Wednesday, May 25, 2016

Link dump posts aren't the best but it's what I've got for you now

Global warmingWhat's really warming the world? (yes, it's people.) It's good to have a few solid links on hand for debates I guess. (OTOH there's that phenomenon where showing people facts that contradict them makes them dig in their heels, not think twice... if you get a digging-in-heels response from this, what are the counterarguments though?)

Economics/basic income: Man, I was going to post some news article but as usual I could just post a much better Slate Star Codex post. This one's about how America's not the Land of Opportunity, and basic income might be the only way we might get there, though even that's not a slam dunk. And follow that link to "how bad are things"... yow.

Algorithmic fairnessThere's software used to predict future criminals. And it's biased against blacks.
UGH DUH OF COURSE IT IS... given that most companies cannot get something as simple as "serve a web page" right, of course they will not get something as subtle as "don't make your algorithm secretly hate black people" right. So what's the answer? Keep it as goddamn simple as you can, and when you have an algorithm that's as powerful as that one, at *least* require that it's open source. (This doesn't solve everything; it can be open source and complicated as hell, and open source doesn't mean that anything will ever change, but "ok company, make whatever algorithm you want, we'll use it to determine people's lives without asking too many questions" is not a solution.)
Edit: OOPS I READ THIS TOO CASUALLY; seems their algorithm is fine, or at least off the hook for now. Do stats right, please. Draw conclusions after you do the stats. Don't start off trying to prove something's racist.
Lessons learned from this whole episode:
- internet writers do things that I do not like. So do all of us. (see also: academics, in every single thing they ever write.) In the long run, we should as a society stop doing that somehow. In the short term, it's at least useful to know that this is the world you're in.
- I'm thinking of stopping using the term "racist." It's basically a swear word at this point.
- Please please do not misinterpret what I'm saying in the previous sentence. Not saying we should stop discussing racism, or that people aren't racist, or algorithms aren't racist, or that everyone should stop using the word "racist", or... any number of other stupid things. I just mean that as soon as "racism" comes up, everyone (including me) gets defensive or aggressive and the conversation (which might have otherwise gotten us somewhere) is usually no longer worth having.
- Please let me know if there are unforeseen negative consequences of the above. Also please let me know if you're a POC and find this a bad or hurtful idea.

Design on the web: Relatedly, here's a nice roundup on the subject of things that web and mobile app designers and writers do that I do not like. Noticeably absent: broad claims like "the internet is making us stupid." Present: a solid list of anti-patterns. (or "dark patterns.")

Python and NLP: On a lighter note, the amazing Allison Parrish, one of my main inspirations in creating Swot Perderder, made a sweet python library that would have saved me a solid bit of hassle: Pronouncing.

SF Rent: Eric Fischer is indeed a hero, and various versions of this article about SF rent prices have been creeping across my internet. Dear authors of the "software used to predict criminals is biased against blacks" article: you should be required to be as thorough and humble as Eric if you want to publish data journalism. (well, maybe that is an impossibly too high standard :)

Brains: Finally, and sorry to end off with a difficult one, but this is important, ok, really smart psychologist argues against the information processing model of the brain. It's gotten us somewhere, but at this point it is doing more harm than good. At first I was all like "yeah but what if you could model all the neurons" but his point about the uniqueness of it all is intriguing.

"even if we had the ability to take a snapshot of all of the brain’s 86 billion neurons and then to simulate the state of those neurons in a computer, that vast pattern would mean nothing outside the body of the brain that produced it."

Dang. Ok, that's got me thinking. Not enough to say anything wise about it yet.
Edit: ok, after conversations with friends, and rereading the article, I guess a lot of it is a giant rant by someone who doesn't actually know what computer scientists and cognitive scientists do, and assumes they're dumb. Eh. I glossed over that to get to the pull quote about simulating a brain in a computer, and that's what I took out of it. And I agree that most of the rest of it is pretty dumb; here's an entertaining take down of some of the ranty bits.

2 comments:

Daniel said...

We should talk about that brain article sometime :) I was not impressed! But maybe I just wasn't charitable.

Dan Tasse said...

Yeah, I do want your take on it! Parts of it did read like a guy who doesn't know what computer scientists do, claiming that computer scientists are claiming crazy things. It was mostly the part I quoted (and following) that got me to thinking, huh, maybe you can't actually do this.

(kind of like when I used to think, well heck, the universe is fully deterministic, so you could simulate all the molecules in a giant computer, so what's even the point of wondering about free will etc, until I realized that the computer would have to be wayyy bigger than the universe and so it's not possible. point is, it's worth talking about whether something is possible within the physical universe not just whether it's possible in thought-experiment-infiniteland.)