Another thing to note: I’m optimizing on a time-horizon of around 10-30 years. Making a business first could easily take 6-20 years.
Ozzie Gooen
Karma: 1,235
I feel pretty flattered to be even vaguely categorized with all of those folks, but I think it’s pretty unlikely of working out that well (it almost always is). If I was pretty sure (>30%) I could make a company as large as Apple/Twitter/Tesla/YC, I’d be pretty happy to go that route.
I’ve chatted to hundreds of entrepreneurs and tried this, arguably, twice before. That said, if later it’s predicted that going the more direct business way would be better for total expected value, I could definitely be open to changing later.
“Metaculus − 70k, Ozzie Gooen − 70k, Jacob Lagerros − 27k. These are small companies that need funding; why are you acting as grant-givers here rather than as special interest investors?”
I’m not sure why you think all of these are companies. Metaculus is a company, but the other two aren’t.
Personally, I think it would be pretty neat if this group (or a similar one) were to later set up the legal infrastructure to properly invest in groups where that would make sense. But this would take quite a bit of time (both fixed costs and marginal costs), and if there are only a few groups per year (one, in this case, I believe) is probably not worth it.
I honestly think this was one of the more obvious ones on the list. 39k for one full year of work is a bit of a steal, especially for someone who already has the mathematical background, video production skills, and audience. I imagine if CEA were to try to recreate that it would have a pretty hard time, plus the recruitment would be quite a challenge.
Happy to chime in here.
I’ve previously worked at Ought for a few months and have helped them make Mosaic. I’ve been talking a decent amount with different Ought team members. We share broad interests of how to break down reasoning, but are executing this in very different ways. Mosaic works in breaking down a very large space of problems into tiny text subproblems. I’m working on a prediction application, which works by having people predict probabilities of future events, and separately share information about their thinking. I think that essentially no one who sees both applications would consider them to be equivalent.
I’m doing very similar work as part of my research at FHI. The plan is not to attempt to become a business or monetize in the foreseeable future. I’ve been around the startup scene a lot before, and have come to better understand the limitations of getting money in business ways. In almost all cases, from what I can tell, experimental and charitable desires become pushed aside in order to optimize for profits. I’ve considered this with Guesstimate. Originally I thought I could make a business that also would be useful to EAs, but later realized that it would be exceptionally difficult. Most realistic business strategies for sales looked like domain-specific tools, for instance, a real-estate-specific distribution application, which would sell a lot more but be quite useless for EA causes.
In this case, my first and main priority is to experiment/innovate in the space. I think that doing this in the research setting at this point will be the best way to ensure that the work stays focussed on the long-term benefits.
Hypothetically, if in a few years we wind up with something that was optimized for EA uses, but happens to be easily monetizable with low effort, then that could be a useful way of helping to fund things. However, I really don’t want to commit to a specific path at this stage, which is really early on.
It’s a general platform, not specific to any domain. In a similar vein to the Good Judgement Project’s platform, but with other features I think will be useful.
I plan to try it out with some groups in the next few months.
Some other options, in case someone is particularly interested:
Guesstimate is being used by several EAs, and could use some development help. It’s making enough from purchasers to cover basic costs, but not enough for myself to fund significant extra development. It’s open source.
I’ve been working on a new forecasting application written in ReasonML/node/graphql. If that sounds interesting to you, let me know!
Happy to advise and help people contribute to either, though both do have learning curves.
For what it’s worth, I’d really like to see more highly speculative ideas like this with simple Fermi estimates of value. Even if it’s ridiculous, it could still have a very small chance of being viable, and of inspiring other ideas.
This may be an April 1st post, but honestly, I appreciate the main idea.
Agreed that tokenizing and markets would be difficult in the short term.
The main possible aspect would be evaluating many projects and estimating the impact of them, as opposed to just giving an ordering for the very top projects. A rubric can be used for evaluations.
For example, say you have some rubric where every project was scored on “importance”, “novelty”, and “quanity” or similar. Then you divy up the prize money proportional to those things on the rubric and make the results public.
That could be an uncomfortable level of transparency for some people, but it would help foster a discussion of which projects are the most valuable.
Kudos for pursuing this and not getting too attached to it as to not be willing to give it up when the evidence showed such made sense.
Here’s a question: do you have thoughts on how this could have failed faster? If you were to go about this again, what would you have done differently in order to spend even fewer resources on it?
Thanks! Some quick responses to parts.
Legal Implications
I think copying PredictIt would be pretty messy. I’m curious about the feasibility of using crypto, similar to CriptoKitties. It seems like several crypto groups did essentially a similar thing, and perhaps in the medium-term, they will be recognized as being legal in the United States.
“If OpenPhil initially promise $10m, and then later bump it to $20m, projects that have already sold their tokens cannot expand to take advantage of this increase, so it is effectively pure windfall with no incentive effect. ”
I agree cohorts are one solution. Another issue though, is that if the audience thought there was a decent chance of a bump, then the (“40% chance there would be a $10mil bump”) would be factored into the price.
Any chance you would consider aspects from my Impact Prizes blog post, for future competitions? https://forum.effectivealtruism.org/posts/2cCDhxmG36m3ybYbq/impact-prizes-as-an-alternative-to-certificates-of-impact
Realize this may be a long shot, but I’d like to see more experimentation here.
It’s also possible there won’t be much competition. There may only be 3-6 entities with serious chances of making an AGI. One idea is to have safety researchers in almost every entity.
It’s definitely an important question.
In this case, the equivalent is a “car safety” nonprofit that goes around to all the car companies to help them make safe cars. The AI safety initiatives would attempt to make sure that they can help or advise whatever groups do make an AGI. However, knowing how to advise those companies does require making a few cars internally for experimentation.
I believe that OpenAI basically publically stated that they are willing to work with any groups close to AGI, but forgot where they mentioned this.
That makes sense to me. When I said “harder to scale”, I mean harder to “put a bunch on top of each other”. In some ways it’s not as elegant.
Agreed that Impact Prizes are one ways that Certificates of Impact could work long-term. Like, one group places $100k of Impact Prizes for 2030, where it will only be used to purchase Certificates of Impact.
Impact Prizes as an alternative to Certificates of Impact
I’d generally agree with that.
Honestly, the technical infrastructure for Certificates of Impact would be very similar to that for Impact Prizes as I discuss them above. I think both would be really interesting to test at larger scales.
Impact Prizes may need less hype though, but may be more difficult to scale.
Not that I know of. Paul has a lot of stuff going on, for one thing. :)
I think some people are still excited about Certificates of Impact though.
Yep, it’s related. I’ve looked a bit into social impact bonds; they actually are quite specific and precise though (they pay with a specific interest rate, for very specific outcomes).
There have been different thoughts on how to use markets for charitable work before. The Impact Certificate Post and its comments list some.
Basically, though it’s a bit extra short when weighted for what we can change. Transformative narrow AI or other transformative technologies could also apply.