“LLMs pose a unique risk to science, education, democracy, and society that current legal frameworks did not anticipate. This is what we call ‘careless speech’ or speech that lacks appropriate care for truth. Spreading careless speech causes subtle, immaterial harms that are difficult to measure over time. It leads to the erosion of truth, knowledge and shared history”
https://www.ox.ac.uk/news/2024-08-07-large-language-models-pose-risk-society-and-need-tighter-regulation-say-oxford
Large Language Models pose a risk to society and need tighter regulation, say Oxford researchers | University of Oxford

Leading experts in regulation and ethics at the Oxford Internet Institute, have identified a new type of harm created by LLMs which they believe poses long-term risks to democratic societies and needs to be addressed by creating a new legal duty for LLM providers.

@FrankPasquale AKA “bull$hit”

@knutson_brain @FrankPasquale

Maybe, but obviously in some country at least HALF the HOMO SAPIENS population is addicted to living being repeating Goebbels inspired CareLess speech

ask #maga and #trump and #putin

I suggest that
#llm #llms #stochasticparrots #machinelearning #ML #AI #artificialintelligence are MUCH MORE apt at saying true things than #trump or #putin and #afd and #RassemblementNationalRN and #milei and #bolsonaro and #stevebannon and #robertmercer and #foxnews and #nigelfarage and #borisjonhson and #ultraright and #rupertmurdoch and #farright and all their propagandists

ouch I forgot #professionaldisinformers like #climatedeniers and #bigoil propaganda cf. #agnotology and the science of programmed ignorance and doubt

That's what I think I observe for now

@stphrolland @FrankPasquale
There's signal, antisignal (of which you speak), and noise (or bull$hit, adopting the Frankfurt definition here):

https://www.youtube.com/watch?v=o0MF8HMUmIk

On Bullshit - Harry Frankfurt (2005)

YouTube

@knutson_brain @FrankPasquale

Your link is super! Deep thought.

BUT THERE'S BEEN an evolution: he says that politician caught lying are threatened by grand jury and public outrage

THIS HAS BECOME FALSE according to me

But overall the analysis is great.

@knutson_brain

You beat me to it. Good summary!

I can recommend reading Harry Frankfurt's On Bullshit. It's a short read.

Related:

https://pivot-to-ai.com/2024/06/29/chatgpt-is-bullshit-scientifically-speaking/

@stphrolland @FrankPasquale

ChatGPT is bullshit, scientifically speaking

In their paper “ChatGPT is Bullshit,” three University of Glasgow philosophers, Michael Townsen Hicks, James Humphries, and Joe Slater, argue that “AI hallucinations” are better described as Frankf…

Pivot to AI

@floppy @knutson_brain @FrankPasquale

I am not convinced about #chatGPT and #LLMS as being only bullshit

I use them as a #wordembedding systems

When I use them, I use them to express me things through the lens of their #wordembedding

and it is "up to ME" to determine and correct what they give me, and it requires further work from me ok, but as is: it can be a precious tool

I still consider it a super nice tool

And maybe we should focus communitcation about #LLM #LLMs on #wordembedding

a #wordembedding systems has no goals for deceiving, nor gains from deceiving, (unless programmed as pre-instructions by OpenAI, or ElonMusk... which is... what they do...)

LLMs do not "care or do not care" - that is not something that makes sense

They output a construction which depends on a wordembedding. And that's all. According to my current undertanding (which is still limited).

EDIT: answered before I read the paper you linked me to. Hope to read it today.