Michiel Bontenbal
banner
mpbontenbal.bsky.social
Michiel Bontenbal
@mpbontenbal.bsky.social
Complex beings as us humans can not be summarised in a few lines but I am here for #AI #climate #EUtech #EU_politics. And geopolitics, unfortunately. Lecturer in AI & IT. Some posts in Dutch. 🇪🇺
Reposted by Michiel Bontenbal
🌱Very proud of our team's latest release 😊 meet Tiny Aya, a massively multilingual model with 3.35B parameters.

Tech report here: github.com/Cohere-Labs/...
February 18, 2026 at 2:16 AM
Reposted by Michiel Bontenbal
More embedding models and an even more reliable inference engine is what you get with @hf.co Text Embeddings Inference v1.9.0 💥

More in the thread 🧵
February 17, 2026 at 4:05 PM
In my view it comes down to this: the top 3-5 models will be proprietary as people are willing to pay for the best models.

Below that group there is a big number of open weights models.

Not SOTA but in many cases good enough and with excellent cost/benefit ratio.
Open models are in a perpetual race to stay relevant at the frontier. While they're doing better than I, and many experts would expect given the cost of models, I don't see evidence that open models are accelerating and surpassing the best closed models.
www.interconnects.ai/p/open-model...
February 17, 2026 at 6:48 PM
Reposted by Michiel Bontenbal
👏 Ukraine regains over 200 sq km in five days – AFP
Ukraine regains over 200 sq km in five days – AFP
Ukraine’s defence forces regained 201 square kilometres of territory between 11 and 15 February, marking their fastest advance in more than two and a half years, according to AFP analysis of ISW data.
www.pravda.com.ua
February 17, 2026 at 6:33 AM
Europe: we should regulate AI

Pentagon: No, don’t you dare to self-regulate AI.
EXCLUSIVE: The Pentagon is considering severing its relationship with Anthropic over the AI firm's insistence on maintaining some limitations on how the military uses its models.
Exclusive: Pentagon threatens to cut off Anthropic in AI safeguards dispute
Anthropic has not agreed to the Pentagon's terms and defense officials are getting fed up after months of difficult negotiations.
www.axios.com
February 16, 2026 at 8:37 PM
Reposted by Michiel Bontenbal
It's available here: github.com/ademasi/ref-... Full disclosure, it was made with Claude. The code is readable. I tested it with less than 5 papers. Always check the outputs ;)
GitHub - ademasi/ref-checker
Contribute to ademasi/ref-checker development by creating an account on GitHub.
github.com
February 16, 2026 at 10:20 AM
Reposted by Michiel Bontenbal
The distinction between grassroots activism and automated influence is collapsing.

We typically worry about "bot farms" or "troll factories." But a distinct, more complex threat is emerging. We call it "Cyborg Propaganda."
February 16, 2026 at 7:59 AM
“Met alleen regulering (van Big Tech) redden we het niet, daarvoor is de machtsverhouding te scheef”

En dat van een hoogleraar Recht.

Ja, als je de toekomst wilt vormgeven, moet je ‘em zelf bouwen.
Europa zit in een noodsituatie, waarschuwt Reijer Passchier, hoogleraar en auteur van ‘De vloek van Big Tech’. Amerikaanse techreuzen overtreden keer op keer de regels, terwijl Europa te afhankelijk is om in te grijpen. Dat maakt politieke moed nu belangrijker dan ooit.
Podcast | Hoogleraar Reijer Passchier: ‘Onze onmacht tegenover big tech bedreigt de soeve­reiniteit van Europa’
In ‘Follow the Money interviewt’ spreken journalisten Annette van Soest en Harm Ede Botje met sleutelfiguren uit verschillende FTM-dossiers. Van toezichthouders en wetenschappers tot politici en juris...
www.ftm.nl
February 15, 2026 at 2:07 PM
Reposted by Michiel Bontenbal
if bluesky was a short film: www.youtube.com/watch?v=mEVl...

(very good short!!!)
That's AI (2026) - Short Film
YouTube video by maximise
www.youtube.com
February 13, 2026 at 11:12 AM
Reposted by Michiel Bontenbal
We’ve been preparing for AI as an Oracle—one brain that answers our questions. But what if intelligence emerges from interaction? What if the Singularity isn’t so singular?
February 12, 2026 at 9:27 PM
Reposted by Michiel Bontenbal
Welcome Euro-Q-Exa, Europe’s latest quantum computer.

Hosted at the Leibniz Supercomputing Centre in Munich, Euro-Q-Exa is the third EuroHPC JU quantum computer to come online.

Strengthening Europe’s quantum ecosystem and supporting researchers & innovators across Europe.
February 12, 2026 at 5:38 PM
Reposted by Michiel Bontenbal
Quite the chart
February 11, 2026 at 7:44 PM
Reposted by Michiel Bontenbal
Our grad-level "Deep Learning" course (MIT's 6.7960) is now freely available online through OpenCourseWare: ocw.mit.edu/courses/6-79...

Lecture videos, psets, and readings are all provided.

Had a lot of fun teaching this with @sarameghanbeery.bsky.social and @jeremybernste.in!
February 11, 2026 at 5:52 PM
Reposted by Michiel Bontenbal
We are announcing "Python Unplugged on PyTV"!

Join us for a free online conference bringing together the leading voices of the #Python community.

