@david_rysk

276 Followers
1.9K Following
250 Posts
Trump quietly shutters the only federal agency that investigates industrial chemical explosions

The Chemical Safety Board, a federal agency that investigates large-scale industrial chemical disasters, will close under Trump's orders.

Grist
Many people don't realize that gatekeeping in open-source communities often also comes in the form of delayed or absent reviews by maintainers. Many times, this is a power play by maintainers to discourage others from expressing their opinions. I've repeatedly observed that when maintainers approach technical discussions as authoritative figures rather than collaborative participants, it ultimately harms the project itself.

Did you know that it is possible to create a fully functional rootfs from an OTA update?
Since the AppleTV HD is so similar to the HomePod 1st gen, we can just use the atv restore ramdisk for the HomePod.

We made a script to create a custom ipsw for the HomePod (using atv ipsw as donor for ramdisk), which you can restore from pwnDFU (using regular idevicerestore) to unbrick your homepod from a software update failure!
Check it out here https://github.com/tihmstar/homepodstuff

GitHub - tihmstar/homepodstuff

Contribute to tihmstar/homepodstuff development by creating an account on GitHub.

GitHub
Huge thanks are due to an anonymous benefactor who had a broken IN1606, which runs the same firmware. The poor thing had been drowned by a leaking A/C pipe and was very much the worse for wear... so of course, they did what any respectable hacker would do.
Tore the CPU off and traced the wiring back to the 20-pin connector, hoping to find JTAG, but finding something better.

Sometimes you just don't have the right connector, none of the alternatives you have will fit, and even your smallest crimp terminal is too big.
So you shove a piece of wire over the pin.
And that, friends, is how I got access to U-Boot on an Extron IN1608 HDBT, dumped the NAND flash, and fixed the corruption which was stopping it from booting πŸ₯³ πŸŽ‰ #ElectronicsRepair

Would you have guessed there was not just a UART, but also an MMC/SD interface on the diagnostic connector? I didn't.

LCM is dead
https://www.geekwire.com/2024/seattles-living-computers-museum-logs-off-for-good-as-paul-allen-estate-will-auction-vintage-items/

and to no ones surprise
livingcomputers.org is gone

will they be nuking the github projects too?
git clone while you can
https://github.com/livingcomputermuseum

Seattle’s Living Computers Museum logs off for good as Paul Allen estate will auction vintage items

Living Computers Museum + Labs on First Avenue South in Seattle. The nonprofit closed in 2020 just before the pandemic and never reopened. (GeekWire Photo

GeekWire
In case anyone's still using v1 or v2 of GitHub actions's upload-artifact/download-artifact: they're being retired in six days. Make sure to migrate your jobs now if you can. https://github.blog/changelog/2024-04-16-deprecation-notice-v3-of-the-artifact-actions/
Deprecation notice: v3 of the artifact actions Β· GitHub Changelog

Deprecation notice: v3 of the artifact actions

The GitHub Blog
Lol β€œbias automation” is accurate
I'm doing a talk on DIY cable TV at @emf next weekend, if you're into #AnalogCable TV please feel free to pop by, or find me in the ECHQ (EMF Hams) village :)
https://www.emfcamp.org/schedule/2024/203-being-elliot-carver #EMFcamp
Being Elliot Carver: How to run your own Cable TV station (or network)

Electromagnetic Field
Oh no. Now that's a true horror game concept. https://store.steampowered.com/app/2800420/Publish_or_Perish/
Steamで40% OFF:Publish or Perish

Publish or Perish is a game that simulates working in academia. In order to keep your job, you must publish scientific work. Reviewers are here to get you and all you can do is try to survive. Can you publish enough papers to enjoy tenure? (Contains practical advice for academics)

Γ—
Lol β€œbias automation” is accurate
@kangaroo5383 it sure is, and jeeezloueeeeez that result that the person is replying to…!
@kangaroo5383 Bias propagation and amplification.
@kangaroo5383 original tweet is from 2021 and answer from 2017? πŸ€”
https://x.com/hatr/status/1361756449802768387
hakan (@hatr) on X

