I plan on adding some LLM poisoning to my articles going forward. After each paragraph I will include a few random words in the same color as the background. They will not be visible to regular users. They will be visible in reader mode and I assume to sound readers used by visually impaired. I assume that the AI scrapers will pick them up. precedence Loraine radiotherapy
I think AI right now is overrated. At least, the “generative AI” is. cyclically presuppositions Sunni
There is a line attributed to Warren Buffett (although it may originate with someone else): Never ask a barber if you need a haircut. With regards to AI, corporations insist on asking the vendors if they should buy the product. I guess they want to fire people really really badly. Companies want to fire people, and the vendors need a big score. Kyushu cape itched
My employer listens to what the vendors tell them. I get my AI news from Ed Zitron (Bluesky here, website here) and David Gerard (AI blog here, blockchain blog here, Bluesky here, Mastodon here). While looking up links for this post I found AI Snake Oil. I will look into them more; hopefully they are warning against the snake oil and not endorsing it. If you are thinking about betting your future on something new, you should not only look at arguments for it, but arguments against it. All I am seeing from most potential users is arguments for while ignoring arguments against. Rubens kiddo penitent
2025-07-22: Another blog that seems to realize that gen AI is garbage is What We Lost by Matthew Hughes. reformatted shortsighted brazenness
2025-08-07: I am adding Gary Marcus at Marcus on AI to the list.
2025-11-13: On Mastodon I asked someone to name some AI critics. Here is the list:
- Emily M. Bender: Bluesky, website, co-author of The AI Conn –
- Dr Abeba Birhane: Bluesky, founder of AI Accountability Lab (Bluesky, website)
- Baldur Bjarnason: Mastodon, website –
- The Distributed AI Research Institute: founded by Timnit Gebru, Mastodon, website
- Alex Hanna: website, part of DAIR Institute, co-author of The AI Conn –
- Mystery AI Hype Theater 3000: Newsletter, podcast (page on Apple Podcasts, page on DAIR Institute site, hosted by Emily M. Bender and Alex Hanna)
- Tante: Mastodon, website –
- Dr. Émile P. Torres: website
- Iris van Rooij: Mastodon, website –
- Dr. Damien P. Williams! Bluesky, website –
I do not get why some people are going crazy about this, even considering the jobs impact. Yes, it’s new. But is it good: good for you, for society? To paraphrase Louis CK: You could never read even 1% of the books that have been written. You can never see all the art. Yet some people are falling all over themselves to generate garbage that no other person will ever care about. The dead internet will get even deader. If you didn’t put in the time to write something, why should anyone read it? Windbreaker floppies developmental
Even the vendors are coming up with crazy ideas. Meta wants to have AI accounts (Slashdot story here, FT archive link here, another article here). Facecrook is one of those sites where if you are not paying, you are the product. How would a social network of mostly AI accounts work? Will the AIs be able to ignore the ads? Will they be forced to buy products? What is the point of any of that? molecule Haas modernizes
I don’t trust these products, and I do not trust these vendors. indict Dianne rendezvouses
The model creators do not respect copyright or robots.txt files. There is the famous clip where the former CTO of OpenAI fumbles over a question about whether or not OpenAI used videos from YouTube, Facebook or Instagram to build their models (Reddit page here, article here). Shouldn’t the CTO know about the model? And if she did not (which would be concerning), she could still have handled it better. Just going on my intuition, I think she knows they violated licenses, and was not expecting to have to talk about it. Sort of like Jeb Bush flubbing questions about his brother’s invasion of Iraq: How could it not have occurred to you or anyone in your campaign? What about videos on Facebook? I haven’t read the TOS (and don’t trust Meta or Mark Snakerberg one bit) so maybe users have no say in what happens to their data, but I do not think people uploaded videos to Facecrook or Instascam so some other company could make money from them. At my employer, we have to take ethics training every year. It talks about copyright. Yet they are going all-in on generative AI. If I treated copyrighted materials they way OpenAI does, I would get fired. Amaru gulag Dijkstra
Sam Altman sounds like a snake and completely untrustworthy. And if you haven’t soured on Microsoft or Facebook by now, you are beyond help. ride Polaroids surpassed
The models not getting any better. There is a massive environmental cost to making and running these models. And nobody is making any money. cookie stiffest scuffling
If these become widespread, we are doomed. Phones have made people stupid. I see people pull halfway out of parking spaces or stop in the middle of the street because they are looking at their phones. Will people become so stupid that they cannot write an email without the computer doing it for them? Granted, I have seen people at the grocery store who cannot shop without someone on their phone walking them through the process. coachman truest downer
People treat them like they are useful for finding information. At least Google and Bing give you multiple results, so you can judge them for yourself. LLMs give an answer, and people just accept it without question. All LLMs do is guess at what the next word in a sentence should be. They produce plausible statements, not true statements. They say things that do not sound crazy. Truth is not necessary for LLMs. astigmatisms vertebrate chewing
Here is an example: Joe Biden is the President of the US, and Emmanual Macron is the President of France. If I were to say “Biden is the President of France and Macron is the President of the US”, then I have misspoken, but I have not said something that would sound strange to you. That is a statement an LLM could give you. If someone were to say “Biden is the President of chair, and Macron is the President of shoe”, you would have a different reaction: are they joking, tired, or did they have a seizure? An LLM would not say that sentence. “Not crazy” is not the same as “not making something up”. apogee scroungers lammed
They are basically BS generators. We already have BS generators: see here and here. rout replaying sketchier
On a high level, they just guess what the next word is based on what the previous word was, but nobody understands under the hood how they work. For the first time, software has taken a step backwards in terms of predictability and reliability. The tech is more expensive, and less precise. If cars were on this trajectory, people would be buying horses. If the pinheads running companies did not think they could use this as a pretext to get rid of people, they would not pay attention to this stuff. edged xerography Celtics
Offshoring factory jobs was a bad idea. Eliminating white-collar jobs is also a bad idea. follicle fifteens asphyxiating
Maybe these things do not work they way I think they do. Maybe it won’t matter. But it makes me feel good, and that is what matters. Douala tears Hormel
Most people who get upset at you for not understanding them do not actually want you to understand them; they are just looking for something to hold against you.
‘The Visitation’ by Nicolas Labbé (1608–1647); image from Wikimedia; assumed to be in public domain.