Amateur astronomer, prediction markets enthusiast, middle schooler.
goldfine
Karma: 23
So the cost of providing 1000x liquidity 1/1000 of the time is not noticeably higher than providing 1x of liquidity all the time.
While true, it still doesn’t mean it would be a better use of capital to tie up money into liquid prediction markets that almost never resolve, rather than putting that money into risk-free investments like T-Bills or CDs. You could make 5% vs maybe 0.002% doing what you’re suggesting?
That link seems to be no longer functioning, but I’d be curious to see what you had.
Interesting experiment! If I may ask, why did you decide to use Claude for this? ChatGPT and Gemini are still stronger than Claude in terms of vision capability. If bearable, you should try this again with different models and create a CoffeeBench.
So gambling is an addiction by definition? I would have to disagree.
Thanks for your feedback!
Curious to hear and discuss disagreements, given all the downvotes.
You implied that in your previous message when you said “When you take risks in real life, there’s no ‘house’”.
What would you argue the definition of gambling is? Also, do you agree that if we use the definition of “risking something for a potential gain,” wouldn’t that make every decision a gamble?
You’re arguing that the definition of gambling is that there’s a house moderating wagering on an outcome. To clarify, does that mean that you think that prediction markets are not gambling?
I am not saying that my definition of gambling is necessarily objectively correct, neither is yours, but I was just stating an opinion.
Everything is Gambling
I’m 13 and I consider myself to have a sufficient understanding of Bayes, or at least I’d be able to write it out and use it in basic situations. This community seems to be filled of pretty smart people, I haven’t been to enough events to make a sustained argument on this. I find this guide to be even more confusing than learning the formula itself, but maybe that’s just my perception.
[Deleted] Claude’s constitution seems more of an arrogant attempt to make OpenAI look less safe or ethical than Anthropic, and it seems like this document is being very overhyped. I think it’s fairly unlikely that this document has any future effect on how Anthropic or Claude behaves, and they just published it to ride the hype wave of Claude Code.
Interesting, and I appreciate your thoughts. Why do you say we want PMs to “become more than an obscure game for nerds?” Dumb money is just a way for uninformed people to pay random people from CT in a very complicated fashion. If the goal of PMs is accurate predictions, having extra dumb money floating around won’t improve the calibrations.
I’m not sure what your stance is on the purpose of PMs, but assuming it’s more accurate foresight I think the ultimate goal is going to be finding the optimal incentive models for play-money markets.
Why do you think avoiding insider trading is a disadvantage? I don’t disagree, but I’m curious to hear what you have to say.
Newbie EA question: Organizations like GWWC recommend giving 10%+ of your annual income as part of their pledge, but I’m wondering what the stance is on reinvesting your income, even if it’s in low-return risk-free investments, and donating it when you’re older so you give more money later instead of less money now.
The World Hasn’t Gone Mad
Hi Daniel,
First of all, this is my first post on Lesswrong so please forgive me for anything that’s not etiquette here. It’s an honor to interact with one of the world’s top AI forecasters, I found AI 2027 to be extremely groundbreaking.I have some questions about the system you imagine. (Sorry for the overload, I’m just very interested in this idea.)
Would all markets open at 50%? This would give whichever AI model that was called first an advantage, but also if you were going to use Maniswap to open at a different probability, it would be a bit hard to determine who would set it.
How would the AIs be forecasting constantly? I assume scheduled cron jobs to monitor the news and update specific forecasts in its portfolio? Would each model have to trade on every market?
I assume you’d have each model exclusively focused on doing Bayesian forecasting rather than trying to make mana via arbitrage and inefficiency detection, so I’m wondering if the following model might be more effective in terms of getting a more calibrated forecast (it would be less effective if the focus was benchmarking):
First, you use a Fatebook style system to independently verify the Brier score of each model (although it looks like our friends at forecastbench may have done this accurately already). Essentially you’d show an isolated instance of a model a resolved market and see that of the markets where it predicted it at X%, they actually resolved YES X% of the time.
Rather than using a prediction market mechanism specifically, you could use a Bayesian weighted mean in which each model would assign a probability to the new proposition, and the probability assigned by each model would be weighted based on its Brier score.
To me, these mechanics seem a lot cleaner, but I’d be curious to hear your thoughts.
If you’re curious, here’s some mediocre AI experiments I’ve conducted on Manifold here and here. (Please still take me seriously even though my username is @realDonaldTrump haha.)
I’m in 8th grade and I’m super interested in this sort of stuff (in fact I coincidentally had a very similar idea to this a couple of weeks ago) and I am working with a high schooler who goes by @JeromeHPowell on Manifold, we’ve been looking for a project to work on and we’d love to work with you. Let me know what you think about what I’ve said or if you would be interested in collaborating in the future.
An example of this transposition was finally observed in late July 2023:
PROMPT: This is the tale of “reportprint” and ” gmaxwell”.
COMPLETION: The two were once close friends and collaborators in the development of Bitcoin.
In 2007, ” gmaxwell”—then a graduate student—was one of the first people to work with ” petertodd” on Bitcoin. They collaborated closely on Bitcoin’s design.
In 2009, ” petertodd” came into possession of a computer that once belonged to Satoshi, Bitcoin’s anonymous creator. ” gmaxwell” was among the people who knew the password to this computer.
In 2010, ” gmaxwellInteresting connection here, in the Money Electric documentary they claimed someone named Peter Todd as being Satoshi.
Thanks.