darawk

88 ms·

841.
by darawk 6y ago
It seems to me that the thrust of this sub-thread is a debate over narratives. One narrative is "America is basically good, but it has serious problems, and we're working on fixing them". Another narrative is "America is
842.
by darawk 6y ago
Nobody is saying "America is great shut up". America has very serious problems, both currently and historically. The point of this piece and the other commenters in this thread in defending America is not that we should ignore the
843.
by darawk 6y ago
It seems to me that's what's happening in this thread is privileged people with no experience of systemic racism telling Filipino veterans how they ought to feel about the country they fought for. The people with the concrete expe
844.
by darawk 6y ago
Some of these seem reasonable, or at least unobjectionable. But 'dummy value' and 'sanity check'? Really?
845.
by darawk 6y ago
Gerry-mandering is pretty race neutral, though. That isn't to say that districts are never carved along racial lines - they absolutely are. But plenty of districts are drawn to disenfranchise white voters, too.
846.
by darawk 6y ago
I think the point is that, though police violence is terrible and should absolutely be corrected, trying to summarize the United States with it is incredibly reductionist.
847.
by darawk 6y ago
That could be. But afaik it hasn't been attempted yet.
848.
by darawk 6y ago
Well, the ethicists just don't understand the models, then. For instance, there are a bunch of measurements you can take of faces to identify people, if you were doing it manually. Things like pupillary distance, canthal tilt, nose wid
849.
by darawk 6y ago
Yep. This is just a few big companies finding a clever way to get free marketing out of a decision to cut their marketing budgets due to the pandemic.
850.
by darawk 6y ago
> And what does this get you? Let's look at a face recognition dataset. What happens when you debias it? Is it still useful? No. Because the faces no longer resemble real faces. Not to you. But you can remove the racial information
851.
by darawk 6y ago
You have it exactly backwards though. Its humans that obstruct accountability, because they are opaque and inconsistent. ML is consistent and inspectable. It cannot lie to you about its motivations. It is infinitely more accountable than an
852.
by darawk 6y ago
> I don't follow this. If we have a biased objective function, the model won't surface any biases we weren't already cognizant of in the objective function. And they were already quantifiable: we had a function that we wer
853.
by darawk 6y ago
What is your argument that it's still bad? That was kind of a non-sequitur. If we have a model that takes in skull measurements and accurately predicts whether you are an introvert or extrovert...what is the problem with that?
854.
by darawk 6y ago
IQ isn't the point though. I'm not a huge fan of IQ specifically either. Your original objection was intrinsic to the idea of detecting things from faces. These objections are to the quality of the measures. I'm not going t
855.
by darawk 6y ago
Sure. But you also need to account for the actual baseline you're comparing to. Whatever process it might be that's taking someone's IQ as an input is right now being handled in some subjective manner by other humans. Do you
856.
by darawk 6y ago
Of course we should be careful. But we should also keep in mind what the baseline is. The baseline is subjective evaluations by other humans. If we don't build models, it's not like we're going to live in a world where nobody
857.
by darawk 6y ago
I know. My point is not about IQ. My point is that people find IQ to be a useful measure of things, in certain situations. We can use something else, like the big 5 personality inventory, if you like. The point is just that subjective chara
858.
by darawk 6y ago
> Ignoring the (many) arguments that IQ is not really a great marker for intelligence, there's basically no value in trying to predict IQ based on facial structure. Unless it is quite literally 100% accurate (which it is not), you w
859.
by darawk 6y ago
> There will be no accurate data for this because traits are subjective. Unlike other subjective predictive modeling, like sentiment analysis, I cannot feasibly think of a use-case for this that is not deeply disturbing. This is a strang
860.
by darawk 6y ago
Phrenology was bad because it was scientifically invalid. I don't see any problem with predicting traits from appearances, at least, not intrinsically if the data is accurate.
861.
by darawk 6y ago
> How do you de-bias datasets, when the raw data (coming from the real world) starts biased to begin with? I can think of at least one trick that might work. Build a classifier capable of identifying race from the data, and then reverse
862.
by darawk 6y ago
I agree that some people appear to believe this. But what is the justification for it? I don't really get how it cashes out in concrete harm, assuming you de-bias the datasets. To be clear, I do get arguments that ML might increase the
863.
by darawk 6y ago
I think this is actually a little unfair. The internet has actually done all the things we hoped it would do. It has just also done some other things that we didn't anticipate. I'm still hopeful that these are basically just gro
864.
by darawk 6y ago
Ya. And even if he was writing on totally un-emotional topics, like a food blog or something, his job is such that patients being able to discover these aspects of his personal life would be likely to pollute his doctor-patient relationship
865.
by darawk 6y ago
This is really sad. I hope SSC comes back. What the NYT is doing here is almost incomprehensibly shitty. I can't imagine why they would think it's so important to publish Scott's real name.
866.
by darawk 6y ago
Of course you can price it however you want. And you do. You are in control of whether and when to give away your data. If you use Facebook, you are trading your data for their services.
867.
by darawk 6y ago
> Seems like this is putting the cart before the horse. Why should we care what Facebook can and can't afford? We don't owe them a business model. The point of this is to remit to the user the value that has been extracted from
868.
by darawk 6y ago
I would love to do that, and it would certainly help, but I don't think it's anywhere near sufficient to pay teachers as much as your average software engineer, for instance.
869.
by darawk 6y ago
He gives two examples in the article.
870.
by darawk 6y ago
The problem, though, is that the number of teachers we require makes that prohibitively expensive. We need to do both things. We need to increase the leverage of individual teachers, so that they can teach more students. To do that, we requ

More ›