diff options
Diffstat (limited to '')
| -rw-r--r-- | 2022/talks/maint.md | 26 |
1 files changed, 23 insertions, 3 deletions
diff --git a/2022/talks/maint.md b/2022/talks/maint.md index 8f179de8..f7f4571d 100644 --- a/2022/talks/maint.md +++ b/2022/talks/maint.md @@ -94,27 +94,35 @@ The initial questions below were answered live, and as they may be found in the ### Live Q&A Q: This seems to assume that there will be $ contributions commensurate with the value of the project vs. everyone freeloading because there is no incentive to pay + A: This concept of economy will make all participants stakeholders, so there is an incentive to pay. There are a number of open questions, though, as this idea is new. Q: Are you aware of projects like OpenQ (https://www.openq.dev/)? Would that fit the model in your opinion? + A: Not familiar with OpenQ Q: I see incredible amounts of overlap with the SourceCred system https://sourcecred.io/ , where attribution of antecedents, graph of contributions, fair-in-hindsight backpropogation is built-in. Are you aware of SourceCred? + A: Not very familiar with it [answered in more detail live, but I've also added more on this after the live section of Q&A below] Q: How is this different from money? Not in some abstract ownership vs attribution way. Open source funding is an incentive problem, which this does not change as far as I can see. + A: Money can be revisited in the future but maybe not immediately. Q: How would you approach a viable experiment? + A: We have a Github Action that automatically creates the boilerplate of files and directories needed for the accounting. Further development through dialogue because of the many questions open. Q: Given that the oversight is a social process, how do you constrain the cognitive and time burdens of deciding the values of attributed contributions? + A: We start with simple heuristics and first tackle the majority of issues. Q: How are the attribution amounts calculated? (ORGA NOTE: restored from a prior version; welcome to remove if this was your question and you removed it intentionally) + A: Dialectical Inheritance Attribution (DIA), a social practice described in the talk. One heuristic strategy that will be used will be "Analyze, Appraise, Anonymize, Attribute." Q: What are your assumptions about human nature vis. self interest vs. altruism? + A: In the live Q&A I meandered on this point, so let me answer more clearly here. ABE does not assume anything about human nature. It is neither cynical nor idealistic. Rather, it achieves the goals I mentioned in the talk in a purely structural way, setting up the source of value in such a way (DIA) that emergent incentives align with the common good. The "common good" emerges as simply being the aggregate of desires in a particular context. Since selfish desires cancel out at a sufficiently large scale, the incentive resultant from DIA forms what is good for everyone rather than what is good for some, in other words, exactly what we mean by the common good. @@ -124,6 +132,7 @@ Another way of saying this is that in an ABE system, you don't have to choose be In response to this, some may say, well then we won't get credit for being "truly" good. To them I would say, it's so much more important for our world to teach people to be better than for us to worry about who is "truly" good on their own. If our world lets selfish and parochial people become cruel and deluded and further the cause of hate and injustice, then we have failed them and each other. Even the worst among us has great capacity for good, and the wonder of ABE is that it could use them to do good, teach them to be good, a kind of "Aikido" that redirects their inclinations to align with (without being limited by) what everyone agrees is good. After all, if we teach moral lessons in the ivory towers of our churches, temples, mosques and synagogues, while also teaching by our actual practices and systems that selfishness and winning and being egotistical are good, then what results can we expect? Q: URL of the project? + A: - The founding documents for the prototype economic system are at https://github.com/drym-org/foundation - The accounting system which you can use in your projects (contains setup instructions) is at https://github.com/drym-org/old-abe @@ -136,6 +145,7 @@ The rest of these questions will be categorized under "ABE Now" -- relating to t #### ABE Now Q: How is this different from splitting donations to my project with my partners? + A: It's like that, but on a grand scale, formalized, and "automated" in the sense that creators are not required to participate in this process. Rather, it is a service provided to members of the community by members of the community. In addition, it also: 1. Recognizes antecedents in both directions. It's not just sharing proceeds from your project with your buddies, but also sharing with creators whose works and ideas are reflected in yours. And likewise, it's others sharing proceeds from their projects with you. 2. Encourages investment. It's easy enough to write a small project with your buddies, but when you have big dreams, you need big resources. If you are doing a startup in today's system, you divide "ownership" shares with your buddies and also with investors with deep pockets who can help you scale your