📺 Live on YouTube
🗓 March 4, 2026
⏰ 11:00 am – 6:30 pm CET

The best of Python, straight to your living room. Save the date: https://jb.gg/ztocbe
February 11, 2026 at 2:01 PM
Reposted by Michiel Bontenbal
We have published the #CFP for our AI and Open Government Workshop, to be held colocated with the International Conference on AI and Law (ICAIL) in Singapore.

Read more:
graus.nu/research/cal...

#AI #OpenGov
Call for Papers: AI & Open Government Workshop at ICAIL 2026 • David Graus
We’re excited to announce the AI & Open Government Workshop (AIOG), co-located with ICAIL 2026 in Singapore on June 8, 2026. The call for papers is now open! Together with Graham McDonald (University ...
graus.nu
February 11, 2026 at 7:07 AM
Reposted by Michiel Bontenbal
"Crying girls will make you rich."

That's the actual pitch from a marketer selling packages of emotional videos that brands use to flood TikTok with undisclosed ads.

I just published an investigation into industrial UGC campaigns. What I found is troubling. 🧵 indicator.media/p/crying-gir...
“Crying girls will make you rich”: How marketers trick users with mass-produced "organic" content
Indicator analyzed UGC campaigns that generated over 2 billion TikTok views and found rampant violations of disclosure rules.
indicator.media
February 10, 2026 at 2:54 PM
Trots op mijn masterstudenten van de HvA Master Applied Artificial Intelligence!

www.linkedin.com/posts/hva-ma...
Afgelopen week was het einde van Blok 2. Studenten presenteerden hun eindproject over het vereenvoudigen van overheidscommunicatie met behulp van AI, met een focus op gemeenten. Het was mooi om te… | HvA - Master Applied Artificial Intelligence
Afgelopen week was het einde van Blok 2. Studenten presenteerden hun eindproject over het vereenvoudigen van overheidscommunicatie met behulp van AI, met een focus op gemeenten. Het was mooi om te zien waar ze zijn uitgekomen. De oplossingen liepen sterk uiteen, met verschillende benaderingen en keuzes. Minstens zo waardevol was de groei die ze hebben doorgemaakt tijdens het proces, in hun manier van denken, onderbouwen en presenteren. Goed gedaan allemaal. Jullie mogen trots zijn op dit werk. Dank aan OVER-gemeenten, en in het bijzonder aan Annamaria van Teeseling , Franny van der Bijl en Michael Schulze, voor hun aanwezigheid tijdens de final expo en voor de fijne samenwerking en waardevolle inzichten gedurende het traject. Ook veel dank aan de gastsprekers, docenten en coaches voor hun betrokkenheid, feedback en begeleiding gedurende dit blok.
www.linkedin.com
February 10, 2026 at 2:24 PM
Reposted by Michiel Bontenbal
Oh, here comes another wave of insanity.

Ukraine is now supposed to surrender in a war it has not lost and willingly ensure its own destruction by Russia because Trump needs to avert a midterm disaster.
February 8, 2026 at 7:52 PM
De #IKEA chatbot is toch ook echt uit het jaar nul. Dit zouden mijn studenten beter kunnen.

#AI #fail
February 8, 2026 at 7:14 PM
Reposted by Michiel Bontenbal
Chinese farmer loads pig onto drone, accidentally flies into power line, causes blackout
February 7, 2026 at 4:38 PM
How it started vs How it's going.

It's been less than a year when @karpathy.bsky.social, one of the best in #AI, coined the term #vibecoding.

Now, he is not emboldend but he feels 'behind in programming'.

If he already feels like this, what do I tell my students? What should I teach them?
February 6, 2026 at 11:23 AM
Great paper: "The Platonic Representation Hypothesis"

Neural networks, trained with different objectives
on different data and modalities, are converging to a
shared statistical model of reality in their representation spaces.

arxiv.org/abs/2405.07987

I made a video of Plato's Cave for my class
February 5, 2026 at 7:57 PM
Reposted by Michiel Bontenbal
Submarine data cables, carrying 99% of intercontinental internet traffic, are the backbone of modern life and the European economy 👩‍💻⚙️🛠️

In the face of increasing risks to this critical infrastructure, we are stepping up our efforts to enhance our security and resilience by
⬇️
February 5, 2026 at 2:32 PM
Reposted by Michiel Bontenbal
‪Gisteren weer een gezellige en volle DIY Sensemakers Lab avond in de maakplaats @obamsterdam.bsky.social
Op 18 februari verzorgt Paul Stefaan Mooij een workshop DIY Smart Glasses. @mpbontenbal.bsky.social mocht vast testen🤩
Kom je ook experimenteren?
sensemakersams.org/events/
Events Archive - SensemakersAMS
sensemakersams.org
February 5, 2026 at 8:24 AM
Reposted by Michiel Bontenbal
To understand how the Transformer came to be, we need to understand why ML researchers focused their attention on, well, attention. Part II of the primer on LLMs and Transformers, on RNNs and the Era Before Attention. The WHERE MACHINES THINK Substack. wheremachinesthink.substack.com/p/rnns-and-t...
RNNs and the Era Before Attention: Part II of Primer on LLMs and Transformers
Well before attention became a thing, deep learning researchers were focused on recurrent neural networks for processing sequences—but there was an elephant in the room, a huge bottleneck
wheremachinesthink.substack.com
February 5, 2026 at 2:24 AM