So I just stumbled into Team Human, because a Youtuber I kind of like (Kraut) published a video for them. They seem to run a campaign targeted at creators, pushing to get videos released on the 10th or 17th. So they must be having a bit of a high point right now. (Idk if cash changed hands.)

Their manifesto tries to amp up societal harms, but the existential risk is snuck in at the end (AI programmed pandemics, etc.) and they want to regulate AI chips like nukes. CAIS (Center for AI Safety) is listed as the sponsor. Seems a little “AI centrist” to me, as in aligned with the less insane billionaires, but still a typical AI safety astroturfing shop. Shame if any of your favorite creators fell for it.

Das man sich gerade auf einem niedrigeren Niveau einpendelt als früher halte ich auch für plausibel. Mit Produktivitätsanstieg hat das aber wenig zu tun.

Da ist nichts spannendes passiert. Professionelle Grafikdesigner werden nicht von Software die funktioniert zu Sloposhop wechseln. Wer eine zuverlässige FOSS Version möchte geht immer noch besser zu Krita oder GIMP.

Etwas Kontext zum Kontext: Als ich noch beruflich mit Leuten zu tun hatte, die ständig allen erklären wollten, dass die Blockchain die Zukunft ist, wurde auch gern das Internet als Vergleich herangezogen.

Eine zuverlässige Quelle zu finden war mir zu viel Aufwand. Ich weiß mittlerweile garnicht ob der ursprüngliche Post was aktuelles war, oder ob nur eine Studie aus 2025 wiedergekäut wurde.

Per Suchmaschine findet man einfach so viel Slop nach dem Schema “Die meisten KI Projekte scheitern… beachte daher diese geheimen Tipps um es besser zu machen!” Darüber ist das Meme.

Ursprünglicher Post in !de_edv@feddit.de wurde gelöscht.

Genauen Link finde ich nicht, referenzierte aber sowas wie das hier: https://deutsche-wirtschafts-nachrichten.de/716921/95-prozent-scheitern-warum-kuenstliche-intelligenz-in-unternehmen-zur-falle-wird

Bro, what’s this (1+x)(1+x²)(1+x⁴)... ellipsis bullshit? You gotta get in on that big Π action!

though the lobotomized themselves weren’t among the loudest advocates of the procedure afaik

Ah well, that’s them computos for ya 🤷

Okay, what’s up with this title? Is this an 4-years-late announcement of support for Theo Seremitidis, who was unfairly dismissed by an australian company over labor safety disputes? Are there other Theos in need of support right now? Just Theo appreciation day in general?

I am utterly sniped. I have to go buy groceries now.

Also works for cocaine, I’m noticing.

Might be best as a comeback, if someone compares AI to a pharmaceutical breakthrough like antibiotics.

Just had a shower conversation with myself, and for some reason the crystal meth metaphor (methaphor?) is really coming together in my head.

“Have you even tried some frontier meth?”

“Sure, you can also get energy with exercising, eating well, and structure. But we don’t have to live like amish people, now that we can have meth!”

Nationalized Open Weight Models -> State-funded AI dispensory

There’s surely a rich bounty of comparisons left to mine.

The bubble will pop. It’s much too big to be sustainable.

But that won’t fix software industry employment because the Large Language Models don’t really need to work that well for this process. They only need to sort of work.

Get ready for the era of digitally sovereign on-premise open weight perpetual software sweatshops.

Ich weiß nicht was der betroffene Remmo beruflich macht, aber sich in die DMs von einem MdB zu schleimen der das Konzept Clan-kriminalität kritisiert und dann damit auf Insta anzugeben, wär schon so ein Arschlochmove den ich vom Mitglied eines kriminellen Clans erwarten würde.

Sorry for being late to the party, but I had to give my two cents on AI detectors. (specifically GPTZero & co., watermarking is different) I think it’s a bad idea to recommend them. These companies are overpromising their accuracy and essentially selling false certainty. This will also not fade over time since the inaccuracy is baked into the tech.

Currently the performance may be good for basic use cases, but recent papers seem to indicate you can still dazzle them by burning more tokens or explicitly prompting against detectors. But from what I understand about GenAI training* this is an eternal cat & mouse game, since the AI vendors don’t want their output to be detectable.

You’re right. The grammer is right, but it still sounds off. Like, idk, someone fully capable of english leaving threats in his second language to be sneaky?

Je, wrong gramnar and the german feels totally off. Much too wooden amyway for a proper death threat.

I think he realized it too and sockpuppeted a better example. Check the comment by “Gerard”

I bumped heads here before on that take, but I can empathize deeply with these people.

I expect that interacting with most of them will be infuriating, since they probably accepted the “this is just how things are done from now on” narrative of AI and are some level of infatuated with the fancy demos.

But this acceptance is also where the inescapable soul-crushing comes from. If you can adopt an internal posture of resistance, the prevalence of AI is still grating but you can adopt a perspective that doesn’t devalue your own skill and makes personal impact conceivable. Many of these folks are pretty far gone, but a pipeline of recovery should be possible from a starting point like this.

The concept seemed interesting to me, but they did seem to enjoy making everything as complicated as possible.

I did find a CLAUDE.md when I was browsing their core repos a while back. That was only the web frontend I believe, not the core algorithm. Though it seems generous to assume they weren’t slopping up that one too.