It occurs to me that despite my lack of expertise in alignment, I am really enjoying the mesa optimization conversation because it does a good job of contextualizing machine learning performance. This is good even for narrow applications.
I feel like this is an excellent middle ground between a full review and relying solely on the reputation of the author, and I am excited to see the eventual list of books which pass the epistemic spot checks.
I second this. The architecture of prediction feels like rationality tooling.
Lately short stories, action, and good prose. Short stories are an excellent antidote to the glut of long book series; they don’t allow enough space for fluff, so I find they are consistently better reads. Also lower investment, which is nice. And good prose is good prose, like always.
A year or so ago I read some of Ursula K. Le Guin’s short stories, and that was when I really noticed that there were levels to the whole business. I don’t recall the story, but the scene which struck me was walking down a road in the autumn. I now suspect that depicting banal events well is a mark of craft in the same way as drawing a circle or squaring an edge.
I favor editing directly, except in cases where a new position has been taken. Then I favor a second post, because this will allow us to capture the development of the ideas. This would include any important reversals or entirely new dimensions of the idea.
Putting all subsequent versions of the same post in a sequence would be a good idea; then people who come to one of the posts could see there were previous versions and/or subsequent versions. I wonder if some kind of tag or titling convention would be helpful.
Is this not the default behavior with the RSS? When I edit one of my posts, it shows back up in the daily post queue.
I am strongly interested. I’ve been trying to get some real math under my belt, but to be frank it isn’t my talent so I am perpetually on the lookout for exposition and exposition accessories.
I still have no idea how forcing really works, but man do I like the cut of Chow’s jib.
I’ve noticed I navigate my entertainment largely by things to avoid. I hate coming of age tales in general and anything involving a school in particular. I despise children-of-destiny stories, which is weird because I’ve always liked prophecies. I avoid books when people talk about the worldbuilding.
This strikes me as strange considering how much of my reading when I was young consisted of a child of destiny who comes of age amid crappy worldbuilding. Maybe it is an acquired sensitivity or something.
I support this idea. Especially if it incorporates some motivation for topology, which weirdly seems to hang out by itself until it suddenly becomes critical.
Models are important. They are arguably more valuable than physical assets.
Models often have uses beyond their original intent.
Models are currently ad-hoc in terms of their location and transmission.
Think of them as a kind of IP. Things like patents and copyrights are strategically managed: they are tracked and controlled; they have a lifecycle; they can be combined to generate yet more value. But models are mostly laying around at the department level, the team level, or individual people’s heads. There is no formal, universally recognized system for them the way there is for IP.
Instead, we should curate them:
Model curation is the lifecycle management, control, preservation and active enhancement of models and associated information to ensure value for current and future use, as well as repurposing beyond initial purpose and context.
There are several areas of curation which can be drawn on for examples: museum; digital; content; biomedical models. The work recommends making this a specific responsibility at the organizational level. There should be a Chief Model Curation Office at the enterprise level with a team dedicated to the purpose, including officers at the program level.
Models should have a pedigree: the origin, verification, enhancements, and uses over time should all be a part of the model object.
Source: Rhodes, D.H., “Model Curation: Requisite Leadership and Practice in Digital Engineering Enterprises,” 17th Conference on Systems Engineering Research, Washington, DC, April 2019. [paper] [briefing]
I had launch codes. I had hidden the map previously in my settings, which also had the effect of hiding the button, which in turn was enough to screen off any buttons should be pressed and would this really work? temptations.
I did keep checking the site to see if it went down, though.
I asked for posts like this not 24 hours ago, and here one is. It’s a Petrov Day improbable!
Interesting. Am I correct that this implies the larger the house’s cut, the more systematically off we should expect the payouts to be? It seems like PredictIt’s 10% effectively moves the point at which it isn’t worth it to correct the inefficiency further out from equilibrium than in the case of free trades.
I would be interested in seeing more applied fields, and also specializations which operate at the intersection of multiple fields. Some examples include:
Operators, in the sense of people with executive responsibility. I have enjoyed reading the after-action reports from the organizing experiences and foundation-forming to come from this website and EA.
Finance, which is essentially the field of applied distribution of risk. We have finance people on here, but there seems to be little content in terms of top-level posts from them (the easiest way to tell there are finance people present is to look at the top-level finance posts and then look at the criticism in the comments).
Industrial or Systems Engineering, which are fields dedicated to integrating other fields and applied optimization of the group all together.
The adjacent memeplex of Effective Altruism seems to have a bunch of operations and finance people in it.
We might consider trying to target people who are connected to teaching or instruction in their area of expertise somehow. I expect the average level of engagement with a new subject is quite a bit deeper here than in most other communities, so we might be in a position to offer an audience of motivated learners as an enticement to them. Simultaneously, the instruction experience will help with the problem of technical posts having too high a threshold to engage with them.
How do you suppose this compares to the likes of Anki or Mnemosyne?
It’s probably worth emphasizing the huge challenges involved in trying to get a coordination process to output something highly specific. This is enough of a problem to start with using specificity to privately evaluate start-ups; how to get from there, through governments, and then through international relations to a highly specific treaty is a good candidate for the most complicated possible task.
I have heard from several angel investors words to the effect of “I don’t invest in ideas, I invest in people.” Which is to say they prefer a good group of founders with a mediocre idea to a less reliable group of founders with a better one.
This seems similar to your high generic competence standard. The hitch is that the preference for a good team over a good idea doesn’t rest completely on the likelihood with which a mediocre idea will be successfully executed, but rather also on the likelihood that this good team will recognize the mediocrity of the idea and shift to a new one successfully. Quoting from Paul Graham’s essay linked above:
So don’t get too attached to your original plan, because it’s probably wrong. Most successful startups end up doing something different than they originally intended — often so different that it doesn’t even seem like the same company.
I feel like the ability to recognize and then articulate value should be included in the idea of generic competence. Likewise for things like opposition research: following the advertising example, I don’t see why we can’t just recurse on execution advantages with the same basic structure of a story. It is like a Value Sub-Proposition Story, where the specific person is the entrepreneur and the specific problem is delivering on some aspect of the Value Proposition (by getting it in front of people).
It still seems useful to the investor to know whether or not execution advantages are specific and what they may be, and therefore also useful to the entrepreneur to articulate them.
I strongly approve of providing less-formal essays to aid with clarity and intuition.
This post reminded me of a video about a different way to prepare textbooks, which was written in Julia. The book is Algorithms for Optimization. It’s beyond my ability to assess content quality, but it seems to look pretty sharp.