Deep in the future, or maybe not as far as you think, exist the cupboards of perfection. When you step into them, you step out perfected in some way.
For some of the cupboards, this is a very simple process. Step in, step out buff. Or step in, step out knowing the rules of Bridge. Or step out, and look like the nearest-looking version of you that would, in conventional terms, be rated a 10. Of course, even these cupboards have many hidden subtleties of operation, but they are at least simple to understand in the broadest sense.
The Ultimate cupboard was an attempt to make a cupboard that would end all cupboards. You step in, and you have all and only those changes made to you that would, in your opinion, improve you.
But reservations soon emerged about the ultimate cupboard. What if, for example, you wanted to know everything that was part of the common knowledge of humanity- all publicly available sciences and arts- but, due to some quirk of your psychology, acquiring this knowledge would ruin you in some way- make you very sad, or make you an edgy nihilist or something like that. As a result, the Plus Ultimate cupboard was created. With the Plus Ultimate cupboard, you would get in and the changes requested would be performed, then a check would be run to ensure that you would approve of how new you had turned out- if this check failed, you would revert versions.
But the problem with the Plus Ultimate cupboard was that sometimes people got in and changed themselves, then got out and realized that, in light of their new perspective, there were other ways they wanted to change. This led to sequences of transformations, some very long. A new model was created, the Plus Ultimate Sequencer, which would immediately make you into the being that would emerge from the end, with the feature of the Plus Ultimate (that each stage is checked by the old you before being finalized) added.
But there was a problem, sure we might have:
A (accepts is an improvement) B (accepts is an improvement) C (accepts is an improvement) D
And so on, but we might get to the end, with, say, Z, but while each step may approve of the following step, Z might look pretty alien to A, maybe A would even be horrified by Z. Thus it would not be, from A’s perspective, self improvement.
So now we come to the Plus Ultimate Sequencer Deluxe. Here, we have a sequence where:
1. A specifies a series of changes
2. B is created satisfying those changes
3. A check is performed to determine whether A approves of B
4. B specifies what changes they would like, if any
5. C is created, meeting those specifications
6. C is inspected by A & B, and if both unanimously approve, the changes are made.
7. The process continues for as many steps as are unanimously agreed to by all proceeding ‘steps’.
[A variant existed that only required the approval of the first version and the latest version, but that had incentive compatibility problems]
This was all very good and well, but a bug of sorts emerged. We’d go A, B, C, D, then perhaps D wouldn’t be able to get consensus on E, so the process would stop, then D would just step out of the Cupboard at the end of the process and then, to arrive at their preferred E, enter the Cupboard again immediately and continue. The makers of the Plus Ultimate Sequencer Deluxe sought a solution and found one in a more dictatorial algorithm, the Iron Ultimate Sequencer.
A would change into B, B would change into C, and so on and so forth, at each stage it would be checked that A approved of each change. When a proposed change was reached which A did not agree to, the last stage that A would agree to, but implanted with a desire not to undergo any more changes, would step out of the cupboard.
Now this all seemed a bit chilling to some, an attempt by the present self to control your future self, so while arguments about whether it should even be allowed were going on a new Cupboard was made. The premise of this one was a little gentler. The Negotiator Ultimate Sequencer works as follows. You get in and enter a virtual space, and an improved version of yourself is created according to your specifications, then you and the improved version of yourself discuss what a further improvement would look like, and then after that you (the original being) decides on a second set of improvements in discussion with your improved self, that being is created. then you three discuss would a further improvement would look like, you come to a decision and so on. Each version of you has input to the discussion, but only the original makes the final call on each new version created. When the original decides there are no further improvements to be made, and that one of the created beings is the most ideal version of them possible, that being steps out of the cupboard.
A serious problem was that most people’s idealised selves are superb persuaders, and some of them could be quite sophistic in getting what they wanted. A patch that prevented the discussing selves from lying during the negotiation, but the fundamental problem remained that superintelligent, charming and knowledge versions of oneself might make one agree to changes that normally one would be deeply opposed to, and total honesty does little to ameliorate this. Various proposals floated around, like talking to one’s idealised selves through emissaries to blunt their charisma, but nothing was settled on as a universal protocol.
Of course the negotiator ultimate sequencer is, once again, vulnerable to the new version of you just not being happy with the outcome, and stepping in the Cupboard again, meaning your veto was for nought. However, the element of negotiation and discussion tends to make it rarer. Attempts to combine the features of the Iron Ultimate Sequencer and the Negotiator Ultimate Sequencer, with a similar “I don’t want to change again” value overwrite at the end of the negotiating sequence and create an Iron Negotiator Ultimate Sequencer were defeated for legal reasons.
Now the models we have described to you so far are just a fragment of the literal millions that were created, chosen to be representative and give you a flavour of the ideas. It was getting out of control. The consumer was suffering from too much choice, thus the MetaUltimate Cupboard was created. You would step inside the Cupboard, and then inside the Cupboard you would find another Cupboard- your own concept of the best possible self-improvement cupboard, a cupboard that would improve you via whatever process you thought was best. Of course all cupboards in the ultimate cupboard line improve you in whatever attributes and regards you think best, but the MetaUltimate cupboard goes through your preferred self-improvement procedure. If, for example, your ideal procedure was “I drink tea at a tea ceremony, and then make my decision on what attributes to improve, no reiterated steps” that would happen.
Of course it wasn’t long before a MetaMetaUltimate Cupboard was created which would allow you to undergo your ideal process of choosing your ideal process for choosing your ideal of self-improvement cupboard. So for example:
Janus enters the MetaMetaUltimate Cupboard. Inside he find a recreation of the Hanging gardens of Babylon in which he talks with three reconstructions of ancient sages, three reconstructions of friends and three reconstructions of artists. On the basis of his reflections during this experience he decides that the ideal way to choose how he should be improved is to have a meeting with everyone he’s ever been close to, the meeting is duly convened with (reconstructed) personas and finally Janus makes up his mind- he’s just going to go through the Negotiator Ultimate Sequencer except it’s a variant where the discussion happens over several days in, let’s say, a Bamboo forest.
Of course, someone soon observes that since the MetaUltimate Cupboard allows you to undergo any process you like of self-improvement (presumably with some limit of computation), and nothing in that rules out reiterated processes- the MetaUltimate cupboard can also serve as a MetaMetaUltimate Cupboard and even a MetaMetaMetaUltimate Cupboard making the production of these models unnecessary . Several people are chided for the waste. The MetaUltimate cupboard reigns supreme then because it can go through any self-improvement process, but its very generality means the mystery remains: Just what is the perfect self-improvement process? How should you prefer to be improved? Presumably, your own preferences enter it the question somehow, but this all gets very complex, because you have preferences over preference formation and process, and preferences over processes of preference formation processes and so on, and questions of dynamic consistency and so on…
I have a feeling that even in terms of our own desires what a more perfect version of us is, is radically underdetermined. There is no unique answer to the question “what a perfect Judith looks like, from Judith’s point of view”.
That's pretty fun idea. I wonder how people's intuitions differ on this, but for me, I feel like it is okay if A, in a chain A->B->...->Z doesn't approve of Z. It still feels like it's an improvement. It also kind of resembles more how people in the "real world" work, continually self-improving themselves by their current definitions of self-improvement only.
I could be millions of diamondoid bush robots, woohoo!