project up to provide the maximum value. It's the same in ABE, except that anyone can be an investor simply by paying money to the project. This allows you to scale up your project by the support of ordinary mortals and not only "angels." It also means that every project will scale up to the right extent -- not too much and not too little -- because there are no incentives to wring value out of projects when there are more efficient ways to get the same amount value -- there are no barriers to becoming an investor, after all. If your particular horse isn't winning, there is no cost, and indeed an incentive, to pick another horse. Of course, in an ABE system, these horses aren't even competing, and there usually wouldn't even be a clear boundary between them! @@ -145,34 +155,44 @@ Q: How is this different from SourceCred and OpenQ? When there are technologies A: Systems like SourceCred are promising, and it's great that they're being developed. Technologies such as these will be indispensable to ABE's operations in the long term. At the same time, I want to strongly emphasize that technology is not the _basis_ of the new system. Instead, the basis of ABE is dialogue and agreement. This is a central idea because it means that anyone who has ever contributed value, and anyone who is contributing value today, and anyone who will contribute value in the future, can rest in the safety of collective attribution and be recognized and empowered -- people like you. Aside from sharing your work, there is nothing technological that you need to do (e.g. record your contribution on a blockchain, or be part of a software project that is using an attribution-oriented compensation scheme such as SourceCred, or have patents on your ideas, or anything else) in order to be eligible to be recognized for what you did, are doing, and will do. Q: What prevents bad actors from taking over? + A: There are many possible kinds of bad actor. + 1. Those who use your project and don't pay. + For now, this is OK and expected. But as the system scales, becoming eligible to _receive_ attributive payments means consenting to participate in ABE wholesale. So the more valuable a project is, the greater is its incentive to participate. + 2. Those who will make improvements and sell independently instead of contributing back. + This person is operating under the assumption that they will be able to generate more money on their own than through others via well-established channels of attribution and use. This assumption is generally unlikely to hold. + 3. Those who attempt to set standards that benefit themselves. + Because standards are set in an anonymized way, such self-serving standards are only likely to prevail on small scales where participants cannot be truly anonymous. At larger scales, this "Dialectical Mirror" ensures that these incentives cancel out ensuring that fair standards win over selfish ones. Additionally, since DIA is applied globally -- that is, the standards agreed upon in special cases are generalized to the maximum extent possible -- self-serving incentives in special cases would be negated by standards decided in the general case. To put this all in simple terms, "desires that benefit only oneself don't scale, desires that benefit all do." I call this the "Good vs Evil" principle. It is a very interesting mathematical property of an ABE system. + 4. Those who do not report payments. + The ABE constitution requires that payments being reported is a collective responsibility -- both payers as well as payees can report it. Payers have an incentive to report it because it counts as an investment. Payees have an incentive to report it because being in non-conformance with the constitution can make the project ineligible to continue receiving attributive payments from the system. But in general, yes it is important to put safeguards in place to protect against identified risks, and no doubt, there is a lot of work to do on this front. If you can think of such risks, you can help by bringing them up and/or helping to implement the necessary safeguards. We're all in this together! Q: If in ABE some portion of payments to my project go to upstream projects, then isn't there less incentive for me to work on my project? + A: Some portion of your revenues go upstream, but by the same token, some portion of revenues of downstream projects come to you. Determining the precise proportions of value is not an easy problem, and it will take time and experimentation to arrive at the "sweet spot" for simultaneously incentivizing future work while fairly recognizing past work. Q: I don't see a license on ABE projects. What gives? -Whether you have a license or not, and whether your project is proprietary or not, it is in all cases eligible to be recognized by ABE (but note that if your code is not open source, then there is less value there to be recognized -- constituting an _incentive_ to release your code). +A: Whether you have a license or not, and whether your project is proprietary or not, it is in all cases eligible to be recognized by ABE (but note that if your code is not open source, then there is less value there to be recognized -- constituting an _incentive_ to release your code). By virtue of this, having a license on ABE projects would amount to introducing