Notices by Baldur Bjarnason (baldur@toot.cafe)

“Schools Using AI Emulation of Anne Frank That Urges Kids Not to Blame Anyone for Holocaust”

https://futurism.com/the-byte/ai-anne-frank-blame-holocaust

I mean, what the fuck are we doing?

“Five things privacy experts know about AI - Ted is writing things”

https://desfontain.es/blog/privacy-in-ai.html

All of this has been known for a while, but this one especially:

> Bonus thing: AI companies are overwhelmingly dishonest

Anybody who is writing analysis, predictions, or planning on LLMs and doesn’t take into account the overwhelming dishonesty of the organisations involved is either deluding themselves or you

In certain circles people tend to cite the cognitive impact of repeated COVID infections as an explanation for the rise in popularity of right wing and nationalist extremism

But, neurological issues with progressive cognitive decline are—terrifyingly—not too uncommon in my family tree and I can tell you this:

Unkindness is not a symptom unless the person was fundamentally unkind in the first place

You can’t blame people’s cruelty on a virus

What's worse is that most of the popular critics of this worldview and state of affairs are running largely on vibes as well. Same methodology. Same cherry-picking of references. Same kind of reasoning through showmanship.

It's quite disheartening on the whole and makes me question the point of writing essays like I have over the years.

One of my realisations I've come to during my newsletter/blogging pause is that the vibes crowd has thoroughly won, both in tech specifically and in general. Facts don't matter. Research doesn't matter. If it has research aesthetics and has the vibes you like, people treat it as truth. Motion and churn with the right vibes count as progress. Revenue is treated as evidence of inevitable future profit, no matter how irrational the underlying economics are.

There's no convincing or reasoning with people if they think your facts have a bad vibe. Explaining things, with references, has no impact because the references are gauged based on vibes and not how well the studies were structured or how well the paper is argued. There is no difference today between decision-makers in tech and the antivaccination crowd. They both operate on the same epistemology and worldview

And if you ever find a manager who is an outright fan of these people: run, don't walk, in the other direction. Do not work for them if you value your mental health.

So, between the Mullenweg/Wordpress, DHH/37 Signals, and Sam Altman/OpenAI dramas, we now have pretty reliable ways for gauging the judgement and critical thinking skills of pretty any given commentator in tech.

Any podcaster or writer who defends these dudes and organisations is almost certain to not be, well, fully using what little capacity for reason nature gave them (to put it politely).

IME, if they show poor judgement on this, they almost always have poor judgement in general

“Teens Are Talking to Pro-Anorexia AI Chatbots That Encourage Disordered Eating”

https://futurism.com/character-ai-eating-disorder-chatbots

This industry fundamentally does not care it's harming people or not.

"Denmark: AI-powered welfare system fuels mass surveillance and risks discriminating against marginalized groups – report"

https://www.amnesty.org/en/latest/news/2024/11/denmark-ai-powered-welfare-system-fuels-mass-surveillance-and-risks-discriminating-against-marginalized-groups-report/

> the sweeping use of fraud detection algorithms, paired with mass surveillance practices, has led people to unwillingly –or even unknowingly– forfeit their right to privacy, and created an atmosphere of fear.

This, and not the AI bubble, is going to be AI's legacy. Accountability sinks that enable harms on a massive scale.

Plan B it is

https://www.baldurbjarnason.com/2024/plan-b-it-is/

“AI overwhelmingly prefers white and male job candidates in new test of resume-screening bias – GeekWire”

https://www.geekwire.com/2024/ai-overwhelmingly-prefers-white-and-male-job-candidates-in-new-test-of-resume-screening-bias/

“Researchers say an AI-powered transcription tool used in hospitals invents things no one ever said - ABC News”

https://abcnews.go.com/US/wireStory/researchers-ai-powered-transcription-tool-hospitals-invents-things-115170291

“AI” models are suffused with US values and, occasionally, those are quite shocking to us non-Americans

An Icelandic police force used a generated image to promote a public notice

People were absolutely horrified

Why?

Because the cop in the image had a gun, in a holster, AROUND CHILDREN 😱

The uniform was also not accurate for an Icelandic cop but what people found obscene was the idea that anybody would carry a gun around children, even holstered

https://www.ruv.is/frettir/innlent/2024-10-20-logreglan-a-sudurnesjum-aetlar-ad-haetta-ad-nota-gervigreindarmyndir-af-logregluthjonum-425104

“The Return”

http://hackeducation.com/2024/10/05/the-return

> I’ll be damned if I’ll go quietly and let the techno-libertarian fantasies strip the future from the rest of us.

> A GitHub [survey published in August](https://github.blog/news-insights/research/survey-ai-wave-grows/) found that 97% of software engineers, developers, and programmers reported using AI coding assistants.

97% using tools that generate 41% more bugs. Fuck all of this. What a fucking clown car of an industry

“Devs gaining little (if anything) from AI coding assistants | CIO”

https://www.cio.com/article/3540579/devs-gaining-little-if-anything-from-ai-coding-assistants.html?amp=1

> Many developers say AI coding assistants make them more productive, but a recent study set forth to measure their output and found no significant gains. Use of GitHub Copilot also introduced 41% more bugs, according to the study from Uplevel, a company providing insights from coding and collaboration data.

Like I've been saying for a while, feeling productive is not the same as being productive

Since every new release by OpenAI is usually followed by nonsense commentary and bubble inflation: a note

Testing a large statistical model with a handful of manual tests ("it solved this riddle!") is like assuming the home remedy you took cured your cold. These are methods that only lead to superstition.

“Going Buildless | Max Böck”

https://mxb.dev/blog/buildless/

I've gone buildless for personal projects for a long while now and am not going back

Also on HTML includes: it makes me sad to think that XML had standardised includes almost 20 years ago. I wasn't a fan of XML at the time but in hindsight it's looking like we, web devs, dumped an entire ecosystem of tools and standards just out of an annoyance with error handling that we then got a worse version of anyway with JSX, setting us back a decade