Apertus | Swiss AI

Switzerland has developed an open set of models, Apertus | Swiss AI, that is trained on a documented training set, “developed with due consideration to Swiss data protection laws, Swiss copyright laws, and the transparency obligations under the EU AI Act.”

EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS) has released Apertus, Switzerland’s first large-scale open, multilingual language model — a milestone in generative AI for transparency and diversity. Trained on 15 trillion tokens across more than 1,000 languages – 40% of the data is non-English – Apertus includes many languages that have so far been underrepresented in LLMs, such as Swiss German, Romansh, and many others. Apertus serves as a building block for developers and organizations for future applications such as chatbots, translation systems, or educational tools.

This project should interest us in Canada as we are talking Sovereign AI. Should Canada develop its own open models? What advantages would that provide? Here are some I can think of:

  • It could provide an open and well maintained set of LLMs that researchers and companies could build on or use without fear that access could be changed/pulled or data logged about usage.
  • It could be designed to be privacy protecting and to encourage adherence to relevant and changing Canadian laws and best practices.
  • It could be trained on an open and well documented bilingual data that would reflect Canadian history, culture, and values.
  • It could be iteratively retrained as issues like bias is demonstrated to be tied to part of the training data. It could also be retrained for new capacities as needed by Canadians.
  • It could include ethically accessed Indigenous training sets developed in consultation with indigenous communities. Further, it could be made available to indigenous scholars and communities with support for the development of culturally appropriate AI tools.
  • We could archive code, data, weights, documentation in such a way that Canadians could check, test, and reproduce the work.

I wonder if we could partner with Switzerland to build on their model or other countries with similar values to produce a joint model?

Colloque CRIHN 2025 en l’honneur de Stéfan Sinclair

Stéfan Sinclair
Stéfan Sinclair

L’herméneutique en pratique : hommage au travail et à l’héritage de Stéfan Sinclair – Une conférence sur l’analyse de texte, la création d’outils et les humanités numériques critiques Université de Montréal, 10-12 septembre 2025 Conférencières plénières : Servanne Monjour (Sorbonne Université) et Lauren Tilton (Richmond University) Programme Mercredi 10 septembre 2025 (Atelier optionnel) 9h30-12h30 Atelier « Introduction to […]

I’m at a colloquium in honour of Stéfan Sinclair, Colloque CRIHN 2025 en l’honneur de Stéfan Sinclair. I ran a workshop on Spyral and later gave a talk about “Epistemologica in Practice: Replication and Knowledge in Spyral.”

The colloquium had a number of great talks, but one I particularly liked was by Parham Aledavood (U de Montréal)  titled “Viral Intelligence: Contagion, Mutation, and the Spread of LLMs.” He talked about how the release of ChatGPT in 2022 was discussed as if it were a viral event infecting us with AI. He walked through the “formulaic plot” from infection to containment. I was reminded of other stories of influence(za) like the story at the end of Lucretius’ De Rerum Natura.

Personal Superintelligence

Explore Meta’s vision of personal superintelligence, where AI empowers individuals to achieve their goals, create, connect, and lead fulfilling lives. Insights from Mark Zuckerberg on the future of AI and human empowerment.

Mark Zuckerberg has just posted his vision of superintelligence: Personal Superintelligence. He starts by reiterating what a lot of people are saying; namely that AGI (Artificial General Intelligence) or superintelligence is coming soon,

Over the last few months we have begun to see glimpses of our AI systems improving themselves. The improvement is slow for now, but undeniable. Developing superintelligence is now in sight.

He distinguishes what Meta is going to do with superintelligence from “others in the industry who believe superintelligence should be directed centrally towards automating all valuable work, …”. The “others” here is a poke at OpenAI who, in their Charter, define AGI as “highly autonomous systems that outperform humans at most economically valuable work …” He juxtaposes OpenAI as automating work (for companies and governments) while Meta will put superintelligence in our personal hands for creative and communicative play.

Along the way, Zuckerberg hints that future models may not be open any more, a change in policy. Until now Meta has released open models rather than charging for access. Zuckerberg not worries that “superintelligence will raise novel safety concerns.” For this reason they will need to “be rigorous about mitigating these risks and careful about what we choose to open source.”

Why don’t I trust either Meta or OpenAI company?

Humanitext Antiqua

Here at DH 2025 in Lisbon, Portugal, I heard a paper about a neat Japanese projects, Humanitext Antiqua – ヒューマニテクスト. They allow you to identify an ancient philosophy subcorpus (eg. Plato) and then ask questions of that. I was able to get all sorts of interesting results since they have trained their system on Greek and Roman philosophers like Aristotle and Cicero.

Here is a reference to the project:

Naoya Iwata, Ikko Tanaka, Jun Ogawa, ‘Improving Semantic Search Accuracy of Classical Texts through Context- Oriented Translation’, Proceedings of IPSJ SIG Computers and the Humanities Symposium. Download link: https://researchmap.jp/n.iwata/published_papers/48448512

Feminist Data Manifest-No

Hannah L. Jacobs presented a great paper on “Critical Refusal, Slowness, and Openness: Possibilities and Challenges in Community-Oriented Digital Archival Initiatives” at DH 2025. She talked about refusing to complete a project once they realized they didn’t really have community approval to share their data. She also pointed to this Feminist Data Manifest-No.

