Skip to content
Notes by Axbom
Per Axbom Akkoma

Folk verkar trötta på digitalisering. Men då kallar vi det AI de nästkommande 10 åren istället. Så blir allt bättre.

Oj vad slut jag är…

🔗 Originally posted on Akkoma

27 Nov 2024, 20:08
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

The algorithm has been in place since 2013. I doubt it’s legal. The defense is “our model doesn’t discriminate"

An inability to understand how algorithms exacerbate inequality, even when not specifically programmed to do so, is one of the most distressing competence shortfalls in the public sector.

@axbom.com: Scoop today in Sweden's daily SvD showing how women, especially with foreign backgrounds, are disproportionately incorrectly marked as 'cheaters' by the Swedish Social Insurance Agency, when they apply for temporary parental benefit (vab). Yes, by an automated system. www.svd.se/a/Rzmg9x/for...

View original post

🔗 Originally posted on Bluesky

27 Nov 2024, 19:37
Share
LinkedIn Bluesky Email
Per Axbom Bluesky reply

↩ Replying to @arn.is

Definitely. I like how an enough mentality encourages me to shout ENOUGH! at bad stuff. 😜

🔗 Originally posted on Bluesky

27 Nov 2024, 16:26
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

Instead of a GROWTH mentality I've been cultivating an ENOUGH mentality.

It's friggin' liberating.

🔗 Originally posted on Bluesky

27 Nov 2024, 16:15
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

Some tasks handed over to automation should never have been tasks in the first place. Or should at least be revised regularly. But instead of questioning and reinventing tasks, many will now run perpetually.

And with no human involved, why would someone question its validity?

per.ax/auto

Automation will resist change

🔗 Originally posted on Bluesky

27 Nov 2024, 16:08
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

Scoop today in Sweden's daily SvD showing how women, especially with foreign backgrounds, are disproportionately incorrectly marked as 'cheaters' by the Swedish Social Insurance Agency, when they apply for temporary parental benefit (vab).

Yes, by an automated system.

www.svd.se/a/Rzmg9x/for...

Hemlig AI såg oskyldiga kvinnor som vab-fuskare

🔗 Originally posted on Bluesky

27 Nov 2024, 07:51
Share
LinkedIn Bluesky Email
Per Axbom LinkedIn

"Systemet pekade gång på gång ut oskyldiga mammor och föräldrar med utländsk bakgrund, låg utbildning eller inkomst.

Kvinnor drabbades särskilt. De fick mycket oftare den högsta risknivån, samtidigt som hälften av dem var helt oskyldiga.

Och trots att kvinnor och män stod för ungefär samma andel fel i verkligheten så pekade systemet ut (konstaterat oskyldiga) kvinnor som sannolika fuskare mer än dubbelt så ofta som (konstaterat oskyldiga) män.

[…]

Vi ser också att systemet var sämre på att hitta män och rika personer som faktiskt gjort fel."

Om och om igen är det de mest utsatta som drabbas av osynliga automatiserade beslut som få (om någon) har insyn i och förståelse för. Och det är är oerhört svårt för en individ att få rätt - man vet ju inte ens varför man felbehandlas. Och det kanske inte handläggaren vet heller.

När det dessutom hänvisas till sekretess och systemen inte får granskas, ja vad lever vi i för samhälle då?

Bara systemet är så kallat "effektivt" så tycks det spela mindre roll idag om system är orättvisa. Effektivt på riktigt är det förstås mest för de redan gynnade - och för makthavaren.

Det är, hör och häpna, faktiskt inte supereffektivt för de som blir oskyldigt utpekade.

https://www.svd.se/a/Rzmg9x/forsakringskassans-ai-for-vab-fusk-granskade-kvinnor-oftare

Läs mer om osynligt beslutsfattande i elementen inom AI-etik:

https://axbom.se/aielementen/

"Ju mer komplexa algoritmerna blir, desto svårare är de att förstå. I takt med att fler är inblandade, tiden går, integrationer med andra system görs och dokumentationen är felaktig, desto längre undgår de mänsklig förståelse. Många företag kommer att dölja egen kod och undvika granskning, och ibland själva förlora hela bilden av hur koden fungerar. Man kommer inte kunna förklara hur eller varför vissa beslut tas. Att avkoda och förstå hur beslut fattas kommer därmed att vara öppet för oändligt färre personer.

Och det stannar inte där. Detta påverkar också människors autonomi. Genom att dölja beslutsprocesser (hur, när, varför beslut fattas, vilka alternativ som finns tillgängliga och vilka personuppgifter som delas) är det allt svårare för individer att göra väl underbyggda val för sitt eget bästa."

Försäkringskassans AI för vab-fusk granskade kvinnor oftare

🔗 Originally posted on LinkedIn

27 Nov 2024, 07:18
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

Ah… Sweden. The committee on AI has released their report to the government today.

The general feeling is a recommendation to go all in on a specific yet ill-defined technology without asking if it’s the right tool for the job.

But they do seem to be very certain that it’s all very urgent and we will all be left behind if we don’t act now.

I always wonder how far behind it’s possible to be left. It’s not like there’s a scale, or a point of reference. But damn, it’s urgent. We should essentially be panicking. You know, because those are the perfect conditions for well-reasoned decision-making.

