jpfed

I'm not even sure where to start, but I have to ask- are there any groups / projects that are trying to make and train an OSS model federated across members' machines?

Someday I'm going to write a paper that achieves SOTA results with a nigh-incomprehensible mishmash of diverse techniques and title it "All You Need Considered Harmful".

These are good examples of mass casualty events that didn't motivate effective responses. For both climate change and guns, there are people that want to keep the status quo as it is. I believe that popular arguments against gun control will be able to stay effective in ways that arguments against climate action will fail.

First, one of the most popular arguments against action addressing climate change has been that climate change is simply fake. A mass-casualty event would make that argument less effective.

The gun control debate in the United States is linked with people's perceptions of personal agency. Often, a mass casualty event perpetrated with a gun is ended by a gun being used on the perpetrator (sometimes, it's the perpetrator killing themselves, and sometimes it's police killing them). Many people imagine it would be possible to end those events as a bystander with a gun. But I think that it would be much harder to get people to fantasize about being a hero that stops a mass casualty event resulting from extreme heat. Such an event starts and ends when Mother Nature says it does, and it seems like it should be pretty clear to people that the only way personal agency is relevant is in the act of preparing the environment.

News media like to have stories they can focus popular attention on, and no one car accident (unless it kills someone famous) will matter to enough people to warrant broader coverage than one's local paper. Despite this lack of media visibility, there actually is some movement at the city level to change the design of roads to reduce the lethality of accidents. This isn't very publicly visible yet, I don't think, and there's certainly very little in the way of public desire to drive less lethal vehicles.

I haven't read the book, but if the event the book describes kills hundreds of people or more, it seems like a plausible seed that news coverage can accrete around.

Tens of citizens' personal information may have been exposed.

Don't get us wrong, eval() is terrible too.

I'm curious about whether anyone has studied using the harmonic mean (or at least, the reciprocal of the sum of reciprocals) as a way of aggregating utilities. I haven't had time to research this, but the thought repeatedly occurs to me as I have kids that are gifted in a school district that is especially concerned with the gap in achievement between the highest and lowest performers. I can't shake the feeling that what they really should want is a measure that prioritizes helping the lowest performers but does not consider the performance of the highest performers to have literally negative value, and the harmonic mean fits that bill.

A group being "above the rules" is, I think, a statement about what the rules are and how they are enforced. It doesn't really hinge on whether any members of that group have, up to this point, violated the rules.

I have long liked and advocated for approval voting (or cardinal methods more generally), or barring that, basically anything but FPTP.

But I'm disillusioned. Many non-plurality methods (including approval) can effectively solve the spoiler problem (i.e. satisfy IIA)... but I no longer think that spoilers are the absolute most important problem to solve.

The most important problem is single-seat districts creating anti-majoritarian, disproportional legislatures. Because of how people geographically sort themselves among the likeminded, even independent districting commissions will not be able to effectively prevent anti-majoritarian legislatures.

For this reason, I believe that we would greatly benefit from explicitly proportional representation. This can be accomplished through cardinal or ordinal means - I no longer care about that dimension. That doesn't mean that we wouldn't enjoy some incremental benefit from approval or really any non-FPTP system. We would. Seattle should go ahead and do its thing. But the fight for better forms of democracy can't stop with any system that retains single-seat districts.

Nazca's include mechanism allows you to separate your code in whatever way you find appropriate.