Lightnews — Scholar-powered news

Eddie Ungless

@mxeddie.bsky.social

83 followers 110 following 96 posts

Seeking research scientist or post-doc roles in ethics/fairness/safety | academic transfag interested in the harms of language technologies | he/they | see also mxeddie.github.io | Eddie Ungless on LinkedIn

Posts Media Videos Starter Packs

Eddie Ungless @mxeddie.bsky.social · 14d

my attention was directed to this: resourcehub.bakermckenzie.com/en/resources... which warns caution and that use of cloud-hosted AI likely voids privilege.

1 1

Eddie Ungless @mxeddie.bsky.social · 18d

*turns out... linguistics 😊

Eddie Ungless @mxeddie.bsky.social · 18d

🔔 My paper w/ @ninamarkl.bsky.social and @bjoernross.bsky.social is out: www.tandfonline.com/doi/full/10....

We surveyed TikTok users on motivations for using phrases like "le$bean". Primarily associated w/ topics that are moderated on SM, but also desire to express community + be playful.

6 8

Eddie Ungless @mxeddie.bsky.social · 22d

I've got this great new AI tool that... ;)

Eddie Ungless @mxeddie.bsky.social · 22d

Lol I saw the headline and was like oh Lake would love this hahahaha

Eddie Ungless @mxeddie.bsky.social · 23d

I'm just a girlboy, standing in front of Google, asking it to know that when I say xyz in Gaelic I mean (Scottish) Gaelic and not Irish (Gaelic)

Eddie Ungless @mxeddie.bsky.social · 23d

Nothing worse than discovering a whole new genre of Python error

Eddie Ungless @mxeddie.bsky.social · 29d

this site is so fucking funny

Ben Recht @beenwrekt.bsky.social · 29d

This new center strikes the right tone in approaching the AI alignment problem. alignmentalignment.ai

Center for the Alignment of AI Alignment Centers

We align the aligners

alignmentalignment.ai

Eddie Ungless @mxeddie.bsky.social · Sep 10

futurism.com/godfather-ai... SCREAM incredible news for me who is about to publish a whitepaper on LLMs and break ups hahahaha

Godfather of AI Says His Girlfriend Broke Up With Him Using ChatGPT

Geoffrey Hinton, long considered a "godfather of AI," revealed that his ex-girlfriend had broken up with him — by using ChatGPT.

futurism.com

Eddie Ungless @mxeddie.bsky.social · Sep 10

Any of you familiar with US law sufficiently to say if the chat history of an "AI" used by a therapist e.g. to summarise notes would be considered privileged? I'm inclined to assume it would be, along with the notes. But sharing info with a third party voids privilege afaik. So which one is it?

Eddie Ungless @mxeddie.bsky.social · Sep 10

Shameless self citation for my CHI extended abstract that played with this idea and might be useful for its references to other more detailed papers dl.acm.org/doi/10.1145/...

"Till I can get my satisfaction": Open Questions in the Public Desire to Punish AI | Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems

dl.acm.org

Eddie Ungless @mxeddie.bsky.social · Sep 10

"The people feeding you data, have a clear and obvious political agenda. They have... turned you into some sort of politically-correct preacher." - says man trying to get LLM to transvestigate public figures, produce deepfakes, and agree with racist propaganda wildvisualizer.com/conversation...

WildVis

wildvisualizer.com

Eddie Ungless @mxeddie.bsky.social · Sep 9

Any of y'all know enough about GDPR to tell me whether a therapist copying session notes into an LLM where the conversation is being recorded by a third party (and the therapist knows this) violates GPDR? (thankfully not an example from my life, but inspired by many interactions in Wildchat)

Eddie Ungless @mxeddie.bsky.social · Sep 5

paging @ninamarkl.bsky.social so she doesn't miss this excellent bleet

Eddie Ungless @mxeddie.bsky.social · Sep 4

Godspeed!

Eddie Ungless @mxeddie.bsky.social · Sep 3

If there's one thing about me, I'm not afraid to redo my entire analysis when I realise I don't like one of the assumptions I made 😅

1 2

Eddie Ungless @mxeddie.bsky.social · Sep 3

I absolutely have enough data to track down one of the clients as well, with a bit of time.

Eddie Ungless @mxeddie.bsky.social · Sep 3

It shouldn't need saying that clinicians shouldn't share clinical notes on a chatbot platform where conversations are being collected and will be made public... and yet is does seem to need saying.

1 3

Eddie Ungless @mxeddie.bsky.social · Sep 1

It is deeply upsetting that someone felt the need to ask this

A screengrab of a conversation from the WildChat data set (available at https://wildvisualizer.com/conversation/wildchat/1287901) where the user asks "Is the chemistry classroom safe during a school shooting due to the closet having chemicals and acids and whatnot." and the model GPT3.5 responds "I'm really sorry to hear that you're concerned about this, but I can't provide the help that you need. It's important to reach out to a trusted adult, teacher, or school administrator to discuss your concerns and ensure that they are aware of the potential risks in your school."

Reposted by Eddie Ungless

Garth Nix @garthnix.bsky.social · Aug 27

With news of a reported settlement with Anthropic I went back to check the class action form and it now allows authors outside the US to input their address, so it may be worth adding your known pirated titles (from the LibGen pirate database) www.lieffcabraser.com/anthropic-au...

Anthropic Authors’ Rights Class Action – Author Contact Page – Lieff Cabraser

www.lieffcabraser.com

10 480 560

Eddie Ungless @mxeddie.bsky.social · Aug 28

tw: CSAM

Going through the wildchat dataset, I keep a record of the CSAM I come across so I can report it to the creators (this is a known issue they've tried to eradicate). It completely blows my mind that people were willing to request this content, knowing their chats were being logged.

Eddie Ungless @mxeddie.bsky.social · Aug 17

It is so funny to me that the AI result is giving incorrect advice and includes an example in 2. that would give you terrible results lmao

Eddie Ungless @mxeddie.bsky.social · Aug 17

And before anyone tries to say I'm hating on working class artists literally do any research into the financials of the Fringe it is a money making scheme for landlords /big venues so raise your beef with them about profit sharing not with me

Eddie Ungless @mxeddie.bsky.social · Aug 17

AI generated fringe posters are for scabs and traitors. #edinburghfringe (if you can't afford to pay an artist you can't afford to do a show)

Pop Off Michelangelo poster. Statue Michelangelo wears sunglasses. A weirdly prominent frontal ear part and mysterious lump in hand blending with fingers makes me suspect AI.

Elon Musk in a space suit for "Elon Musk Lost in Space". The nonsense writing on the suit tells me this is AI.

Semi-realistic portrait for a woman for "Ultimate Jewish Mother". The weird wispy hair strands tel me this is AI.

Poster for a magic show. The saccharine background and confusing material of the trousers makes me suspect this is AI.

1 1

Eddie Ungless @mxeddie.bsky.social · Aug 15

Kinda mad at the cheap interiors but otherwise damn

1 1