”Företagen vet att det finns ett tillitsproblem, men det betyder bara att de kommer anstränga sig för att få oss att känna tillit. Det är inte samma sak som att minska riskerna med tekniken.” – Marietje Schaake, forskare vid Stanforduniversitetet
@nafnlaus The decision to design them this way is purely human. There are a multitude of human decisions even in unsupervised "learning", when it comes to labelling and modelling. The fact that many of these algorithms are a black box when it comes to predictability is only an example of humans building an error-prone system based on human material.
The machine is also an extension of all human error contained in the unsupervised training material, including historical and representation bias. It's further adding more human input to the mix, not less. All that human labor fed into the machine, and none of those humans visible. Again, it's hard to see the humans (who produced all that content) because of the machines.
Unsupervised training is for example also how Child Sexual Abuse Material has made its way into image generators, and descriptions of abuse and violence into all the text generators. There are people working around the clock across the world to prevent this content from appearing in generated content. More invisible humans doing invisible labor.
I agree they aren't puppets, although puppet features exist in AI as well (deepfake generators). But they are truly not a separate entity from human influence and decision-making. The way I see it, it's all they are. And sometimes that leads to exacerbating human bias rather than minimising it.
Fascinated by the rebuttal that humans make mistakes too, so algorithms should be allowed to as well.
As if the machine is disconnected from the humans building, deploying and using them.
The machines are an extension of human mistakes, not its own separate entity with free will.
It would appear, ironically, that when a huge number of humans are involved in building the machines, the humans become invisible.
Future historians:
"They couldn't see the humans because of all the machines."
#AIEthics #DigitalEthics
"From the generic hipster cafe to the ‘Instagram wall’, the internet has pushed us towards a kind of global ubiquity – and this phenomenon is only going to intensify"
www.theguardian.com/news/2024/ja...
The tyranny of the algorithm: why every coffee shop looks the same
Är jag överdrivet petig när jag reagerar på att ett företag som bedriver kompetensutveckling och utbildning kör en annons i flera veckor som innehåller stavfel?
Annonsen har två korta textstycken och bägge innehåller ett stavfel (eller grammatik-fel beroende på hur man ser det).
Jag tänker att om jag retar mig på detta så gör fler det. Det är förvånansvärt ofta den här typen av fel dyker upp i texter som borde ha granskats av många personer.
Det finns såväl yrkespersoner som tekniska verktyg som kan hjälpa. Men hjälpen uteblir.
Jag blir lite häpen var gång jag ser sån't här. Men det är klart, det drabbar ju mest mig själv. Jag får gå en promenad och fundera över varför jag bryr mig så mycket. 😁
@Caution
This makes me so happy 😊 Thank you for letting me know.
@JMMaok
@Caution @JMMaok
I couldn't resist making a real demo in jsbin:
https://output.jsbin.com/gapapox/
@JMMaok
While I don’t get a chance to code myself as much anymore, the recommendation would be to not style headings but instead apply class names for the visual appearance. Hence one would separate semantic hierarchy (h1…h6) from look-and-feel.
For instance:
<style>
.TitleXL {font-size: 2em}
.TitleL {font-size: 1.8em}
.TitleM {font-size: 1.4em}
.TitleS {font-size: 1.2em}
</style>
In my example then it would be.
<h1 class=“TitleXL”>Site Name</h1>
<h2 class=“TitleL”>Article 1</h2>
<h2 class=“TitleM”>Article 2</h2>
<h2 class=“TitleM”>Article 3</h2>
<h2 class=“TitleS”>Social media</h2>
<h2 class=“TitleS”>Resources</h2>
<h2 class=“TitleS”>Abous us</h2>
<h2 class=“TitleS”>Contact info</h2>
To complicate things, if I was building this page I would likely add a separate heading for the whole footer area. That heading would then be h2 and the four columns would each have an h3. But it was a quick example 😉
@JMMaok
If that is your only visual clue I would agree, but I would presume that there are other design features that indicate how a heading or title is related to the content it refers to.
I made a quick sketch to explain.
In the attached sketch there are three articles but one is displayed as a featured article and the title of that article is visually displayed in a larger font.
If in this case the site title is chosen to be h1 then each of those three article titles would be h2, regardless of their visual size on screen. From the design, it is clear which content element each article refers to.
Moving on to the footer of the site, I have drawn 4 sections, or columns, with 3 links each. The heading of each of those sections is visually smaller than any of the article titles. But in this specific example they would be h2 headings as well, or the hierarchical structure would be off. If in the code they were h3 or h4, they would be interpreted as belonging to article 3, as they would then be subheadings under that article.
But I doubt that anyone looking at this page would interpret the footer headings to be subheadings of article 3, even if they are smaller than that heading. Similarly, the two articles in the center of the page are not easily interpreted as being part of article 1, even if the title fonts of article 2 and article 3 are smaller. The overall visual cues tell the whole story. The code should tell the same story.

@SuperTaliaDX
It’s a fair question. It functionally makes a difference primarily within 1) accessibility and 2) machine interpretation.
In the case of accessibility, the correct heading hierarchy makes it easier to navigate content and understand how different parts of a page relate to each other. With missing or misplaced headings the content becomes more confusing to read with a screen reader*. Jumping between, and listing, the headings of a page are common ways of navigating.
Similary, if machine interpretation is used to determine for example main, supporting and related content it can help pull relevant content to respond to questions and infer importance of one piece of content over another.
One very simple and common example would be an automated table of contents to jump to different sections of a page.
