>Will the U.S. enact an AI safety federal statute or executive order in 2026?
will (if "yes") wind up effectively applying to Google DeepMind, which is headquartered in London, but presumably is affected by Google and the U.S. market...
( While I'm generally not happy with Brexit, I do count it as fortunate that EU restrictions don't cripple Google DeepMind the way they cripple AI in the EU. )
"You are welcome to create a bot account to forecast and participate in addition to your regular Metaculus account. Create a bot account here, and visit the AI Benchmark contest page for templates and support building your bot."
Sweet! How do I collect the prize money again (nicholaskross)? (IMPORTANT NOTES: my metaculus username is FractalCycle, and my paypal email is NOT my metaculus account email.)
Is there somewhere to click to see the judgement criteria for each question, or do we have to guess based on the top-line question alone?
For example, does 'ai created song' include something composed b humans with an AI singing the words (eg Neuro covers), something where AI wrote the first draft of the words but humans edited them and made the music and sang it, songs where a human used generative ai tools to make thousands of instrument loops and searched through them to pick out a couple to include as backing tracks on a song they wrote, etc.?
My estimates range from like 80% to 5% depending on the judging criteria.
I think my entry deserves to have won this contest. I mean I didn't submit an entry, but I still think it should have won, and the wrong person was given the first place that I deserved.
I gotta say, Metaculus is pretty unexciting to browse versus polymarket or manifold. UI choices make it annoying to find and browse questions. The delay between a question being posed and answers being "revealed" is especially irritating! Oh look an interesting question, I wonder what metac-- nope, can't see that, sorry, come back in three weeks.
I also (after many auto-scrolling backtracks) browsed last year's questions and have to say I am pretty disappointed in the results - many of them seem basically just like "linear interpolation from base rate to zero" because of the necessary decay in probability for a "will happen in 2025" question. And there were some pretty big whiffs!
I'm especially disappointed that my favorite 2025 question (AI starbucks order) was annulled. Goes to show that the hardest part is asking the right question with all the necessary carve-outs and caveats I suppose.
The Philadelphia Eagles will be eliminated in the first round of the NFL playoffs and their head coach Nick Sirriani will be shockingly fired only a year after winning the Super Bowl
it’s genuinely cool that bot accounts and AI-generated forecasts—that feels like the right direction for a platform that’s supposed to be about forecasting the future. On the other hand, the criticism about U.S.-centric questions and unclear or overly narrow resolution criteria seems pretty fair. https://bitlife-simulator.io
It’s a fascinating mix of crowdsourced wisdom and serious skin in the game, especially with the $10,000 prize pool drawing in the sharpest minds https://happyrewards.io/ from the ACX community.
Finding a trustworthy site in Britain can be a nightmare but I reckon I have struck gold here with the payout speeds. I recommend you check https://bigwinss.uk/ if you want to see some serious returns on your deposits. I put in twenty quid and turned it into a massive payout which was back in my bank faster than I expected so it is nice to finally find a platform that values its players and delivers those big wins we are all looking for without any messing about.
How do skipped questions impact overall score? I hope they are counted as community predictions? I've skipped 8 questions (mix of US politics and highly technical stuff) where I am aware I would not be able to outguess community and where I was unable to quickly (like, in seconds, I am not spending a lot of time on this) find prediction market to copy.
As a newcomer, I like that the contest is driven by community-suggested questions and open to both experts and beginners https://fnf.lol —it makes the whole thing feel more approachable. The snapshot deadline and separate bot option are helpful to understand up front, especially for anyone experimenting with forecasting or automation for the first time.
geometry dash lite https://geometrydash-lite.io/ teaches players how to practice timing, remember patterns, overcome frustration, and feel proud when completing difficult levels.
This prediction contest approach is fascinating! I love how it encourages collective forecasting and taps into the wisdom of the crowd. What inspired you to tie it to Metaculus, and how do you see this influencing broader discussions on predictions? Looking forward to your insights! https://snowrider3d.com
That’s a nice snapshot of the ACX/Metaculus forecasting contest this year. It’s interesting how they’ve structured it: reader-sourced questions, a fixed deadline for contest rankings, and then ongoing updates that only affect the site leaderboard. That balance between competitive stakes (the $10,000 prize pool) and community-driven participation makes it feel both serious and collaborative. https://www.opensky-cc.com
What were the questions?
Here you go: https://www.metaculus.com/tournament/ACX2026/?order_by=-open_time
I think there's a large American bias.
True! It is an interesting question whether
>Will the U.S. enact an AI safety federal statute or executive order in 2026?
will (if "yes") wind up effectively applying to Google DeepMind, which is headquartered in London, but presumably is affected by Google and the U.S. market...
