"Deployment is not justified unless it is proven that the benefits greatly outweigh the harms."
Damning letter from over 400 scientists from 30 countries (16 EU states) against a social media ban.
csa-scientist-open-letter.org/ageverif-Fe...
"The conversation about AI is a marketplace of competing science fiction narratives" is bang on
x.com/dkthomp/sta...
Bad timing: Europe's AI talent is already fleeing, 100+ companies with €2.4T in market cap are begging the EU to make life easier for AI developers, and delegates at the New Delhi summit say the EU "shot itself in the foot."
Read more:
progresschamber.org/insights/eu...
Article 95(3): AI providers draw up the Codes. Other stakeholders "may" be consulted, not "shall."
Civil society input is valuable, but they're guests. They can't use the Codes working groups to derail the process and rewrite the law. 📜
Non-industry groups derailing the discussions, politicising the sessions, and claiming companies were "watering down" already-settled rules.
Codes are supposed to be industry-led, but the AI Office's workshops were dominated by groups that aren't building AI models or systems.
What's going on? ⚠️ Non-industry actors hijacked the process.
We took part in all three working groups for the Codes on Transparency. What did we see?
The AI Office's draft Code is quietly undoing the tense negotiations over the obligations of AI models vs AI systems.
The draft Code puts certain obligations on model developers, despite the fact that AI models simply cannot comply with the requirements set for systems.
The process for developing the EU's AI Act's Code of Practice on Transparency - guidance for companies trying to understand the Act - is a mess 🧵
Experts in forecasting, biosecurity, and virology overestimate the biorisks posed by AI by almost 5x. Actual risk is the red line v. forecasted risk intervals above.
Interestingly, superforecasters outperformed domain experts.
forecastingresearch.substack.com/p/how-well-...
🎤 Joined session I of The Parliament's Tech, Digital and AI Summit.
🇪🇺🇺🇸 With MEPs Salla & Yon-Courtin, Commission official Brice Allibert, and TPN's Dan Nechita, we spoke about EU-US relations, Europe's Trump cards, and, of course, a few shout-outs for EU-INC!
The revolution will be unglamourously gradual
AI stoking more panic than actually causing damage to Saas
Uber could be to self-driving cars what Nvidia is to AI
"But a growing number of forecasters now say the economy’s dependence on AI was overstated" has to be the most predictable AI event of 2026
www.washingtonpost.com/technology/...
Bringing new meaning to "undercover agents"
www.politico.eu/article/isi...
Australia is the only Full democracy (according to the Economist's index) with a ban. EU states are set to follow ⏳
DeepL CEO warns that AI Act risks becoming a "competitive disadvantage" for Germany and the EU.
Researchers and founders are motivated by scale, "which environment will truly enable them to realise their ambitions and visions?”
open.spotify.com/episode/6QH...
The actual simple fix: stop requiring banners for low-risk tracking, contextual ads, fraud prevention, and audience measurement.
Keep consent requirements where they matter, for truly personalised tracking. Fewer banners, same privacy protections.
progresschamber.org/insights/eu...
Oh, and they're exempting "media services" from the browser setting, so you'll still get bombarded by consent screens from news sites. New Media like blogs and Substacks? No exemption.
The Commission's solution moves consent to browser settings. One click applies everywhere. Sounds convenient until you realise users will likely choose a blanket "refuse all" at the browser level. That kills cookies for the services that rely on them to stay free.
Cookie banners exist because the EU requires "informed consent" for data collection. Turns out 76% of users find pop-ups irritating.
🍪🇪🇺 The EU finally admits cookie banners are broken. Users waste 600 hours a year clicking through them, and the Commission says, "This is not a real choice."
So what's their fix? A browser-level setting that will break the ad-funded internet 🧵
Only 2 more years before one of the most hysterical takes on AI is proven false
By Harari, Harris, and Raskin www.nytimes.com/2023/03/24/...
EU has "shot themselves in the foot with the AI Act"
www.politico.eu/article/wor...
🇮🇳 🤖 Here's what Europe's AI leaders had to say at last week's Summit:
Macron: AI is a "civilisation-scale" innovation
LeCun: AI could boost productivity by about 0.6% per year, which is "actually quite big!"
Mensch: "The people who are not catching that [AI] wave are going to be left behind"
A/: Risk aversion, regulation (esp. labour laws), and lack of reform.
worksinprogress.co/issue/why-e...
Are social media bans, however well-intentioned, junk policymaking?
Those who spend their time advocating for children’s rights and welfare say that bans won’t work, and only rinse responsibilities from policymakers, platforms, and parents 👇
eurochild.org/resource/eu...
🤯 Europe accounts for over half of global industrial robotics exports.
💡 14/20 countries with the highest "robot density" are European
🚗 Deep industrial capacity across automotive, machinery, plastics, F&B
🇪🇺 4/5 world's largest exporters are European (Germany, Italy, Denmark, Sweden)
Why are European tech founders leaving Europe?
1/ Access to capital
2/ Regulatory environment
Several drivers are interlinked: a nimbler business environment attracts capital, customers, and talent.
www.stateofeuropeantech.com/
🇬🇧🇪🇺 Brejoin would be a colossal win for the EU's tech leadership
(graph from @lugaricano's latest: substack.com/home/post/p...)
