free_rms

11 ms·

61.
by free_rms 6y ago
Teaching is functional. Performative adherence to other people's politics is less so.
62.
by free_rms 6y ago
Another way of saying it is that if a paper/technique is badly missing some interesting ethical implication, that's a great invitation for someone else to publish and set us all straight.
63.
by free_rms 6y ago
I wasn't talking about the actual researchers, I was talking about the 'ethics' people insisting on these sections (who may also be actual researchers). I would expect most of the papers who have to fill in some impact statem
64.
by free_rms 6y ago
Set aside the details, it's clear that the definition of "ethics" is "things which are currently fashionable in US coastal universities". The fact that many of these technologies could be used to steer a cruise miss
65.
by free_rms 6y ago
There's also * this generation of language models leaning into transfer learning reducing the total number of training runs for different applications * TPUs being more power efficient than GPUs (the numbers they used in the paper were
66.
by free_rms 6y ago
They'd probably want to finess it a little bit more than that, but yeah? If you're going to bring carbon emissions into it, mentioning current mitigations as well seems relevant unless you're writing an op/ed piece.
67.
by free_rms 6y ago
The tone of parent comment was "Google should tolerate research that's critical of Google", which I'm sympathetic to. My rejoinder was that if it's inaccurately critical of Google, like hyping carbon impact withou
68.
by free_rms 6y ago
What if it's an inaccurate negative light? The Gebru paper makes a lot of hay about carbon impact, but Google's put a ton of effort into mitigating that, they've been carbon neutral on paper since 2007: https://
69.
by free_rms 6y ago
I don't need credentials to read a paper which muses about current events. You want to outsource your thinking to the credentialed? Go ahead but the rest of us don't have to. The paper is out there and it's quite easy to re
70.
by free_rms 6y ago
Easy, crank inflation back up. Coincidentally, we've got a covid relief bill, and you can see what the wealthy and power fear above all else -- blanket grants to the people that make them less beholden while increasing inflation. Nobod
71.
by free_rms 6y ago
It's fascinating how nobody can say how this paper was such an ethical stab in the heart for Google, it's all about process. The paper's out there! You can read it!
72.
by free_rms 6y ago
Yeah, the way it was handled sounds like it came across someone's desk at the last second and I can understand frustration on her part.
73.
by free_rms 6y ago
Being told "you're multiple years behind the state of the art on this, please reference it" is a reasonable ask.
74.
by free_rms 6y ago
What was she blowing the whistle on specifically? Carbon emissions from a company that's been "carbon neutral" on paper for a decade?
75.
by free_rms 6y ago
The paper she was fired over was leaked, though, and it's.. not novel. It's the same things everyone else has been saying for years (global warming bad, also don't build a racist speech reproducer), and Google wanted her to
76.
by free_rms 6y ago
There's no certification for being an Official Journalist and that's a good thing.
77.
by free_rms 6y ago
But your caveats mean that, in practice, they can come after anyone who actually came into contact with the material unless it was dead-dropped to them without their knowledge, and even then, maybe they were 'encouraging' it. See
78.
by free_rms 6y ago
Yeah, but those horrors made people rich. Cutting emissions makes nobody rich. Developing fusion, carbon recapture schemes, we need to be thinking in terms of boondoggles that get the job done.
79.
by free_rms 6y ago
Singlow isn't talking about blind both-sidism, though, they're talking about skeptical reading of both (or many!) sides to zero in on the truth.
80.
by free_rms 6y ago
Cutting emissions by regulation is a continually losing battle. You can't tell people to live a lower-quality life and expect to win elections, and you'll have the combined lobbyists of a lot of different interests against you. It
81.
by free_rms 6y ago
God I wish we could Manhattan Project this shit. I'm a layman too but I get the impression that the theory is basically there and we just need to figure out the engineering. That's a solvable problem with lots of money and lots
82.
by free_rms 6y ago
This is a cool example of how difficult life can be if every figure of speech is interpreted to try and find offense, somehow, whether historically accurate or not. I agree with sister comment that the English history probably informs the e
83.
by free_rms 6y ago
There are all sorts of things that people broadly agree are offensive that can be or have been produced by ML models. Why not stick to those and discuss mitigation strategies? The paper could be vastly improved and still make it's po
84.
by free_rms 6y ago
I'd fire her in response, it's still a firing. This technicality matters for unemployment and COBRA, and it has a concrete definition.
85.
by free_rms 6y ago
> The answer to this mystery was obvious, but only in hindsight. There are far fewer black people in FFHQ than whites or asians. I found a classifier, got it to identify an order of magnitude more black faces than I had before, retraine
86.
by free_rms 6y ago
Ok, fair enough, the more sophistication you add, the less explainable it gets. That being said, you can look at magnitude and direction of various coefficients and conclude certain obvious things. Even the case where reversing the correla
87.
by free_rms 6y ago
One way to define politics is "difference of opinion over ethics".
88.
by free_rms 6y ago
Yes, a linear regression is explainable. You know what your features represent, and the regression puts a straight-up co-efficient on each of them. You can look at the weights you trained, see how strongly your input features contribute to
89.
by free_rms 6y ago
The paper sounds boring, though. She hits some boring bureaucratic resistance, lost her shit, called everyone racist and gave ultimatums. She chose this hill to die on. If she were uncovering some major ethical scandal, I'd be on her
90.
by free_rms 6y ago
Alright, here's my substantive criticism of how CRT affects various groups in practice (not doxxing my membership in any of these groups): Elite coastal white: Absolutely not threatened. Beneficiary of the system and knows how to nav

More ›