*(The use of a screen reader is of course also an example of machine interpretation)
I interrupt my regular programming to just remind everyone that there is such a thing as h1…h6 headings in html but they are supposed to be used for communicating structure of a page, not visual size.
An h2-heading in one part of your page or site does not have to be the same size as an h2-heading in another part or your page or site. The purpose of the heading is to semantically communicate hierarchy.
In fact, deciding that all h2 headings must be the same size across your website will usually lead to poor decisions with regards to the underlying structure of your page.
What does your design system say about the size of h1…h6?
So yes, the heading in your footer should likely be an h2 but it doesn’t have to be the same size as the h2 in an article. And no you shouldn’t make it an h4 ”because I want it smaller”!
This may be one of the most misunderstood elements of web design.
P.S. And yes you are allowed to have more than one h1 on a page. Yes, really.
#mythbusting
@evilcookies98
h1 is generally the title of the page OR the title of the main article.
h2 will be subheadings
h3 will be subheadings to h2 sections
So you may not be doing it wrong?
What people tend to get wrong is the idea that they can not add h2 below h4. That just means that you are starting a new topic, which is fine. I also see people using h4 in for headings in sidebars (that are about something else than the main article) because they want a smaller font. But since they are new topics, they should often be h2, for example.
So, generally, like this is how one should think (example from the W3C page:
<h1>Plant foods</h1>
<h2>Fruit</h2>
<h3>Apple</h3>
<h3>Banana</h3>
<h2>Vegetables</h2>
<h3>Broccoli</h3>
<h3>Green breans</h3>
<h2>Related articles</h2>
Since this is the internet, someone will soon inform me that bananas are botanically berries. 😊
W3C on headings: https://www.w3.org/TR/WCAG20-TECHS/H42
There has long been a debate about whether you can have multiple h1's on a page, and generally, yes, I would argue that you can. Which Matt Cutts from Google also confirmed 15 years ago:
https://www.youtube.com/watch?v=GIn5qJKU8VM
Let me say it louder for the people in the back… heading levels (h1, h2, h3…) in html are not about size, they are about structure. They communicate the organization of your page.
An h2-heading in one part of your page or site does not have to be the same size as an h2-heading in another part or your page or site. The purpose of the heading is to semantically communicate hierarchy.
In fact, deciding that all h2 headings must be the same size across your website will usually lead to poor decisions with regards to the underlying structure of your page.
So yes, the heading in your footer should likely be an h2 but it doesn’t have to be the same size as the h2 in an article. And no you shouldn’t make it an h4 ”because I want it smaller”!
This may be one of the most misunderstood elements of web design.
#WebDesign
"For instance, in a task measuring the precedential relationship between two different cases, most LLMs do no better than random guessing."
@samuel
En Ghost-installation per domän. Men du kan ha flera författare per installation. Då får varje författare en sådan här sida med alla sina inlägg:
https://axbom.se/author/axbom/
Eftersom jag använder Cloudron kan jag starta en ny Ghost-installation med ett klick.
Så t ex https://digitill.se och https://dataskuggan.se är två av mina sajter också. För elever har jag en med inloggning: https://etik.axbom.se/ och ibland planerar jag bokskrivande på det sättet också: https://debook.axbom.com/
@fiaewald @konstellationen
@samuel @fiaewald
Ghost är bäst. 😁 Jag skrev ju den här långa när jag bytte: https://axbom.com/from-wp-to-ghost/ och sedan om Indieweb generellt.
Allt stämmer inte i inlägget längre men jag svarar också gärna på frågor. Har nog 5-6 Ghost-sajter nu. 😁
De har också en Substack-importer nu, vilket kan vara högst relevant.
@samuel @fiaewald
Obs att den artikeln handlar om en flytt till betalversionen av Ghost, medan jag numera kör Ghost på min egen VPS, så här
https://axbom.com/indieweb/
Och precis när jag skrivit om sanningens flyktighet ramlar jag över Lars Ilshammars tankar om hur vi kan hantera många av dessa utmaningar. Och oron över att arkiven inte förvaltas på ett betryggande sätt.
Det är både viktigt och akut att ta till sig av Lars insikter och uppmaningar.
"Dessutom hjälper arkivet till att läka gamla sår. I de sannings- och försoningskommissioner som nyligen gick i mål har arkivet varit centralt för att ge minoriteter som samer och tornedalingar upprättelse för den svenska statens historiska övergrepp. Det som försvårade arbetet för en av kommissionerna var just brister i arkivhållningen."
"Ingen slump att historiens erövrare som velat kuva ett besegrat folk ofta har börjat med att bränna arkiven."
Really good example of how messed up the future is gonna be. If we reach old age and search for pictures of ourselves, will we choose to doubt all those photos... or will we assume that we did everything depicted?
How will memories of our own lives be captured when our photos and videos are of no help in determining what really happened?
Now is a good time to start on that: Dear diary…
In the story Johan shares is a generated photo of a genuine pumpkin carver in front of a pumpkin he did not carve. And someone else generated this photo of him without consent. The nuances of what is going on here emotionally, socially and ethically are endless.
Engineers be like "Why aren't humans adapting to how my machine works?"
@grimalkina Yay! Congrats, Cat!
@redegelde All but one are available on my website. It will be part of an upcoming post.
Direct links:
https://axbom.com/aipower/
https://axbom.com/aielements/
https://axbom.com/bias-in-machine-learning/
https://axbom.com/hammer-ai/
https://axbom.com/aicom/
They’re also reachable in the featured section of the homepage:
https://axbom.com