Posts
Comments
Partial Update
I'm fully funded! In fact, as of writing. I have nearly 3x more funding than I asked for! The money really started rolling in after Alex Tabarrok plugged my project on marginal revolution, which I did not expect.
I'm not sure why people continued to give me money after the project was fully funded but I'm very thankful. (Of course, the more money I get the longer I can work on the project increasing the chance of success). The fundraising only officially ends 15 September, so I'm not going to give a full report until then.
I was hoping to do some neglected chores while the campaign ran, but since I'm already fully-funded, I feel I can't leave people hanging. So right now I'm collecting people who would like to use my platform to produce public goods to solicit requirements from them. If you're interested join our Discord if you haven't already.
I agree with this comment except for these two points
About your element/matrix example, the problem is indeed free-riding, but DACs would only solve the problem if he wouldn't keep developing it unless he got a certain amount of funding (which would make it a coordination problem). If he doesn't want to condition the development on a certain level of funding, then funding will be based only on the goodwill of people. So DACs also don't solve cases where a reward is deserved, but you're gonna do the thing regardless.
I don't think he is going to do it regardless. Presumably if he doesn't get enough money he has to shut-down his company completely or reduce the amount he spends on open-source and focus on the more profitable consulting side of the business. If DACs were used they could invest more in the open source.
I think the focus on public goods in the post is a bit confusing. Not because it's a problem to have and offer that as the main motivation for this, but because it makes it seem like it's only relevant for public goods, where it's still as relevant for private goods if there's a big upfront cost to producing any at all.
Private goods already have a mechanism to solve this problem: investment. That's why I think DACs are not that relevant for private goods.
I think these examples are consistent with my views: Sometimes public goods are provided because the demand for them is so great they people overcome the co-ordination problem or just allow people to free-ride (lighthouses) or sometimes public goods are not that public (bee pollination) and so private enterprises pay for them for very little free-riding.
The problem is not the public-goods are never provided, but that the are under-provided: either they are not provided or are of low-quality relative to what is economically efficient.
I think it's about 10%, but I'll need more experience to see how much it actually works out on average. For comparison, I think normal crowdfunding has about 5-8% fees, but only if you succeed.
will you be happy if this only works for club goods? I think here the private entrepreneurial case is much stronger, and even in the original paper a lot of the discussion of strongest equilibria is for $K=N$ and club goods.
No, because of the loss of value due to unnecessary exclusion.
What mechanisms other than "culture" can you think of to stop the "raise a bunch of money, then put ads in anyway"
So I think for the platform I'm creating, I'll only allow projects that are public goods (so if it's media/software it must have a permissive license). That means that anyone can take the public good with ads and just remove the ads. That's the main reason why open-source doesn't have ads, because everytime someone tries to add ads, people fork the project to remove the ads (on Linux usually the package maintainers will remove the ads).
is there any knowledge of Kickstarter or GoFundMe being approached and asked to implement this functionality?
Alex Tabarrok said he tried in one of the videos linked. I haven't spoken to him about it, but I assume they weren't interested.
I was hoping that proving the concept would cause more established players to adopt the system.
But what I’ve done in the past is that people sign up with my platform, get a code, and put the code in their pubic comment that they can attach to a contribution on the third-party platform. Then if they want to claim any rights attached to the contribution from me, I check that the code is the right one, and if it is, believe them that they’re either the person who made the contribution or that the person who made the contribution wanted to gift it to them.
Oh cool, that's a good idea. Then you can piggy back off existing crowdfunding platforms instead of making your own one.
people sign up with my platform,
Do you have a link? It sounds cool. I want to check it out.
Well, let’s say I barely have the money to pay for my own cost of living or that I consider a number of AI safety orgs to also be the even-more-cost-effective uses of my money.
I think I see your point. I agree that DACs don't solve this type of free-riding.
Yes. Thanks. Fixed
I haven't read the book, could you provide such examples?
I think there are obvious cases of where public goods version of something is under-provided (lower-quality) compared to the club good version of that thing. e.g. Proprietary software being more polished than open source software, HBO higher quality than YouTube.
Where I live the police are useless so neighborhood watches have sprung up in the wealthy areas. However, my impression is that the neighborhood watches appear much later and only in richer areas than you'd expect in an efficient market. Particular many people got private security years before neighborhood watches, and lower-income areas don't have neighborhood watches even though they would probably be willing to spend 1/1000 of their income (just pay 1 guy in the community of 1000 people) for more safety.