I keep reading the phrase “task force” in the report (which is entirely in Swedish, and that’s really the only action point in English) and all I can think is Space Force.

Anyway, here’s the report: https://www.regeringen.se/contentassets/11431c781ff746b3b7a49312fa4de624/ai-kommissionens-fardplan-for-sverige.pdf

Steve Carell as General Mark Naird in the Netflix series Space Force. The show is a workplace comedy-drama series that centers on a group of people tasked with establishing the sixth branch of the United States Armed Forces, the United States Space Force. Season One follows the efforts of General Mark Naird (Carell) to get

🔗 Originally posted on Akkoma

26 Nov 2024, 22:11
Share
LinkedIn Bluesky Email
Per Axbom Bluesky reply

↩ Replying to @alexanderskogberg.com

Det där är ändå ett förbaskat bra rim. 😁 👏

🔗 Originally posted on Bluesky

26 Nov 2024, 17:33
Share
LinkedIn Bluesky Email
Per Axbom Akkoma reply

@mkj I tend to agree. Hopefully this will boost my blogging. 😁

🔗 Originally posted on Akkoma

26 Nov 2024, 14:35
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

Considering moving back to Mastodon. I love Akkoma, but I don’t have the time to maintain my installation.

On my Cloudron installation, Mastodon would be essentially maintained for me.

What saddens me most is that despite all the marketing of the ease of switching instances I will again lose my history of posts on my timeline (and there’s that one time I lost a few thousand followers).

Thankful I have a backup on micro.blog though.

This also means my posts can’t be this long anymore (Cloudron is kinda locked in that way) but maybe that’s a good thing, haha.

🔗 Originally posted on Akkoma

26 Nov 2024, 14:00
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

Får jag ett AI-genererat julklappsrim så kastar jag paketet i brasan.

🤖→📝→🎁→🔥

🔗 Originally posted on Bluesky

26 Nov 2024, 10:25
Share
LinkedIn Bluesky Email
Per Axbom Akkoma reply

@kallekn ❤️😂

🔗 Originally posted on Akkoma

26 Nov 2024, 10:14
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

Får jag ett AI-genererat julklappsrim så kastar jag paketet i brasan.

🔗 Originally posted on Akkoma

26 Nov 2024, 10:12
Share
LinkedIn Bluesky Email
Per Axbom Bluesky reply

↩ Replying to @handle.invalid

Certainly! It's this one: "Creator labeler".

bsky.app/profile/crea...

https://bsky.app/profile/creatorlabeler.bsky.social

🔗 Originally posted on Bluesky

26 Nov 2024, 07:27
Share
LinkedIn Bluesky Email
Per Axbom Bluesky reply

↩ Replying to @loukas.bsky.social

I played squaredle.com 11/25:
27/27 words (+2 bonus words)
🔥 Solve streak: 1

Squaredle

🔗 Originally posted on Bluesky

25 Nov 2024, 20:56
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

You may have seen this quote before. The quote is from @mollycrabapple.bsky.social and it appears in this column by @bcmerchant.bsky.social:

www.latimes.com/business/tec...

And it's still true.

“There is no ethical way to use the major AI image generators,” Crabapple says. “All of them are trained on stolen images, and all of them are built for the purpose of deskilling, disempowering and replacing real, human artists. They have zero place in any newsroom or editorial operation, and they should be shunned.”

🔗 Originally posted on Bluesky

25 Nov 2024, 20:04
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

I’m not surprised they don’t really care. I was a bit surprised they put it this bluntly.

“To make our lawyers happy:"

Screenshot from a user screen in a Microsoft application where more information is requested from the user (name, email and location).

A box is preselected alongside the text “I would like to receive information, tips, and resources related to Microsoft developer tools and services, including Azure DevOps,Visual Studio Subscriptions, and other Micrososoft products and service. Plus a link to a privacy statement.

Below this is a Continue button.

And below this, with an arrow added to the screenshot to draw attention to it, is the text:

“To keep our lawyers happy:
By continuing, you agree to the Terms of Service, Privacy Statement, and Code of Conduct."

🔗 Originally posted on Bluesky

25 Nov 2024, 12:29
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

This is very clear, succinct and spot on. I immediately support these principles.

@thewhynotlab.com: AI regulation will not protect #fundamentalrights despite claims of the opposite. Read my thoughts on how we could rethink #digital regulation to truly center #humanrights & ultimately #democracy 👇 www.thewhynotlab.com/blog/rethink... Find Holistic AIs reg tracker here: tracker.holisticai.com/feed

View original post

🔗 Originally posted on Bluesky

25 Nov 2024, 10:30
Share
LinkedIn Bluesky Email
Per Axbom Bluesky

When I made my poster "If AI was like a hammer" I got some pushback for 'exaggerating' how much more energy intensive it is than using other tools for similar results.

When it comes to overall climate impact I think I may have been soft-pedaling.

www.thetimes.com/uk/technolog...

‘Thirsty’ ChatGPT uses four times more water than previously thought

The poster "If AI was like a hammer" and its full explanations are available in this post:

axbom.com/hammer-ai/

If a hammer was like AI…

🔗 Originally posted on Bluesky

24 Nov 2024, 11:38
Share
LinkedIn Bluesky Email
← Newer Older →
  • Notes by Axbom
  • RSS

© 2026 Notes by Axbom