Thank you everyone for participating in this market. Its been very informative and fun to learn both sides of this market, about writing styles (thanks @NcyRocks ), about the inconsistency of various Gpt-predictors and reasons behind it, and many interesting conversations on discord dms.

The market will resolve to YES. The essay was indeed submitted to Stanford by an applicant (personal identification details removed from the essay) who was turned down that year. The Prompt B is written by ChatGPT after i directed it to add more personality to the answer.

Thanks again for participating, feel free to message, suggest ideas, or continue discussions. Have a good day guys :)

Ran this through GPT-2 Output detector and A showed as 99% real while B as 77% generated from GPT-2. Fairly trivial and I put together an extensive YouTube video about how this works. https://www.youtube.com/watch?v=whbNCSZb3c8

@PatrickDelaney strong words for a man with with only 1 share 🧐

@PatrickDelaney I have no idea what the reputation of the market maker is and there are comments below including types which confuse me. Also there's no real way to make any money on this anymore unless you bet in the 100's+ because it's so obvious.

@PatrickDelaney ironically, "typos" not, "types"

@PatrickDelaney that's a fair point.

The typo below was due to manifold showing the person's position as opposite to their position, leading to a bit of confusion.

I'm new to Manifold, and your point about reputation is fair (I'm not betting for play money, I'm making this market because it tickled my curiousity and now in hindsight, well, didn't expect this much swing of opinion over what's human written). I'm very amused, however, that this much commotion happened over what i thought was fairly obvious. So, this market was successful in my eyes.

@PatrickDelaney hahahaha

@firstuserhere regarding the first point - next to a person's name, it says "X is predicting (yes/no) at Y%" but that yes/no status is often different from their real/final position.

@NcyRocks what do you know that we don't???

@jonsimon Nothing special, just the way ChatGPT writes. It loves putting in things like ‘I live a zero-waste lifestyle’ and would be unlikely to yield the ‘soccer’ bit, the ‘I think better at night’ bit, or to change the subject as quickly as A does.

@jonsimon Put it this way: if it’s B, I’m going to be very impressed with @firstuserhere‘s ability to make ChatGPT write with different idiosyncrasies, and to write like ChatGPT.

But I have a lot of faith in my own ability to pick up on writing styles.

@NcyRocks this market was made 3 months ago when i just joined manifold. You pointed out how it tends to gravitate towards a centre, doesn't "dive in" to a character, and I've found myself thinking about your phrase "ability to pick up on writing styles" quite a few times since. I've not yet tried this particular prompt with GPT-4 yet, especially Bing, but I'd be surprised if it wasn't even harder to distinguish the human and gpt-4 one now (especially @BoltonBailey's bit about how a human is likely to overshare their perceived flaws, and wonder if GPT-4 is good at it)

It seems trivial to talk about the ability of LLMs to write good text nowadays because of how often we interact with them. We've seen other people's brilliant prompts, our own interactions and stories etc etc (I've lost count of how many tens of thousands of words I've had GPT-4 generate for me) but every once in a while, I just pause and reflect at just how good this stuff ended up being, from where we were just a few years (or even months) ago. Just makes me feel some sense of awe whenever I step back and see it

I don't know how any of you people are confident enough to have bet this all the way up to >90%... at this point I'm counter-betting just because the upside potential is so high

plot twist, they're both bots

@Odoacre haha but no. One of them (of course, with personal details which have been removed here) has been seen by the stanford admission committee people

I feel like this is obvious

@anne I agree, but I see you bought NO shares (while I hold a lot of YES). I’d like to hear why it’s obvious to you that it’s B?

Here's one that was generated by ChatGPT, prompted by me just now:

Hi future roommate,

I'm really excited to meet you and be your roommate for the upcoming academic year at Stanford!

A little bit about me: I'm originally from the East Coast, and I love exploring new places and trying new foods. I'm also really into sports, both playing and watching. In my free time, you can usually find me playing basketball or cheering on my favorite teams. I'm also a bit of a neat freak, so you can count on me to help keep our shared space clean and organized.

One thing that's really important to me is staying on top of my studies, but I also believe in the importance of taking breaks and having fun. I hope we can balance each other out in that regard and create a positive and productive living environment for ourselves.

I'm looking forward to getting to know you and making the most of our time together at Stanford. Let's make this a great year!

Best, Tetraspace

This seems stylistically closest to the second one, and straight-up uses the same phrase "I'm also a bit of a neat freak" (Same as Bolton Bailey's ChatGPT generation: Which one is human, which one is AI (Stanford Roommate essay)? | Manifold Markets)

@Tetraspace If you think B is by chatgpt, then you should buy YES shares, not NO shares!

@firstuserhere This is a very cool market! Was the human written one written by an actual Stanford applicant? (Feel free not to answer).

@BoltonBailey message me on discord? firstuserhere#1188

My output from GhatGPT with this prompt:

Hey there future roommate!

I'm really excited to meet you and start this new journey together at Stanford. A little about me - I'm from a small town in the Midwest and am the first in my family to go to college. I'm a bit of a neat freak and love to keep my living space organized, so hopefully that won't be too much of a problem for you. In my free time, I enjoy playing soccer and hiking in the beautiful California wilderness. I'm also a bit of a foodie and love trying out new restaurants and cooking new recipes. I'm open to trying any type of cuisine, so if you have any recommendations or want to grab a meal together, let me know! I'm really looking forward to getting to know you and making the most of our time at Stanford.

Best, [Your Name]

@BoltonBailey i don't know if clarification is needed but the market resolves to No is B is chatGPT and Yes if A is ChatGPT written

@firstuserhere Your prompt literally says "Resolves to YES if A is human written."

@BoltonBailey yes, and the line above it explicitly mentions that one of A and B is Human, the other ChatGPT

@firstuserhere In your description you say

"Resolves to YES if A is human written."

In your comment right above this, you say

"Yes if A is ChatGPT written"

Which is it?

@BoltonBailey oh I'm sorry, made a typo in the comment. The description is correct.

@BoltonBailey any way to remove the erroneous comment? Perhaps by reporting it?

@firstuserhere I see the issue - my comment is marked saying I bought NO, but I actually sold that and bought YES.

Not sure if reporting the comment is appropriate given it's an honest mistake and the context makes that clear.

@BoltonBailey yeah, that's why i replied in the first place and then checked users, and it showed reverse position

@firstuserhere regardless, to everyone who sees this, please stick with the description of resolution

@firstuserhere resolution in the description*