For Route B, I'm not sure I can find a super compelling sentence, that's why I thought it would be easier to just have something I could point to (Hey look at all these cool things we managed to raise money for, I can help you raise money too!).
For Route A, I'd would be surprised if there was a trusted influencer who would risk their reputation on this weird financial scheme, unless there were at least several examples showing that it worked. I think what I'm doing is a prerequisite for this route.
So the idea is that you only force the people to pay who actually are willing to pay. Obviously in the real world, you don't know who these people are. In the post I wrote:
The theoretical model of Dominant Assurance Contracts assumes away some things that you have to deal with in the real word
- Perfect information about pricing: In the example problem above, we assumed that 10 villagers were willing to pay $15 to pave the road. In real life you do not have that information and risk over-pricing/under-pricing your contract. Presumably the contracts run in the real world that failed were over-priced.
So in the real world an asteroid deflection DAC risks being over-priced (and then we all die from an asteroid) or under-priced (some people free-ride). I still think this is an improvement over other mechanisms.
Maybe we shouldn't call it a Refund Bonus but something like "Attention Compensation". I would say that you are expecting to profit (3) derived from your efforts of reading the contract and agreeing to pledge and then not actually getting what you paid for (not 4).
Personally I think it's irrational to give money expecting a DAC to fail. make-yass made a post about this. I think the long-run social equilibrium is that people realize that it's dumb to gamble on DACs failing and don't do it.
Thanks for the feedback.
I think there is a massive product problem with the idea -- people don't understand it, think it is a scam,
I think you hit the nail on the head. I agree that this is the main problem.
If your efforts were more directed at the problem of getting people to understand and be excited about crowdfunding.
That was the point of this post? It's possible that I'm doing a bad job at this. Do you have any suggestions for what I should be trying to do instead?
So initially I wanted to price at $829 but https://manifold.markets/moyamo/how-much-money-will-my-metacrowdfun only gave me like 30% chance of succeeding. So I changed the price to $629 which gave me 45% chance of succeeding.
There are people willing to take financial risks for profits.
then if you do offer to do the project using a dominant assurance contract, then either the project doesn't get funded and you get screwed, or the project does get funded and (some of) the funders get screwed.
The way I think of it is that there is a co-ordination problem. People want public goods, some people want to provide it, but they have difficulty coordinating. So either the project gets funded: yay people get the public good they wanted. Or it does not and people didn't really want that public good anyway (and you pay the cost for poorly predicting peoples preferences and sucking up their attention). So to me it's more of a win-win.
All of it. I'm going to cover the transaction costs.
I really like that. It draws attention to the fact that the "bonus" is compensation for time and effort spent on trying to buy a product that was never delivered.
This is the Howey test
- An investment of money
- In a common enterprise
- With the expectation of profit
- To be derived from the efforts of others
I think it fails 3 and 4 simultaneously. There is no "expectation of profit to be derived from the efforts of others". If the contract succeeds then others make an effort but you do not make a profit. If the contract fails you make a profit, but not from the efforts of others.
Thanks for you comment, it's very helpful.
Sorry if I missed it in the article, but who are the parties you would like to get on board to put up the refund bonuses?
It would the producer of the public good (e.g. for my project I put up the collateral).
Can you bootstrap them by using DACs to fundraise for bigger and bigger refund bonuses?
Possibly? I'm not sure why you'd do that?
Are you aware of any issues with securities law, since people can make monetary profits off refund bonuses?
I disagree that a Refund Bonus is a security. It's a refund. To me it's when you buy something, but it comes broken, so the store gives you a voucher to make up for your troubles.
I imagine (no idea really) that the law of the country of the person who starts a fundraiser will be what matters legally.
I'm in South Africa but from what I can tell, if you work with US dollars and do something illegal the FBI will come after you, so I wouldn't be confident that only South African law applies.
Can you maybe just pick out a charity fundraising platform that has funding thresholds and refunds (and is used by consequentialists), and refund bonuses to everyone who contributed to failed fundraisers?
This is actually a cool idea. I don't know how I'd manage to get people's details for giving refund without co-operating with the fundraising platform, and my impression is that most platforms are hesitant to do things like this. If you know of a platform that would be keen on trying this, please tell me!