There was a great question about whether one can mention in a grant that one wants to go slow and that the community may refuse to be studied. Our grant system rewards and supports innovation, not slow research. I’m reminded of The Slow Professor. Perhaps it is tenure that makes slowness possible, not grants.

Reviews in Digital Humanities

Home page for Reviews in Digital Humanities, a journal edited by Dr. Jennifer Guiliano and Dr. Roopika Risam

Listening to a great talk by Paul Spence and Maria Jose Afanador-Llach (and others) on “Evaluation models, global diversity and DH” I learned about the Reviews in Digital Humanities. This looks like a really useful resource for students.

More generally their paper highlighted the importance of local evaluation models that are appropriate to the community.

It is clear that the MLA Guidelines that I worked on, back in the day (and have been updated), are influential, but they need to be adapted.

Colloque « DH@LLM: Grands modèles de langage et humanités numériques » @ IEA & Sorbonne U

DH@LLM: Grands modèles de langage et humanités numériques Colloque organisé par Alexandre Gefen (CNRS-Sorbonne Nouvelle), Glenn Roe (Sorbonne Université), Ayla Rigouts Terryn (Université de Montréal) et Michael Sinatra (Université de Montréal) En collaboration avec l’Observatoire des textes, des idées et des corpus (ObTIC), le Centre de recherche interuniversitaire sur les humanités numériques (CRIHN), l’Institut d’Études […]

Today I gave a keynote to open this symposium on Large Language Models and the digital humanities, Colloque « DH@LLM: Grands modèles de langage et humanités numériques » @ IEA & Sorbonne U. I didn’t talk much about LLMs, instead I talked about “Care and Repair for Responsibility Practices in Artificial Intelligence”. I argued that the digital humanities has a role play in developing the responsibility practices that address the challenges of LLMs. I argued for an ethics of care approach that looks at the relationships between stakeholders (both individual and institutional) and asks how we can care for those more vulnerable and how can we repair emergent systems.

Brandolini’s law

In a 3QuarksDaily post about Bullshit and Cons: Alberto Brandolini and Mark Twain Issue a Warning About Trump I came across Brandolini’s law of Refutation which states:

The amount of energy needed to refute bullshit is an order of magnitude bigger than to produce it.

This law or principle goes a long way to explaining why bullshit, conspiracy theories, and disinformation are so hard to refute. The very act of refutation becomes suspect as if you are protesting too much. The refuter is made to look like the person with an agenda that we should be skeptical of.

The corollary is that it is less work to lie about someone before they have accused you of lying than to try to refute the accusation. Better to accuse the media of purveying fake news early than to wait until they publish news about you.

As for AI hallucinations, which I believe should be called AI bullshit, we can imagine Rockwell’s corollary:

The amount of energy needed to correct for AI hallucinations in a prompted essay is an order of magnitude bigger than the work of just writing it yourself.

CSDH/SCHN Congress 2025: Reframing Togetherness

These last few days I have been at the CSDH/SCHN conference that is part of the Congress 2025. With colleagues and graduate research assistants I was part of a number of papers and panels. See CSDH/SCHN Congress 2025: Reframing Togetherness. The programme is here. Some of the papers I was involved in included:

  • Exploring the Deceptive Patterns of Chinook: Visualization and Storytelling Approaches Critical Software Study – Roya Sharifi; Ralph Padilla; Zahra Farhangfar; Yasmeen Abu-Laban; Eleyan Sawafta; and Geoffrey Rockwell
  • Building a Consortium: An Approach to Sustainability – Geoffrey Martin Rockwell; Michael Sinatra; Susan Brown; John Bradley; Ayushi Khemka; and Andrew MacDonald
  • Integrating Large Language Models with Spyral Notebooks – Sean Lis and Geoffrey Rockwell
  • AI-Driven Textual Analysis to Decode Canadian Immigration Social Media Discourse – Augustine Farinola & Geoffrey Martin Rockwell
  • The List in Text Analysis – Geoffrey Martin Rockwell; Ryan Chartier; and Andrew MacDonald

I was also part of a panel on Generative AI, LLMs, and Knowledge Structures organized by Ray Siemens. My paper was on Forging Interpretations with Generative AI. Here is the abstract:

Using large language models we can now generate fairly sophisticated interpretations of documents using natural language prompts. We can ask for classifications, summaries, visualizations, or specific content to be extracted. In short we can automate content analysis of the sort we used to count as research. As we play with the forging of interpretations at scale we need to consider the ethics of using generative AI in our research. We need to ask how we can use these models with respect for sources, care for transparency, and attention to positionality.

Welcome to the Artificial Intelligence Incident Database

The starting point for information about the AI Incident Database

Maria introduced me to the Artificial Intelligence Incident Database. It contains summaries and links regarding different types of incidents related to AI. Good place to get a sense of the hazards.

The AI Incident Database is dedicated to indexing the collective history of harms or near harms realized in the real world by the deployment of artificial intelligence systems. Like similar databases in aviation and computer security, the AI Incident Database aims to learn from experience so we can prevent or mitigate bad outcomes.