Martin Ruskov @mapto@qoto.org

@tedunderwood.com that's actually amazing. It shows people capture the original intent, leaving the exact formulation as secondary. Then, of course ~200 characters are not enough for a detailed argument.

Off topic 1: one could say that LLMs successfully replicate this particular behaviour

Off topic 2: provides preliminary evidence why a training set from twitter is inevitably of poor quality

@TheConversationUS as far as I'm concerned, it matters less whether the message was written by GenAI, than whether the person behind it really meant what is written. And this could not be the case with or without GenAI

Sounds too much like the Kremlin:
"Around the same time, White House chief of staff Susie Wiles forced a meeting of Trump’s most trusted advisers. The problem: No one was being honest with the president about the domestic impact of the war.

Privately, Wiles had expressed fears that the inner circle’s rose-tinted retelling of the conflict would leave Trump oblivious to the political reality of the war, just months ahead of a contentious midterm season, reported Time magazine earlier this month."
https://newrepublic.com/post/209262/donald-trump-iran-war-plans-screaming-aides

"Among those monitored were a Palestinian academic invited to give a guest lecture at Manchester Metropolitan University and a pro-Gaza PhD student at the London School of Economics, according to internal documents.

In October 2024, the University of Bristol provided the firm with a list of student protest groups it wished to receive alerts about, an internal university email suggests. It included pro-Palestinian and animal rights activists.

In total, 12 universities paid the firm to monitor campus protest activity. Others include the University of Oxford, Imperial College London, University College London (UCL), King’s College London (KCL), the University of Sheffield, the University of Leicester, the University of Nottingham and Cardiff Metropolitan University."
https://www.aljazeera.com/news/2026/4/20/uk-universities-pay-to-spy-on-students-social-media-accounts

@tedunderwood.com @axeghostgame.bsky.social fair enough, my "legally unambiguous" is an inappropriate exaggeration. In a system of settlement agreements (Disney, News Corp, Springer and counting in the case of OpenAI), it is easy to get confused about the specific unresolved issues.

So according to this one review article:

1. A tendency might start to emerge to consider commercial AI systems (the technology) fall within fair use

2. But the (business) practice of accessing the original materials is not rarely through acts of piracy. This includes allegations of seeding pirate torrents to enable the anonymous acquisition of pirated content.

https://copyrightalliance.org/ai-copyright-lawsuit-developments-2025/

“Someone who says ‘I’m against abortion but says I am in favor of the death penalty’ is not really pro-life,” Leo said. “Someone who says that ‘I’m against abortion, but I’m in agreement with the inhuman treatment of immigrants in the United States,’ I don’t know if that’s pro-life.”
https://www.huffpost.com/entry/pope-leo-breaks-down-why-maga-is-not-really-pro-life_n_68dd802ae4b0c450ba64c434

@axeghostgame.bsky.social @tedunderwood.com those that don't find it are by judges that haven't brothered to ask for the training data. Those that find it, are able to confirm it even without requiring the data. It's the misguided idea of accountability of the "don't ask for permission, but for forgiveness later" mantra.

@tedunderwood.com and yet, we are talking of black box systems that have numerous issues, copyright infringements and illicit content being only two of the clear-cut and legally unambiguous ones.

The technology is one thing, the market is a different one. Legislation (already in the US and continuously more so in the EU) is deliberately closing its eyes of this distinction in the interest of multibillionaires. If not corruption, this is inequality politics at best.

Edit: see discussion below for an admission that "legally unambiguous" is an unnecessary exaggeration.

One of the core problems of GenAI is that it's trained on junk data no one has ever read and reviewed. Once again we see that's a lesson not learned, because in their endless quest to more data for more-of-the-same models, GenAI companies have found a new source of mediocre training slop: work-related chats.
https://gizmodo.com/failed-companies-are-selling-old-slack-chats-and-email-archives-to-train-ai-2000747916

What are they thinking? Have they never participated in such conversations not to know that in the context of remote-first work, these are the equivalent of watercooler conversations? Noise is the norm there, and transformer models are supposed to filter this out? All this without considering the survivor bias of failed companies (nice pun).