This is disturbing:
How far away are we from lawyers on both sides arguing based on invented case law and the judge allowing it because his case law is also invented?
AI invents things. I ask it for simple book lists, and 40% of books on them are invariably completely made up. I asked for a list of Mexican presidents since 1990, and was horrified to find Porfirio Díaz on it. AI does this with the simplest tasks. It creates very biased narratives that aim to please the person who wrote the prompt. That gigantic, wealthy law firms use AI to write their briefs is extremely disturbing.
It’s annoying to me that people refer to the fake texts generated by AI as hallucinations. AI is not human. It’s an algorithm that creates plausibly sounding texts that might of might not coincide with reality.
Clarissa's Blog 
“gigantic, wealthy law firms use AI to write their briefs is extremely disturbing”
They can’t distinguish from things that current can do (write meaningless boilerplate) and things it can’t (actually access data in a way that human users need).
This is because they can’t make that distinction in real life… they can’t distinguish between bureaucratic forms that have to be done and the actual job they’re being paid to do.
I can’t decide whether the AI train is going to crash like a train into a mountainside when the tunnel stops or if it’s actually intended to turn people into bobbleheaded beasts who play with it while they await the truck to the abbatoir….
LikeLike
It scares me that even very educated people don’t understand how it works. These lawyers actually submitted these briefs to the court in a case worth billions of dollars.
LikeLike
“people don’t understand how it works”
Been listening to a youtube channel with an AI skeptic (Mo Bitar) and some others and the cluelessness of people who should know better is astounding and terrifying. He also untangles the hype about the ‘dangerous’ AI that seemed conscious (either branding or the people making it are clueless idiots).
Another video (new channel forget the name) is from someone training AI and finds it getting ‘dumber’ not ‘smarter’. I forget if he explicitly says that’s because of the algorithm or uses other terminology but it comes down to the same thing.
LikeLike
https://maldonpress.com/books/the_haxton_review.php
LikeLike
I have asked ChatGPT to write a couple book reviews for me that focus on the potential reader versus the book promos that focus on selling the book. I have been pleased with what I have received so far. I have also received ‘good’ answers to various questions but you need to phrase your question clearly and double check any answers.
LikeLike
I heard just a few hours ago, about an immigrant doctor in Victoria, Australia, using AI to fill out the details on a medical death certificate, with completely hallucinated details about the patient – phone number, who their carer had been, and so on. It was all found only by accident. I presume the doctor (who also spoke poor English) knew nothing about AI and just assumed that it’s accurate… Hallucinatory auto-completion is a whole new kind of false information for our information society to deal with.
LikeLike
“a whole new kind of false information”
Hugo Hofmannsthal, in his libretto to Elektra said it well… Klytamnestra speaking to her confidante/servants about Elektra who is being (suspiciously) nice to her.
“Zerrt ihr mich mit euren Reden und Gegenreden nicht zu Tod?Ich will nicht mehr hören: das ist wahr und das ist Lüge. Was die Wahrheit ist, das bringt kein Mensch heraus. Wenn sie zu mir redet, was mich zu hören freut, so will ich horchen, auf was sie redet.”
“Aren’t you dragging me to the grave with your words that go back and forth? I don’t want to hear anymore: This is true and that is a lie. No one can discover what the truth is. If she’ll tell me something, that would make me glad, then I want to hear what she has to say.”
LikeLike
AI results are definitely getting worse. They are longer, more rambling, more likely to contain serious mistakes.
LikeLike