( While I'm generally not happy with Brexit, I do count it as fortunate that EU restrictions don't cripple Google DeepMind the way they cripple AI in the EU. )
I've been doing my PhD on AI forecasting systems. May I submit a set of AI generated forecasts fron one of my lab's prototypes?
Per the meatculus link:
"You are welcome to create a bot account to forecast and participate in addition to your regular Metaculus account. Create a bot account here, and visit the AI Benchmark contest page for templates and support building your bot."
Sweet! How do I collect the prize money again (nicholaskross)? (IMPORTANT NOTES: my metaculus username is FractalCycle, and my paypal email is NOT my metaculus account email.)
Man I used to love doing these every year, but now that it's my day job it feels less like a break and more just like doing work.
Is there somewhere to click to see the judgement criteria for each question, or do we have to guess based on the top-line question alone?
For example, does 'ai created song' include something composed b humans with an AI singing the words (eg Neuro covers), something where AI wrote the first draft of the words but humans edited them and made the music and sang it, songs where a human used generative ai tools to make thousands of instrument loops and searched through them to pick out a couple to include as backing tracks on a song they wrote, etc.?
My estimates range from like 80% to 5% depending on the judging criteria.
Yeah, you can click on each question to get more information, including the resolution criteria. For the AI songs one, just go here (https://www.metaculus.com/questions/40969/ai-song-in-top-20-before-2027/) and scroll down.
Extremely disappointing questions, these are mostly just American questions you can find on any prediction market already.
I think my entry deserves to have won this contest. I mean I didn't submit an entry, but I still think it should have won, and the wrong person was given the first place that I deserved.
I gotta say, Metaculus is pretty unexciting to browse versus polymarket or manifold. UI choices make it annoying to find and browse questions. The delay between a question being posed and answers being "revealed" is especially irritating! Oh look an interesting question, I wonder what metac-- nope, can't see that, sorry, come back in three weeks.
I also (after many auto-scrolling backtracks) browsed last year's questions and have to say I am pretty disappointed in the results - many of them seem basically just like "linear interpolation from base rate to zero" because of the necessary decay in probability for a "will happen in 2025" question. And there were some pretty big whiffs!
I'm especially disappointed that my favorite 2025 question (AI starbucks order) was annulled. Goes to show that the hardest part is asking the right question with all the necessary carve-outs and caveats I suppose.
The Philadelphia Eagles will be eliminated in the first round of the NFL playoffs and their head coach Nick Sirriani will be shockingly fired only a year after winning the Super Bowl
it’s genuinely cool that bot accounts and AI-generated forecasts—that feels like the right direction for a platform that’s supposed to be about forecasting the future. On the other hand, the criticism about U.S.-centric questions and unclear or overly narrow resolution criteria seems pretty fair. https://bitlife-simulator.io
It’s a fascinating mix of crowdsourced wisdom and serious skin in the game, especially with the $10,000 prize pool drawing in the sharpest minds https://happyrewards.io/ from the ACX community.
If I don't have time to keep reaffirming my prediction on Metaculus, does it make sense to let it "never expire"?
Finding a trustworthy site in Britain can be a nightmare but I reckon I have struck gold here with the payout speeds. I recommend you check https://bigwinss.uk/ if you want to see some serious returns on your deposits. I put in twenty quid and turned it into a massive payout which was back in my bank faster than I expected so it is nice to finally find a platform that values its players and delivers those big wins we are all looking for without any messing about.
How do skipped questions impact overall score? I hope they are counted as community predictions? I've skipped 8 questions (mix of US politics and highly technical stuff) where I am aware I would not be able to outguess community and where I was unable to quickly (like, in seconds, I am not spending a lot of time on this) find prediction market to copy.
As a newcomer, I like that the contest is driven by community-suggested questions and open to both experts and beginners https://fnf.lol —it makes the whole thing feel more approachable. The snapshot deadline and separate bot option are helpful to understand up front, especially for anyone experimenting with forecasting or automation for the first time.
geometry dash lite https://geometrydash-lite.io/ teaches players how to practice timing, remember patterns, overcome frustration, and feel proud when completing difficult levels.
How can they give the correct prediction? What do they rely on? https://blockblast.org/
This prediction contest approach is fascinating! I love how it encourages collective forecasting and taps into the wisdom of the crowd. What inspired you to tie it to Metaculus, and how do you see this influencing broader discussions on predictions? Looking forward to your insights! https://snowrider3d.com
That’s a nice snapshot of the ACX/Metaculus forecasting contest this year. It’s interesting how they’ve structured it: reader-sourced questions, a fixed deadline for contest rankings, and then ongoing updates that only affect the site leaderboard. That balance between competitive stakes (the $10,000 prize pool) and community-driven participation makes it feel both serious and collaborative. https://www.opensky-cc.com