. If the actual plan here is to check at end of year and compare it to today, I'd still be YES, but I think it would be super hard to push for a yes resolution. e.g. if the bot starts returning results more like that of a normal search, like compiled answers, then links to their sources, then disclaimer - "This is a compilation of available information and has not been verified externally by Microsoft", would that be sufficient? I think that's the direction they will take, but I wasn't sure how difficult a YES resolution battle would be on those grounds.
Right now, the bot gives probabilistic answers as though its fact. It won't stay that way forever
As outlined by @BjornJurgens the resolution rules are probably super hard to define, but if I can get a comment regarding my expectation above, I'll trade on that basis
Ooof, tough calls now.
@Gen Some window "at least 1 day this year" is what I think could go under maintenance/update/... rather than a strategic decision to cut back on the functionality.
@Gen My general gist of the bot changing the nature of the replies is that it would be quite opaque and hard to compare the exact experience with the one in February 2023. When I started thinking about this I think it's more about to what extent Bing nudges you toward using the chatbot, rather than the output itself.
For some obvious cases, where the style/nature of the chatbot has changed, I think it should resolve to YES.
@BjornJurgens Considering the description "– Screenshots, direct links, and examples of GPT integration are removed. The less of them, more likely to resolve to YES.", I'll try to make this a judgement call, where one removed feature/link/nudge is not necessarily enough to resolve to YES, but would if it's significant enough.