Léo Varnet @leovarnet@qoto.org
- Institution
- CNRS, ENS Paris
- Research topics
- Auditory perception, psycholinguistics, hearing loss.
CNRS researcher at École normale supérieure Paris. Auditory perception, psycholinguistics, hearing loss. My toots are searchable #tootfinder.
Joined Oct 2022
Léo Varnet
boosted
As someone who has participated in multi-year edits wars over, yes, Nazi shit, I will say that my biggest concern here isn’t about unedited LLM text hitting wikipedia articles—that’s v bad but probably largely fixable—but with the way Talk page sophistry is about to become absolutely fucking unmanageable as malicious editors set chatbots to do their infinite argumentation for them
To generalize: LLMs on the web’s surfaces are bad. LLMs in the backstage are much worse.
https://www.vice.com/en/article/v7bdba/ai-is-tearing-wikipedia-apart
@danielyuksek Thank you for reviving this topic! One limit I see to this line of reasonning is that it assumes that biases are related to the perspective of the individual, and not to that of society itself. If you think of gender biases as a difference in individual values/point of view, how would you explain that, in a jury, female faculty exhibit a bias against female applicants? (https://www.pnas.org/doi/10.1073/pnas.1211286109). This being said, I agree with you that being aware of its own bias plays an important role in dealing with them! (https://www.nature.com/articles/s41562-019-0686-3)
Happy to introduce the new release of the fastACI toolbox (v1.3) for investigating auditory perception using reverse correlation (https://github.com/aosses-tue/fastACI/tree/v1.3). This release includes new functionalities and examples, e.g., the possibility to replicate Ahumada's seminal reverse correlation experiment on tone-in-noise perception!
Léo Varnet
boosted
So folks, don't forget to check if your Ikea chair is compatible with your screen. I'm not kidding. A thread 🧵.
I've had the problem with my new screen for several weeks now, that every now and then the screen goes black for a few seconds.
Today I had enough and wanted to investigate the problem. So I started changing all the cables, plugging the screen into a different socket, and and and. Nothing helped.
We have started a wall of faces in the lab @lsp_ens. Not only will we finally know everyone's name, but taking pictures was also a lot of fun.
Léo Varnet
boosted
This is real-world XKCD
See full thread: https://twitter.com/rachel_l_woods/status/1649790435533922307?s=61&t=Ugdi4XBKf_2ovJ1y9hKs4w
@ceoln you were right, it was the ID indeed... But the plugin doesn't work with qoto apparently https://qoto.org/@freemo/110236883563374813
Léo Varnet
boosted
Does anyone else feel that modeling work has a shorter lifespan than experimental work? When I'm searching the literature, I'm less likely to be interested in a modeling paper from e.g. 20 years ago than an experimental paper from the same time.
Modeling work is so dependent on the assumptions and mindsets (and computational tools) of the time, it seems it best serves to move thought forward in the moment, but not (at least on average) last a long time.
@w101 can you give me a hand with this issue maybe...? Thanks in advance!
Hello #Fediverse! Some #MastodonHelp needed here! I'm trying to include a mastodon feed to my #Wordpress website, but I need my "user ID" (https://wordpress.org/plugins/include-mastodon-feed/#description)... Does anybody knows where to get this info on @QOTO #QOTO
Sorry to ask such a basic question... Can anybody help me, @freemo maybe?
Léo Varnet
boosted
Dooling and Danks provocatively argue Psychology is not ready to move beyond hypothesis testing - in 1975. https://link.springer.com/content/pdf/10.3758/BF03336685.pdf?pdf=button They claim our understanding of underlying theories is often too limited to generalize beyond the levels of the independent variable chosen in a specific experiment. I feel this is true in most researchlines in psychology. This also makes it difficult to interoret effect sizes meaningfully, build more generalizable models, and predict outside of direct replications.
PhD offer in Brussels
A fantastic #PhD opportunity to work work on adolescence as a second sensitive period for complex auditory processing with a fantastic supervisor @AxelleCalcus, in a fantastic city! #auditory #development #PhDPosition
Léo Varnet
boosted
Sur twitter une bonne âme a partagé l'extrait. Difficile de trouver meilleure description de la situation actuelle.
Léo Varnet
boosted
Sorbonne Paris Nord de nouveau bloquée par nos étudiant(e)s ce matin.
#greve13avril
Léo Varnet
boosted
A propos de #sondages, je trouve cette ligne de défense des instituts particulièrement intéressante : "Par définition, les sondages ne se trompent jamais, car ils n'ont pas vocation à prédire" (Laurence Parisot, ancienne présidente de l'#IFOP et du #MEDEF) ou "les sondages sont des photos, ils n'ont pas de caractère prédictif" (Roland Cayrol, ancien directeur de l'institut #CSA). Effectivement, on peut vouloir mesurer l'opinion à un instant t et non prédire les résultats d'une élections... Mais sans ce dernier aspect pour le relier à un point de référence réel, le sondage devient un indicateur circulaire qui ne mesure rien d'autre que "ce qui est mesuré par un sondage". #pseudoscience
Je viens de finir "10 leçons sur les sondages politiques" d'Alexandre Dézé @umontpellier -- visé par une plainte en diffamation pour avoir critiqué la méthodologie d'un sondage de l'#IFOP (cf. "L’IFOP tente de faire taire les universitaires" @Mediapart https://blogs.mediapart.fr/les-invites-de-mediapart/blog/060422/l-ifop-tente-de-faire-taire-les-universitaires-soutien-notre-collegue-alexandre-deze )... Et pourtant l'auteur ne fait que dire ce qui est répété par le milieu scientifique depuis des décennies: les #sondages souffrent de faiblesses méthodologiques caractérisées, qui en font un très mauvais outil prédictif.
A propos de la prétendue scientificité des sondages, cette citation de Pierre Weill, fondateur et ancien directeur de la #Sofres, est assez éloquente : "Au fond, c'est comme de la cuisine. Il y a une part de pifomètre, mais cela n'empêche pas de suivre quelques recettes" (https://www.lemonde.fr/a-la-une/article/2011/03/08/redressement-des-sondages-pifometre-et-autres-recettes_1490007_3208.html) #LibertesAcademiques #pseudoscientificpractices #pseudoscience
- Institution
- CNRS, ENS Paris
- Research topics
- Auditory perception, psycholinguistics, hearing loss.
CNRS researcher at École normale supérieure Paris. Auditory perception, psycholinguistics, hearing loss. My toots are searchable #tootfinder.
Joined Oct 2022
