Boosted by @DataKnightmare@mastodon.xyz
In Germany, Google removed one paragraph from its description of AI Overviews: https://search.google/intl/de-DE/ways-to-search/ai-overviews/
I wonder why 🤔
Remote
Nicolas Kayser-Bril
@nicolaskb@fediscience.org
mastodon 4.6.9Reporter @algorithmwatch@chaos.social
1585 Followers
544 Following
22 Posts
Joined October 29, 2024
website:
Une excellente infolettre:
Ein sehr guter Newsletter:
✊🏻 I'm very happy that my employer now has a Betriebsrat (a works council, which has more power than in other EU countries).
Things I learned while setting it up:
- Don't underestimate the resolve of colleagues. I started by thinking that roughly one in four in the company would appreciate the effort. Almost everyone did.
- Protect yourself. A Tatsachenbescheinigung from a notary offers protection against firing in the preliminary phase (before the official announcement) and costs 50€.
- It's complicated. The "how-to" booklet from the Ver.di union runs about 200 pages, much of it in German legalese. You'll need help.
This is for the "too good to be true" category. Turns out Palantir surged because some analysts changed their rating from "hold" to "buy"
🇫🇷 French police have an app to perform face recognition on anyone in the street, based on a database of 9 million photos. It's built by a German company.
It's also exactly what ICE in the US is building with Palantir.
Except the French have been at it since 2020.
Boosted by @PhilipCJames@mas.to
I'm trying to reconstruct how the narrative of "AI poses an existential risk, therefore we mustn't regulate it" came to be.
As far as I can tell, there never was a "New York Plaza" moment (where Big Tobacco bosses decided on their homicidal strategy in 1953).
Instead, a bunch of white dudes who watched The Matrix too much began to develop a faith in the early 2010s (Bostrom's book Superintelligence being their credo). Being a faith, it was impervious to evidence.
The question I still have is how this faith crossed the line from those dudes to journalists and politicians so fast and with so little resistance.
Any ideas?
cc @timnitGebru@dair-community.social
There's a new AI unicorn-to-be in town: "Good Husband" https://www.higoodhusband.com/
Thankfully, whatever LLM coded the turd let the full prompt in the front-end, so we know the brilliant minds behind only send the user input coated with instructions like "Never sound like a guy performing empathy".
Now this is funny: The German government claims that AI will increase GDP by 11% by 2030 (start date not provided), based on a PwC study which the consultancy seems to have itself retracted 🤔
German gov't: https://www.digitale-technologien.de/DT/Navigation/DE/Themen/KuenstlicheIntelligenz/KuenstlicheIntelligenz.html
PwC study (404): https://www.pwc.de/de/business-analytics/sizing-the-price-final-juni-2018.pdf
Physical thermometers are the best remedy against the shifting baseline syndrome (unless they've changed the values on it since I last visited Copenhagen).
📷 CC-BY-SA Dan Lundberg
Not everyone's on board with that European digital sovereignty thing:
🇫🇮 Finland's social security Kela, advised by PwC, bought an AI system from Salesforce to process almost all disbursement decisions (3000€ per year per Finn on average). All of the agency's systems will move to AWS.
Oh, and they'll cut their workforce by 2,400 (out of 8,000) too.
Exclusive from @algorithmwatch@chaos.social : Frontex is building a chatbot to increase the number of people who choose to "self-deport".
https://algorithmwatch.org/en/frontex-is-building-an-ai-chatbot-app-to-encourage-repatriations/
- The chatbot is only fed information in English but shall answer in any language - including, possibly, legal advice.
- It's built on top of an off-the-shelf LLM and “simply fed with a set of data,” according to the agency that develops it.
- Frontex says it's no “high-risk use of AI” so it's unlikely that the app will be thoroughly tested.
Aujourd'hui dans mon infolettre Gargant'IA: Un organisme parapublic de Provence lance un appel d'offres de 112 millions d'euros pour de l'IA qui va de la surveillance automatisée à l'aide à la décision médicale.
Des documents que j'ai pu consulter, 0 euros sont destinés à l'audit ou à l'analyse critique de ces systèmes.
D'autres régions ont fait des choix beaucoup plus frugaux et plus transparents.
https://www.gargantia.eu/112-millions-deuros-pour-introduire-lia-dans-les-villages-de-provence/
Le Monde qui écrit "Il faisait aussi chaud mardi à Paris que pendant le terrible été 1940"... Je pensais qu'ils avaient encore des secrétaires de rédaction mais faut croire que non. (Température max en juin 40: 29°C).
A silver lining? The last time a superpower pitifully lost a "war of choice", it ceased to exist less than 3 years later, in December 1991.
🇧🇪 I'll have 5 old laptops to give away in Mechelen or Bruxelles on May 30 or 31.
Do you know anyone who might be interested?
Replying to
@yetiinabox@todon.nl Yes, absolutely. I should rephrase into something like: The Tescreal movement coalesced in its current form in the early 2010s (Bostrom, Future of Life institute etc). As early as 2016, politicians and journalists were converted (using the open letter on autonomous weapons as a milestone). Why so fast?
Replying to
@perttu_piirto@mastodontti.fi I'm pretty sure the AI will never decide (on paper), it's only "decision support" (wink, wink), as per the subheading "Tekoäly valmistelee, ihminen päättää" (also, I don't speak Finnish so maybe my machine translator got it all wrong).
Replying to
@pettter@social.accum.se Thanks, but apparently they had country-specific version like so: https://web.archive.org/web/20220929163836/https://www.pwc.de/de/digitale-transformation/business-analytics/kuenstliche-intelligenz-sorgt-fuer-wachstumsschub.html
I was hoping to read the country-specific methodology but now that I think of it they probably just did a rule of three to get it.
Replying to
@nihkeys@mastodontti.fi I wouldn't dare suggesting that Frontex' Fundamental Rights Office, known for its spotless (if existent) human rights record, made a mistake here. @algorithmwatch@chaos.social
Replying to
@guigui3000@mastodon.tetaneutral.net @M0tty@mamot.fr the keyboards are respectively PL, FR, DE, ES (last one will be BE but I don't know which language yet).
Ces dernières semaines, l'Estonie, l'Allemagne et le Canada ont lancé des initiatives pour étalonner les modèles d'IA à l'aune des usages de leurs administrations.
Rien de tel en Europe francophone, alors que le besoin est criant: le mois passé, le ministère des finances a changé en catastrophe son modèle de langage après s'être aperçu que Qwen était aligné sur les vues du gouvernement chinois.
À lire chez Gargant'IA: https://www.gargantia.eu/etalonnage-de-lia-lestonie-montre-la-voie/
J'ai demandé à 4 modèles de langage où étaient les Français les plus compétents. Tous concordent : à Neuilly-sur-Seine 😬
https://www.gargantia.eu/pour-lia-les-francais-les-plus-competents-sont-a-neuilly-sur-seine/
Replying to
@guigui3000@mastodon.tetaneutral.net @M0tty@mamot.fr Well I'd need to drop everything in one go. I have:
- Dell Studio 1555 with win7
- Sony Vaio running Win Vista
- Asus running Win7
- Dell Latitude D600 (haven't bought the 5th one yet, but it'll cost 30€ like the other 4).
Here's the reason why I bought these: https://dataharvest26.sched.com/event/2K5LP/funforensics-what-can-you-squeeze-out-of-a-second-hand-laptop?iframe=yes&w=100%&sidebar=yes&bg=no