Above the Fold: Manifold integrates GPT-3

Manifold plays chess, draws penises, and scores at the 99.5th percentile in a prediction contest

Hello! I was going to title this post as “Manifold draws penises” and create a prediction market on what its email open rate is compared to our usual newsletter but decided I wasn’t brave enough to do it. Anyways, you’re probably curious as to the context so I’ll just let you start reading. Enjoy!

Manifold plays chess (and loses)

A few months ago Alex, a 2000 rated player on lichess, challenged Manifold to a game of chess. Manifold loss spectacularly and resigned after just 12 moves with people manipulating the game so Alex would win.

We are currently in the middle of a new game, this time against a level 3 stockfish bot. For reference, this bot’s skill is roughly comparable to a human who has 100-500 hours of chess. So not amazing, but not terrible either.

The structure of the game is slightly different this time, with votes on the next move only influenced by YES shareholders. The more shares you have, the more powerful your vote. This would seem robust, except users have created a second market to bet on! This way they can hold Yes shares in the main market to allow them to vote, whilst holding a larger NO position in the secondary market.

Top trader Jack shared some thoughts on Discord for how to make future games more robust, although being able to manipulate the games seems to be half the fun of it!

“ok here's the ruleset proposal I currently have in mind for Manifold Plays Chess 3:

• One main market: Will Manifold win?

• For each move, the participants get 1 day to vote on a shortlist of moves (approval voting by making comments with proposed moves and liking them).

• We create a market for each of the top N moves: If this move is chosen, what will be the win probability on the main market after 5 moves? These run for another day and then the one with the top probability is chosen (the others resolve N/A). There will be anti-snipe rules (e.g. use the average probability over the last 5 hours, or allow the market to extend if there's excessive volatility just before closing). The main concern is whether people manipulate the main market (i.e. buy a bunch of NO and try to force a blunder) but I anticipate the market will aggressively correct the main market price based on stockfish probabilities.

You could require people to buy in a YES stake in the main market to be eligible to participate in the move selection process.”

Chat GPT

Chat GPT-3 has taken the world by storm over the past few months and it’s been fascinating to see how its impact on the education system has transpired. In recent news, it passed the final exam for a master’s program. Some are even suggesting it could replace google, although I remain skeptical of this claim. If any readers have a strong conviction it will cause a significant reduction in google usage, make a market with reasonable parameters and I’ll bet against you.

One of our most popular markets with 560 traders is whether Google will include a chatbot of its own soon. What do you predict?

We’ve been using GPT-3 to automatically assign groups to newly created markets with none assigned. It has had… mixed results and has left us seeking non-AI solutions to improve correct group assignments in the short term.

Fortunately, where the Manifold devs have fallen short with integrating Chat GPT-3, our users have made sure we still have plenty of markets about it. My personal favourite is this market which incentivises users to use it to draw penises to win mana.

Jokes aside, most of our user’s markets have been about the highly anticipated launch of ChatGPT-4. Our GPT4 speculation group has lots of interesting markets, one fun one being will GPT-4 get the Monty Fall problem correct?

Our markets think it is unlikely GPT-4 comes out this quarter, so you have some time to think of your own predictions to add!

Manifold plays Model UN

Some of our users are looking to organise a Model UN on Manifold. You can apply to participate here! The format is still being discussed, but each player would represent a country and pursue different objectives. With everyone predicting how the world shapes out as the game unfolds.

Scott Alexander’s Prediction Contest

The results are in for the 2022 prediction contest and prediction markets delivered in spectacular fashion.

Who Predicted 2022?

Prediction markets did extraordinarily well during this competition, scoring at the 99.5th percentile - ie they beat 506 of the 508 participants, plus all other forms of aggregation. But this is an unfair comparison: our participants were only allowed to spend five minutes max researching each question, but we couldn’t control prediction market participants; they spent however long they wanted. That means prediction markets’ victory doesn’t necessarily mean they’re better than other aggregation methods - it might just mean that people who can do lots of research beat people who do less research.2  Next year's contest will have some participants who do more research, and hopefully provide a fairer test.

If Manifold Markets performs as well as last year we should outperform everyone except the top 15 performers out of 3500 entries. Do you think we can once again score so highly now that our markets have significantly more traders than they did this time last year?

Manifold Markets Subreddit

We have a subreddit: https://www.reddit.com/r/ManifoldMarkets/
We are going to start posting video content containing our markets there! We tend to prefer discussion on Manifold itself, but want to encourage users creating content surrounding our markets and sharing them for that sweet sweet karma.

FRI is hiring!

If you are looking to begin a career in forecasting, the Forecasting Research Institute is taking applications for a number of positions until February 1st.

Forecasting could use more gender diversity - EA Forum

Apparently, a lot of men have been applying, which isn’t a huge surprise as many adjacent fields are relatively male-dominated. One of our key goals with Manifold is to lower the barrier to forecasting for everyone, so it was a huge pleasure to see this comment on the above post:

“Thanks for writing this, I think it's really important and well expressed. Diversity of thought and experiences should be valuable .

I will say as a woman who was skeptical about her personal fit regarding forecasting I found playing around Manilfold Markets and trying to make some forecasts helpful and the community there is friendly. So if anyone wants a easy way to just try for themselves what forecasting might look like head to: Manifold Markets”

-Rina

Site Updates

• New site-wide search

• Improvements to feed algorithms

• Mobile app push notifications for @mentions, replies, and comments

• Infinite scrolling related markets at bottom of market page

• Updates to Maniswipe (Wip)

• Changes to signed-out view