Colleagues of mine analyzed A.I.-based job interviews. The software promises to be able to detect personality traits and be "faster, but also more objective". Turns out: Just placing a bookshelf in the background, changes the results significantly. https://t.co/G9yx0abkfb

X (formerly Twitter)
@klorydryk why does it have to be an answer? Pretty sure it’s broadly applicable
@klorydryk @kangaroo5383
They aren't actually relatedβ€”it's not an answer. Someone paired the two posts together thematically.
@artemis I do understand now :)
@kangaroo5383 Can’t wait to see AI ranking candidates based on the volume of their skull. πŸ’€
@meduz @kangaroo5383 Brb, moving closer to the camera in order to win the job interview
@wh0sthatd0g β€œAs an AI, I can only attest the candidate lacks of honesty by trying to game the evaluation, which is an expected behaviour from people with a smaller skull. Thinking outside of the box disorganises corporates. I filed a police complaint for attempt to create chaos, the candidate is probably already in jail now and thus not available for the position.”
@kangaroo5383 what's fascinating is they accidentally made a tool to measure their biases, allowing us to reverse engineer factors they actually consider, even if they don't realize it themselves.
@iris @kangaroo5383 Yeah. What *I* found interesting is that I noticed my initial β€œgut reaction” to the pictures mirrors the AI result. Seeing someone in front of a bookcase makes me feel more relaxed and comfortable with them vs a blank wall. In a real interview situation i’ve learned to use my initial reaction more for normalisation self-diagnostic purposes than as a diagnostic of the candidate.
@kangaroo5383 I'll have a side of accountability circumvention with my deception engine
@kangaroo5383 Without a doubt, the same conclusions would be met if the test was with / without glasses.

@kangaroo5383

Imagine a future in which your access to income is contingent on your ability to perform β€œagreeableness” for an LLM algorithm.

@HeavenlyPossum @kangaroo5383 Sobering. Gonna be the same for getting rentals soon, if not already? πŸ₯΄

@kangaroo5383
May you please post the original link?
Only the first part web.br ... is visible.

Thank you πŸ˜€

Objective or Biased

An exclusive data analysis by BR (Bavarian Broadcasting) data journalists shows that an AI for personality assessment can be swayed by appearances. This might perpetuate stereotypes

BR

@vdaron @JFG @kangaroo5383

Most people seem to be in denial that "Machine Learning" is fundamentally and unavoidably "Bias Learning."

And having learned bias, it's automated and legitimized.

@kangaroo5383 this is why I am considering wearing ivy league hoodies from schools I didn't attend
@kangaroo5383 While it's definitely a much lesser issue than the other biases under examination, here, I think it's worth noting that any computer that assumes *less* neuroses in the person with the extensive library clearly doesn't know a lot about people who read
@kangaroo5383 at least they are using OCEAN and not Myers-Briggs
@kangaroo5383 who is that guy, I want to cite properly?
Fairness oder Vorurteil?

Eine exklusive Datenanalyse zeigt: Videosoftware, die mittels KΓΌnstlicher Intelligenz die PersΓΆnlichkeit von Bewerbern einschΓ€tzen will, ist fragwΓΌrdig

BR
@KathrinM @muellerwhh @kangaroo5383 danke fΓΌr den link. 3 jahre alt, da hΓ€tte ich noch lange die timeline durchscrollen kΓΆnnen ;)
@KathrinM @kangaroo5383 I was looking if the term "bias automation" was old. I find it witty and to the point. So I would be happy if I could cite another source than an anonymous user on X.
@kangaroo5383 I think this software was funded by discrimination lawyers. Imagine how much money they are going to collect from the companies that use it!

