Emma

I’d like to coin a new term: “algoholic” (I think you all know what it means…)

Coming soon: ERCIM Forum Beyond Compliance Let's talk about digital ethics in academia with international experts on: geopolitics, data altruism, GenAI & more... 🗓️ 29-31 Octobre 2025 📍 Rennes, France 💡 Free 🔗 edu.nl/v8re9 @Inria_Rennes #CCNEN #AIEthics

Meta has released a huge new AI language model called OPT-175B and made it available to a broad array of researchers. It also released a technical report with some truly extraordinary findings about just how dangerous this machine can be. 🧵 #AI #OPT175B

Hinda Hanned questioning responsible AI at #ICTOPEN2022 and pointing at what it would require to do better in the future, in practice 💡@hanedhinda @CAIL_Amsterdam 💡

To find more about check out this video:youtube.com/watch?v=CMXu66…

Don't miss the #ICTOPEN2022 Conference! @hanedhinda will give a keynote presentation on "Defining and mitigating algorithmic bias: a practitioner's perspective". @emma_beauxis has also prepared a poster presentation on "Unsupervised bias discovery for Fair AI".

@emma_beauxis & @michaelcochez presenting their work on Unsupervised Bias Discovery for Fair AI at #ICTOPEN2022

@kendraschaefer Very interesting, thanks! Yet… users could choose among predefined tags. Perhaps companies are reluctant to give users such control, because it reduces theirs? And impedes them selling profiling software?

That was likely due to technical issues. Many algorithms use pre-set tags (like "conservative" instead of "right-wing"), and allowing users to modify tags could cause coding havoc. Plus, imagine if users mis-spelled tags when changing them. Chaos. 10/11

Well, here they are. Yesterday, China's cyberspace watchdog, the CAC, finalized China's groundbreaking new rules on recommendation algorithms. They take effect in early March. Thoughts on the final version below. 1/11 cac.gov.cn/2022-01/04/c_1…

"Responsible Recommenders in the Public Library" Our PhD vacancy on translating ethical principles into metrics, experiments and recommender system design, will close in 3 days. With @informusiccs @jrvosse @KBNLresearch and me cwi.nl/jobs/vacancies… @CWInl @cultural_ai

There's an opening for an associate professor (tenured) focusing on social data science at our department. academictransfer.com/nl/307283/asso…

3 weeks ago LAION-400M dataset (now a billion+), first Image-Alt-text pair dataset of this scale was released. @vinayprabhu, @MannyKayy & I dug into it arxiv.org/abs/2110.01963 Long tread 1/ Warning: paper contains NSFW content that may be disturbing, distressing &/or offensive

Large scale datasets sourced from the web are plagued w/ problems including inclusion of problematic content & seemingly obvious go to solution is filtering but filtering itself raises its own set of problems & I am grateful for this paper laying out the nuances beautifully 1/

Over the past few weeks, the Facebook Papers have reaffirmed that FB has fueled the spread of hate speech & misinfo around the world. But there’s a crucial piece missing from the story. FB isn’t just amplifying misinfo. The company is also funding it. technologyreview.com/2021/11/20/103…

@MarankeWieringa Great question! I’ve no answer but just an extra option perhaps: using dots instead of hyphens. Visually it’s more consistent (tiny thus more space, no extra symbol to recognize). Technically you’d need to make a domain + a sub-domain (may edit a config file, but not too tricky)

Visually impaired tweeps! I have a question. I'm about to register a domain name and I was wondering what kind of set-up would be better for screenreaders. The domain name will be made up out of two words. Will it help you if there's a hyphen between the words?

PhD vacancy at @CWInl on "Responsible Recommenders in the Public Library," translating cultural values and ethical principles into concrete metrics, experiments and recommender system design. Co-supervised by @informusiccs @jrvosse @KBNLresearch and me cwi.nl/jobs/vacancies…

How many #platforms in the UK pay a living wage? Source: Fairwork UK Ratings 2021 bit.ly/3kSLwlL #LivingWageWeek

@Abebab (indeed, and being stupid or delusional remains very possible)

an impossible task too

@Abebab Formalizing intelligence & sticking to a single definition of it (or sets of tools/methods) is actually a definition of stupidity (imho)

can't decide if this is naivety or ignorance

I have yet to come across work on AGI that I can take seriously

@rajiinio Let’s also imagine that they restore it, at some point. Eg, after people extensively tag each other, thinking there’s no more facial recognition at play.

To put this in context: Facebook pioneered the use of deep learning for facial recognition with the DeepFace model in 2014, because they had more face data than anyone else at the time (arxiv.org/pdf/2102.00813…). For them to now (finally) reject this technology is a big deal.

Proud of Sarah-Jane too! And grateful to @dvdgrs & @RandstadNL for giving her such a great support. It was wonderful to embark in such experimental work together, thank you!