Archives, page 7

[Past]
Post-move
Date: 2018-Apr-14 19:08:48 EST

During the years I blogged little and G-plussed much, I kept a journal. I find myself using it a lot less as this comes to take on some of its function. Much of what I write is organising my thoughts and imposing meaning on my existence. If that's useful or interesting to others in some fashion, great. If not, oh well.

I'm pretty much entirely moved now; my old apartment just has dust and a vacuum cleaner that are eager to meet. I might swing by there later today to do that and then take the vacuum cleaner over to the new place. The old landlord is being uncommunicative; I may have invited mild drama in giving them phone rather than paper notice, albeit early. And then late paper notice, wherein I refused to get my intent to move out notarised. The new place is still slowly being unpacked, and I have an excess of boxes, but this is a big part of how I've chosen to live my life; given an annoying task I add a habit to my life to chip away at it, rather than sprinting.

I feel bad for the people trying to land bpfilter modules into the Linux kernel, significantly because they wanted to introduce a very specific capability, and people realised that it might be better to grow the idea into a general facility, and work out what that facility looks like. It's probably the right engineering decision, because it avoids a kernel full of small similar things, but to be in the shoes of that would-be author is not a happy thing. But so it is in engineering.

I'm amused at this idea and identity of digisexuality. There's an automatic inclination I think for people, single among billions, to attempt to find new ways to distinguish themselves, and craft terms and identities to do so. Maybe part of a storyteller instinct. Even when the results are ridiculous. Not that the need is, but the crafting of thick identities out of narrow specifics, and then having done that the need to backfill a lot of content into something that barely needed a term to begin with. We can and probably should just ignore those efforts, denying them the attention needed to catch on.

I'm weirded out to see that some tech companies say blatantly false things in their SEC filings. In the above link, the claim that "A Database is at the Heart of every Application" is nonsense. Tempting to start a project digging up more nonsense that makes its way into Edgar.

I've written elsewhere recently on forms of Feminism and where I see myself; about to wrap up the topic for which this has been a good map. As noted before, I consider myself a Classical-liberal feminist (and gender-role abolitionist) with some liberal reservations. I think the majority of concerns of gender come down to public or private coercion, I don't care to adjust any biological differences that might exist, but I have a small value for groundbreaking levels of representation, in order to help pave the path for people outside whatever statistical distribution they belong to's norm. And in education I accept more invasive rebalancing because of how necessary it is for everything else.

I think there's been a lot of bad commentary on Charles Murray, but this, which claims that his works suggest bad policy, is on. The bad commentary remains my primary concern, as it's drawn a lot of attacks from anti-free-speech progressives, and I think we need to be able to talk about anything, even bad ideas. Plus I remain generally neutral on the scientific content on Murray. When Murray starts talking about libertarian politics (And he does), I think that's where we should point it out and push back hard (although, again, in appropriate ways that don't stop him from talking or block him from having a willing audience). On the article's discussion of substantitve versus opportunity equality, if we talk about an absolutist version of the former, then I advocate a mix. But if common notion of the former is already mixed, then I might fit into that camp.

This makes some interesting statements linking role-playing to AI training. Resampling in statistics has already done this in a sense, and a lot of CNN-training already involves things like this as well. I think this is good advice people are already doing, and they're already looking to do the obvious extensions of it into broader realms of application. Still, for people less-familiar with the field, it's an interesting read.

It's best not to take hard stances on what happened from a single account, but this story about the early days at Tinder and Bumble sounds like some seriously unpleasant mess. And a reminder that just because a relationship is consensual, that's just a first bar on whether it's appropriate. The sharpest argument here is that even a consensual relationship can make both people incapable of being fair to each other in other contexts, in ways where their mutual fair treatment is obligatory. Like at work.



Justice as contingent results
Date: 2018-Apr-18 22:49:43 EST

This demonstrates the failing of some ideas that both liberals and conservatives occasionally flirt with - that it's valid to aim for equality of outcomes from our legal system, rather than equal treatment under neutral laws. Locally this often comes in the form of attacks on Broken Windows policies, on the theory that such laws have a disproportionate impact on lower-income neighbourhoods. The thought being that we should empathise with criminals on these topics, from turnstile jumpers to breaking of windows not because it's enforced unevenly (or that the laws have too harsh a penalty), but merely because of the impact.

I can have some sympathy for the other arguments, but bundling an impact argument in sours me on any further discussion. And likewise with the linked article above with people upset that Trump's people are being prosecuted and there are not many people on the other side being prosecuted. If there were legitimate concerns about selective enforcement narrowly tailored around Trump, sure, that would be unacceptable, but the mere fact that Trump's people get more focus plausibly can (and likely does) come from the fact that they're breaking various laws a lot more. And if that produces varying outcomes, it should. Attempting to rebalance based on that is inappropriate.



Denying the Power of Words
Date: 2018-Apr-20 02:07:25 EST

There's a beautiful paradox in modern (and perhaps historical) American society, and certain philosophical trends in European society that influenced us. Words have a lot of ability to affect things; en masse they decide if a leader is effective or weak (particularly in monarchies), they can push on our emotions both directly and in how we define things, en masse they help establish or destroy notions of legitimacy (gaslighting is direct manipulation of this facility to destabilise somebody's worldview), and they help shape our well-being, our feelings of safety and its lack, and all sorts of other things.

A direct, full-throated recognition of this is a powerful reason to drop a lot of constraints on speech, whether done as laws, as expectations of violent or career-altering responses, or as social norms. We can prevent all kinds of harms this way. It also, depending on how we recognise it, leads to ideas of crimes of honour (insult somebody in some ways and they might, in order to preserve their status which is essential to their mental well-being which you probably unjustly decided to step on, need to make things right).

So, in contrast to a lot of other societies still extant in the world, we have decided to go for free speech as a legal and social norm, and when it comes up against all these harms, we decide that the harm done by restricting speech, on whole, is worse than any of the harms that speech can do. We get a lot out of this bargain. Some of us consider free speech a direct good (I do); a good independent of its effects. Some consider this a good bargain even if they see free speech as a means to an end, perhaps noting that while it offers potential for a lot of potential harm, it also lets the targeted person blow off steam, laugh at the person who went after them, and lets them go after them right back (as speech, at least).

The way we resolve that tension then, is to dismiss the harms that can come from speech, and teach our children to try to ignore them (sticks and stones, etc). It's not always simple.



Edge of understanding
Date: 2018-Apr-20 17:31:41 EST

I've been helping with a paper at work, and one of the things it talks about is use of a machine learning technique, applied framewise over time, that will threshold (kinda) if/when data becomes significant enough to become an ROI. I feel I'm on the very edge of an insight that the algorithms in play (a CNN), when viewed stepwise (more or less a 90 degree rotation) might take a very different form and be much more efficiently calculated were they revised towards thresholding rather than independent (but stateful - based on data available so far) invocations. It's a strong instinct, but it's just at the edge of my slippery grasp; if I understood a bit more I feel I could figure out a general transformation of algorithms, or at least develop a means to do so, to realign it along these lines.

A lazy version of this would just be to memoise a lot of the intermediate state.