Source for the upper post:
web.archive.org/web/2021021619…

Direct link to the interactive site presenting the work:
web.br.de/interaktiv/ki-bewerb…

The @fasterthanlime quote is from its own independent context:
web.archive.org/web/2018092807…

@muellerwhh
@JFG
@klorydryk
@kangaroo5383

hakan on Twitter

β€œColleagues of mine analyzed A.I.-based job interviews. The software promises to be able to detect personality traits and be "faster, but also more objective". Turns out: Just placing a bookshelf in the background, changes the results significantly. https://t.co/G9yx0abkfb”

Twitter

@kangaroo5383 See also "empiricism-washing"

https://pluralistic.net/2023/07/26/dictators-dilemma/

> This is what Patrick Ball from the Human Rights Data Analysis Group calls "empiricism washing": take a biased procedure and feed it to an algorithm, and then you get to go and do more biased procedures, and whenever anyone accuses you of bias, you can insist that you're just following an empirical conclusion of a neutral algorithm, because "math can't be racist."

Pluralistic: The surprising truth about data-driven dictatorships (26 July 2023) – Pluralistic: Daily links from Cory Doctorow

@kangaroo5383 yikes! I wish I could say I’m surprised

@kangaroo5383

"Bias automation" is fantastic!

@kangaroo5383 relavant: during the (ongoing) pandemic many public experts appeared on tv remotely from their own homes and this twitter account rated their bookshelves https://x.com/bcredibility?lang=en
Bookcase Credibility (@BCredibility) on X

What you say is not as important as the bookcase behind you.

X (formerly Twitter)
@dominic LOL that’s amazing… but what if my bookshelf is also my liquor cabinet x bare ware πŸ€ͺ
@kangaroo5383 I think it really depends on what types you have and how that goes with the literature on display
@kangaroo5383
I've done a number of interviews with a guitar hung on the wall behind me and I'd love to know if that improves my rating.
@kangaroo5383 It's the latest version of screening employees with a Myers–Briggs test. It's about as scientific and meaningful. But that's less important than being cheap and automatable, which is vital when screening large numbers of candidates.
@kangaroo5383
What you wear is supposed to be the most important thing in job interviews with real people so ...
@kangaroo5383
Would be fun to get access to that software and optimize to get the best background

@lilacperegrine @kangaroo5383 It would be especially cool if you could make it real weird like those "place a sticker exactly here on the 50 km/h sign and the car thinks it's a little child" hacks.

Like, the HR software thinks you have great leadership skills because of the mauve and teal pentagram in your lower left. Technomagic!

@clacke @lilacperegrine or those anti surveillance ware that makes you invisible / or super great leadership skills must hire 🀣
@kangaroo5383 well, the machine learned the bias automatically from the input data, the developer didn't need to put his own bias in the system.

@gkrnours What you are missing is that someone decided the criteria for inclusion in the training set.

@kangaroo5383

@clacke @kangaroo5383 @clacke @kangaroo5383 I might have phrased my point awkwardly. Both names are different viewpoints on the same device. Machine learning is what it is and Automate bias is what it does.

funny thing, I looked up Merriam-Webster for the definition of machine and it's not about automation, it's about repeatability. A specific learning machine should always extract the same bias from a dataset. Which is nice for finding bias if you want to remove them

aspragg (@aspragg@ohai.social)

@kangaroo5383@mastodon.social See also "empiricism-washing" https://pluralistic.net/2023/07/26/dictators-dilemma/ > This is what Patrick Ball from the Human Rights Data Analysis Group calls "empiricism washing": take a biased procedure and feed it to an algorithm, and then you get to go and do more biased procedures, and whenever anyone accuses you of bias, you can insist that you're just following an empirical conclusion of a neutral algorithm, because "math can't be racist."

ohai.social
@kangaroo5383 thanks for reminding of that. It's probably a good argument for focusing on the output, reproducing bias, rather than the method.
@kangaroo5383 @BirdFriendCoral The Bookcase Credibility twitter account was on to something, apparently.