Oh I should spruik my Summer/Winter School talk.

Come see me try and explain my entire PhD thesis in 12-ish minutes! I worked with a community in regional Australia to find pathways to the Sustainable Development Goals at the local scale.

I'm talking at 10am AEST/0:00 UTC on Saturday 30 July.
Sign up here and they'll send you the link summerschool.scholar.social/20

Hey y'all, if you signed up for @bstacey exciting Summer/Winter school talk on the prices of physics textbooks, it'll be starting in an hour-&-an-half @ 14:00 UTC! Check your email for the link (make sure to peep that Spam folder just in case!)!

Also, check out all our talks here:

Thinking about taking notes with a text rather than on a text. Reading a text without taking notes is fast but I forget a lot. Reading while taking notes on the text is a slow grind of synthesizing, transcribing quotes, etc. to generate a reference document. Reading without the pressure of creating that mnemonic is quicker, freer, and allows for more creative thought around the text (which is usually the point anyway) while still creating a mnemonic for the text's structure.

Proud to announce: tenacity pays off, I am moving up in the academic pyramid! Starting in September my next step will be as a candidate over at University. 🥳 I get the opportunity to work on human factors in organizational , supervised by Kate Labunets and Slinger Jansen.

Looking forward to expand my network with new connections. Anyone from here on Mastodon? :boost_requested:

And don't let me started on how we consider the data in the overall "trustworthy AI things": it's quite overlooked, really. We just collect data from the wild that was constituted using microlabour platforms that create a new proletariat, we scrape data from the web without considering any consent framework. How to integrate the data (or a representation of the environment) into our research, and how to do it well (eg without killing the planet and enforcing marginalized groups exclusion)?

Show thread

Another topic is the explainability and accountability of artificial intelligence program. how to ensure stakeholders can query a program / the program's owner and require accountability for how they were categorized by the program?

Show thread

But it only addresses a small part of the topic.

For instance, regarding the privacy of a data-based program. Those big monsters tend to do a lot of memorization (and they were trained in what a consider an unrespectful way, but I digress) . Is there a way to quantify how much they memorized? Could we "patch" those programs, or design them in a better way so that you don't end up finding personal data while querying such program?

Show thread

Basically you have the whole "formal safety" stuff I made my PhD on. Since data-based programs are so huge, it's very complicated to have sound guarantees that they behave according to their specification. But it also comes from the fact that it tries to work on very abstract, intricated specifications. For instance, "do I need to pet my cat again today, or go out to buy them treat?" require to define the very high-level concepts of "cats", "treats", "outside", and combine them together.

Show thread

Sorry for the lack of activity by the way, I had my PhD defense in november and had quite some trouble to adapt to my new working environment. But now it's becoming better, so I'm back!

I'm now a trustworthy artificial intelligence researcher and, huh, part of the work is to figure out exactly what "trustworthy" means?

Today at lunch had the pleasure of listening a talk by @Cyborgneticz. It was really nice, even if (or because of?) it knew almost nothing on the subject. Thank you so much for sharing! I hope your PhD defense is a success!

(if you haven't yet, you can still join the next talks of @SummerSchool)

Here is a

Representations of AI - Why we need to do better by @tgupper will now be taking place on Friday 29th July 2022 at 12:00. You might want to check your availability on the website.


« Longtemps, la résistance du milieu des pompiers aux logiques néolibérales ou aux idées réactionnaires et racistes s’est nourrie d’une éthique forte et solide. Ils étaient fiers d’être, au sens plein, au sens noble du terme, un service public. » monde-diplomatique.fr/2017/03/

My parents (and some of my colleagues) are teasing me, basically "I deserved it by taking the train and not the plane".

Yeah, sorry for trying to not kill us even more with unnecessary carbon emissions I guess.

Show thread

There's more of it though. I decided to go to the conference venue (Wien) in train, rather than plane. From Paris to Wien, it's about 11h of train. Aaaand I missed a connection in München, so I will probably not be at my hotel before 23h :')

Show thread

On my way to the IJCAI-ECAI conference, presenting at the AISafety Workshop (although I will totally try to attend to the whole conference).

First onsite conference since three years, I must say I'm both happy and stressed.

apparently "y2k22" is a legitimate type of bug cropping up all over the place because... people are storing dates in yyMMddHHmm format... as 32 bit integers.

yep, software was a mistake. lets start over and try again, folks.


For all women in cybersecurity, here is a workshop that might be of interest: casa.rub.de/en/wisc-workshop

Show older
Scholar Social

Scholar Social is a microblogging platform for researchers, grad students, librarians, archivists, undergrads, academically inclined high schoolers, educators of all levels, journal editors, research assistants, professors, administrators—anyone involved in academia who is willing to engage with others respectfully.