Put differently, I would love a non-speciesist world and a world where funding in AI safety is allocated efficiently. I’m currently working only on the second problem, so one could argue that I’m freeriding on others’ solutions to the first.
I don't quite understand this point. You could work on AI Safety and donate to animal charities if you don't want to free-ride.
Btw., I think it would be useful to mention what the article is about in the title. I would not have read it (even though I’m very interested in the topic) if Dony hadn’t told me that it’s about DACs!
You're right. I didn't want the title to just be "Dominant Assurance Contracts" because I assumed that most people have never heard of them and tried to come up with something more interesting, but maybe enough people on lesswrong have heard of them so I should probably be more straight forward.
I created a manifold.market.
Yes, individuals have the incentive to mitigate existential risk, but only in their lifetime, and possibly the lifetime of their grandchildren.
Institutions can last many generations and also allow people to coordinate and work together. In theory it's difficult to form a company with the sole purpose of mitigating existential risk, since investors will be pushing you to grow big or make huge profits (in practice it seems like Conjecture managed to do this?). With an eternal company the bondholders want your company to not take risks for big-profits.
One concern on the alignment of executive compensation is that it's especially hard to get executives to care about what happens after they die, unless their perpetual bonds go to their heirs, unlike a regular pension. Even then, they or their heirs can sell those bonds, no? At least in the US, we have laws setting time limits on constraints about how heirs can use or dispose of property left to them.
This is a good point, but I think empirically people don't really divest/diversify their inheritance? This is something that could be tested. In theory the perpetual bonds of a large eternal company should be some of the safest assets to hold, similar to U.S. bonds. So I don't think most people will want to sell these for other assets.
When I look at the world's actually existing very old companies, [...] Do we, or should we expect to, see any signs that these kinds of companies ae unusually motivated to reduce existential risks?
Possibly. I'm not sure we should look at what long-lasting companies do now, since we can't be sure that they will continue to last long (maybe by now the management has gone bad). It would probably be better to look at how long-lasting companies behaved in the past.
it's pretty questionable whether "corporation" is the unit of institution to focus on.
I agree. AI Safety is a public good and so suffers from the https://en.wikipedia.org/wiki/Free-rider_problem and so even if you had eternal companies, they would have to co-ordinate some how. But I think it would be easier for eternal companies to coordinate on AI Safety compared to normal companies.
I'm also pretty skeptical that slack is compatible with financial metrics as the primary optimization lever, whether amortized or instantaneous.
I'm not sure what you mean by this. I think a lot of companies already give their employees a lot of slack? e.g. Apparently Google used to allow every employee to spend 20% of their time on pet projects (although I've heard this practice no longer exists).
Also, it's unclear [...] that theoretical stock value deviates much from perpetual bond value. Both are quite sensitive to perceived stability of company.
Surely their must be a difference. A bond is not exposed to the "upside" of the companies profits only the "downside" of them defaulting. I think maybe a good analogy is how parents behave with their children. Parents are much less exposed to the upside of their children's accomplishments (if you start a business and become a multi-millionaire you're parents see very little of that money), but are much more exposed to the downsides of their children's failures (if your business fails, then they might have to live with you, increasing rent and food costs). Understandably parents tend to push their children to go for safe jobs (accountant, doctor, lawyer, programmer) rather than doing risky jobs with high upside (actor, artist, musician).
I think in the same way an eternal company (if the incentives work) will behave in a less risky way.
I like this idea, because I'm too lazy to review pull requests. It would be great if other people could just review and vote on them for me :P
I'm looking for feedback on my understanding of Corrigibility.
I skimmed CHAI, Assistance Games, And Fully Updated Deferece . Is the key criticism that The Off-Switch Game ignores the embeddedness of the human? i.e. The agent will stochastically model the human and so will just try to predict what the human will say instead of asking it? This limitation is mentioned on Page 3.
Is this criticism correct or am I missing something else?
I tried diving into All of Statistics but I found it to be way too concise. I didn't get past Chapter 3. In particular Chapter 3 felt like a list of distributions and some arbitrary properties. It felt like I wasn't really getting an intuition for what these distributions represent or why these properties are interesting. In the end I dropped the book because I felt like a wasn't really learning anything.
My negative experience with this book is likely a result of me having no previous experience with statitistics.