something distracting which has no bearing on the process. Additionally, as ABE endorses non-ownership, that essentially puts these projects in the public domain. Projects that are not owned don't need licenses. After all, who would be in a position to issue such a license if no one has special privileges to begin with? Q: OK, but why not use the Unlicense or Creative Commons? -Licenses like the Unlicense, well intentioned though they are, don't really help because they offer a glimpse at an open and free world that they don't provide any means of attaining, leading to complacency on the part of the user. To be fair, we owe such licenses a debt of gratitude as they have helped us get to the point where people are more receptive to the idea of non-ownership. But such crutches hinder us now -- if a potential user sees the Unlicense and if this vision of a free world takes the place of the need for a real solution in their minds, then no one is better off for it. On the other hand, a declaration of non-ownership is, to paraphrase Leo Vivier, a thread of curiosity that you can follow to reveal more complete answers. Follow that thread, friends! +A: Licenses like the Unlicense, well intentioned though they are, don't really help because they offer a glimpse at an open and free world that they don't provide any means of attaining, leading to complacency on the part of the user. To be fair, we owe such licenses a debt of gratitude as they have helped us get to the point where people are more receptive to the idea of non-ownership. But such crutches hinder us now -- if a potential user sees the Unlicense and if this vision of a free world takes the place of the need for a real solution in their minds, then no one is better off for it. On the other hand, a declaration of non-ownership is, to paraphrase Leo Vivier, a thread of curiosity that you can follow to reveal more complete answers. Follow that thread, friends! Q: DIA sounds like an involved process. How can it be done efficiently enough to usefully keep up with the pace of contributions to a project? -We use an idea that we call "Renormalization," which I'll explain soon. First, the system, at least at the initial stages, assumes that preferences in the system are consistent. If a person says they like A better than B and B better than C, we assume that they will like A better than C. By making this assumption, we can reduce the hard problem of appraising the value of a contribution to a project to the problem of simply appraising its value in relation to any other single aspect of the project that has already been appraised, and then "renormalizing" (i.e. ensuring the proportions total to 1, or a 100%) the attributed proportions to include the newly created value. For instance, it's hard to say how valuable a particular bug fix is to a project, but it's much easier to say how valuable it is in comparison to another bug fix that was already appraised. So, once there is a seed of appraised contributions, it becomes much easier to appraise new contributions. Periodically, the process of DIA would be conducted afresh to apply the standards more rigorously. This is analogous to a similar algorithm followed in the field of robotics, where a robot navigating a large room can have a rough idea of its position even if it is unable to see its surroundings, by maintaining an internal model of its own movements until visual data is available. Likewise, we can maintain useful appraisals of the value of pull requests even before we have had a chance to conduct the full process of DIA, which may be done at a much less frequent rate (e.g. monthly or quarterly) than the frequency of contributions. +A: We use an idea that we call "Renormalization," which I'll explain soon. First, the system, at least at the initial stages, assumes that preferences in the system are consistent. If a person says they like A better than B and B better than C, we assume that they will like A better than C. By making this assumption, we can reduce the hard problem of appraising the value of a contribution to a project to the problem of simply appraising its value in relation to any other single aspect of the project that has already been appraised, and then "renormalizing" (i.e. ensuring the proportions total to 1, or a 100%) the attributed proportions to include the newly created value. For instance, it's hard to say how valuable a particular bug fix is to a project, but it's much easier to say how valuable it is in comparison to another bug fix that was already appraised. So, once there is a seed of appraised contributions, it becomes much easier to appraise new contributions. Periodically, the process of DIA would be conducted afresh to apply the standards more rigorously. This is analogous to a similar algorithm followed in the field of robotics, where a robot navigating a large room can have a rough idea of its position even if it is unable to see its surroundings, by maintaining an internal model of its own movements until visual data is available. Likewise, we can maintain useful appraisals of the value of pull requests even before we have had a chance to conduct the full process of DIA, which may be done at a much less frequent rate (e.g. monthly or quarterly) than the frequency of contributions. #### ABE Future |