Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
digitalcourage.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
Diese Instanz wird betrieben von Digitalcourage e.V. für die Allgemeinheit. Damit wir das nachhaltig tun können, erheben wir einen jährlichen Vorausbeitrag von 1€/Monat per SEPA-Lastschrifteinzug.

Administered by:

Server stats:

811
active users

digitalcourage.social: AboutStatusProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.8

#hallucinations

3 posts3 participants0 posts today
Public *
Aaron

WTF, #Google? This image appears nowhere in the article. I wonder how the #autistic author would feel about having this image attached to his article in your summary card. Given the title of the article, this seems especially egregious.

What you are doing is not merely #AI #slop. It is harmful. I posted just a day or two ago about the #ableist #hallucinations about autistic people that were presented as fact in a search for "Cassandra syndrome" -- which are still there, btw. What other misinformation are you feeding people about us? People who don't know to be skeptical about your presentation of autism.

EDIT: I was wrong. It wasn't Google's doing this time. You can see the image in the preview in the next post. It's #PsychologyToday that is responsible.

#ActuallyAutistic
@actuallyautistic

Disconnected puzzle pieces with letters on them, arranged on a white surface to wonkily spell out "autism". Below that, an area describing the linked article. The article is from Psychology Today and is titled, "When a friends [sic] says they're autistic, an uninformed reaction can be ruinous." Below that, it says "21 hours ago".
Replied in thread
Public
Petra van Cronenburg

@chema I'm a professional writer and journalist and I would *never ever* let #LLM #proofread or my texts. Never. Only real humans.

This has to do with the fact that good #editors and #proofreaders check a lot more than just word sequences. Above all, they have a feel for individual style (something that is completely destroyed by #ChatGPT) and know the briefing or exposé. @writers

BTW, #genAI can't even #hallucinate. For #hallucinations you need a brain. 😉

#noAI#writer#writersCommunity
Public
PrivacyDigest

Judge admits nearly being persuaded by #AI #hallucinations in court filing

A plaintiff's law firms were sanctioned and ordered to pay $31,100 after submitting fake AI citations that nearly ended up in a court ruling. Michael Wilner, a retired US magistrate judge serving as special master in US District Court for the Central District of California, admitted that he initially thought the citations were real and "almost" put them into an order.

arstechnica.com/tech-policy/20

A robot hand places blocks spelling "AI' on a wooden table that also contains law books and a lady justice statue.
Ars Technica · Judge initially fooled by fake AI citations, nearly put them in a rulingBy Jon Brodkin
Public
Chris Vitalos

If true, #hallucinations cast serious doubt on whether the end goal of #AGI can be achieved with today’s #LLM architectures and training methods.

While ongoing research explores #RAG and hybrid models and inference techniques, no implementation to date has fully eliminated flawed reasoning.

What consumer would trust mission-critical decisions if an AGI is known to confidently state falsehoods?

newscientist.com/article/24795

New Scientist · AI hallucinations are getting worse – and they're here to stayBy Jeremy Hsu
#GenAI#AI
Public
Miguel Afonso Caetano

Companies are spending billions building lying machines AKA LLMs:

If the market could pay just a fraction of the gazillions of dollars that they pay to train these Large Language Models to hire journalists, fact checkers, technical writers, editors, there would be much less need from users to rely on LLMs from the beginning. Most of the time, chatbots are solutions looking for problems that, ultimately, only end up creating more (bigger and larger) problems.

"One recent study showed rates of hallucinations of between 15% and 60% across various models on a benchmark of 60 questions that were easily verifiable relative to easily found CNN source articles that were directly supplied in the exam. Even the best performance (15% hallucination rate) is, relative to an open-book exam with sources supplied, pathetic. That same study reports that, “According to Deloitte, 77% of businesses who joined the study are concerned about AI hallucinations”.

If I can be blunt, it is an absolute embarrassment that a technology that has collectively cost about half a trillion dollars can’t do something as basic as (reliably) check its output against wikipedia or a CNN article that is handed on a silver plattter. But LLMs still cannot - and on their own may never be able to — reliably do even things that basic.

LLMs don’t actually know what a nationality, or who Harry Shearer is; they know what words are and they know which words predict which other words in the context of words. They know what kinds of words cluster together in what order. And that’s pretty much it. They don’t operate like you and me.
(...)
Even though they have surely digested Wikipedia, they can’t reliably stick to what is there (or justify their occasional deviations therefrom). They can’t even properly leverage the readily available database that parses wikipedia boxes into machine-readable form, which really ought to be child’s play..."

garymarcus.substack.com/p/why-
#AI #GenerativeAI #LLMs #Chatbots #Hallucinations

Marcus on AI · Why DO large language models hallucinate?By Gary Marcus
ExploreLive feeds
About