Notices by Baldur Bjarnason (baldur@toot.cafe), page 2

Tech is also extremely bad at software. We habitually make fragile, insecure, complex, and hard to maintain code that backs poor UIs. The best case scenario is that LLMs accelerate already broken software dev processes in an industry that is built around monopolies and billionaire extremists

But, sure, feeling discouraged by the state of the industry is "like quitting carpentry as a career thanks to the invention of the table saw"

Whatever

Reading articles in my feed reader used to be a safe haven, a space where things were quieter and marginally more sensible than social media proper, but lately reading through it just makes me grumpy.

The sceptical got "podded" one by one until only weirdos and the innately distrustful were left speaking. Critics stopped talking because saying anything could have ended their careers.

Belief in the bubble spread until it felt like you were living in the second half of a "body snatchers" movie

The "pod people" effect has kicked into overdrive with the AI Bubble

Most people with influence in tech seem to have been "podded" already, those without influence seem to accept it as a given, and those still voicing criticism are explicitly on the outside, no longer a part of their prior field

Remember, the facts haven't changed

LLMs are still:

- dangerously biased
- electricity hogs that threaten the energy transition
- engines of non-deterministic volatility
- prone to catastrophic errors
- points of centralised control over speech and work

and more

This is, needless to say, incredibly demotivating to feel this isolated in your field. The last time I felt like this, in Iceland during its bubble, I literally left the country. But at least then I could continue working in my field—the same job even, remotely

Iceland rebuilt itself to be a genuinely better place after the bubble popped. Still run by oligarchs and quite corrupt, but much better than it was before

I'm not sure there will be anything left of tech to rebuild after this pops

"What do you mean this is a dangerous bubble? My uncle just got rich buying a second flat with a 100% mortgage. Nobody ever loses money from real estate."

Anecdotal evidence of success from peers was seen as proof that bubble talk was lies and scaremongering, that there weren't any systemic issues

One symptom of a deep bubble is what I call the "pod people effect". Basically, people who seem to go to sleep one day sceptical about the bubble and wake up believers—they've changed their minds purely on social proof and personal impressions without a shred of genuine evidence.

The 2007 Icelandic finance bubble, for example, was all-encompassing here. Everything was of a mind with the bubble. Research said it was great. Media hyped it throughout. Everybody around you was a true believer

A few quotes from an excellent podcast discussion, "You deserve better brain research", between a neuroscientist and a psychologist (@grimalkina and @analog_ashley ) on the "Your Brain on ChatGPT" paper

https://www.changetechnically.fyi/2396236/episodes/17378968-you-deserve-better-brain-research

> Because I'm not even sure, like bioRxiv would accept this.

The whole thing is worth a listen (or a read, the transcript is workable).

On the one hand, legal fights in the US have massive consequences not just in the US but also around the world

OTOH, the US is a dysfunctional state run by chucklefucks and grifters which with its chaotic legislative bodies and judiciary means it's basically Calvinball with real-world consequences

It's not exuding a sense of stability and sound governance is all I'm saying

When people call LLMs “useless” they’re generally being kind as it’d be more accurate to call it harmful, dangerous, toxic, or risky.

It’s like calling white asbestos “useless”. Technically true in most western countries because you literally can’t use it, but it kind of elides the reason why

“ChatGPT Has Already Polluted the Internet So Badly That It's Hobbling Future AI Development”

https://futurism.com/chatgpt-polluted-ruined-ai-development

Also, a recurring security concern (and bandwidth abuse issue) with training data sets is that they generally don't store copies of the data and have few safeguards checking if it's been changed

So, y'know, if a pre-2020 domain expires and is replaced entirely with slop, many datasets will mark the slop as pre-2020 data

I seem to be the only person in the world who doesn’t want multitasking on his iPad. The reason why I like it is it’s my “do only one focused thing at a time” device. All multitasking gives me is a bunch of random features I accidentally trigger all the time until I find a way to disable them.

Avoiding generative models is the rational and responsible thing to do – follow-up to “Trusting your own judgement on ‘AI’”: https://www.baldurbjarnason.com/2025/followup-on-trusting-your-own-judgement/

“Disney, Universal sue image creator Midjourney for copyright infringement | Reuters”

https://www.reuters.com/business/media-telecom/disney-universal-sue-image-creator-midjourney-copyright-infringement-2025-06-11/

I mean, we all knew this was going to happen eventually, right? And that they're going to want to make an example of them, right?

The reason why “AI is inevitable” is a leap of faith and not an actual argument is that these systems still cost substantially more to run than they earn, cost a fortune to build and maintain, and have already hit diminishing returns

That last part means that any efficiency gains found will be spent improving the model instead of lowering costs

A not-so-short blog post on cognitive hazards, homeopathy, "AI" and the epidemiology of beliefs.

Trusting your own judgement on 'AI' is a huge risk: https://www.baldurbjarnason.com/2025/trusting-your-own-judgement-on-ai/

Comforting thought of the day: none of the software industry’s popular success metrics would catch a Boeing 737 Max style failure—catastrophic failure a year after delivery. So, you can literally murder scores of people with a software defect and still get that bonus at work for a successful project

That’s a less comforting thought if you’re a software user, admittedly

Flipping through my feed reader and it really feels like the pro-AI blog posts have ramped up considerably in both intensity and volume over just the past week, all filled with massive caricatures of “AI” criticism and a blind faith in what the tools are promising to deliver

What’s worse is I think most of the posts are 100% sincere

All else being equal, pointing out that the argument laid out in a company’s blog post, if sincere, would indicate a relaxation of standards and quality, is always worthwhile when said company is in the infrastructure and hosting business, irrespective of whether it changes anybody’s mind about “AI” or not