Maybe this speaks to my social and social media networks, but the NDP leadership race isn’t breaking through. I’ve had zero conversations about the race or who should win. I left the party years ago, but I’m still a little surprised how little I’m hearing compared to the last leadership race.
Nailed it
all the way down
Join the Lancaster-Manchester Environmental #DH Seminar on March 11 @ 3pm UK (online) for a talk by @jimclifford.bsky.social & @historyjacob.bsky.social:
"Solving OCR: Using olmOCR to Follow Commodities across the British World"
www.eventbrite.co.uk/e/solving-oc...
#dhist #ocr #envhist 🗃️
(a) a foundational layer showing the changing geography of colonial power; (b) mapping career progression between colonies; (c) a future system to compare how colonial officials write about places and events in contrast to other voices in the archives; (d) who knows…
I'm working on a text mining/knowledge graph project using the Indian Office Lists and Colonial Office Lists are the source material. I've extracted the biographies of Indian Office officials from the 1937 list into a knowledge graph visualized here:
jburnford.github.io/indianoffice...
My biannual reminder, if we don’t agree to peer review requests, the whole system collapses.
And, faculty in at research universities with reasonable teaching loads, need to carry most of the weight. If you or your students want to publish, you need to say yes, even when you are busy.
This feels like a pitch for a Dorchester Review article.
Both are possible and there is value in testing what it can find. But you need to triangulate the answers, follow the footnotes, test the arguments and be really careful.
This problem also shows up when the answer isn’t known. This is where academics and students need to be very careful. Gemini will write a full essay to answer your question whether it finds new evidence pointing to the truth or comes up empty and instead piece a compelling but wrong answer.
This is an important measure of LLMs. The underlying engineering makes them want to answer all questions. But what if the question is nonsense? The examples are extreme, but you could seem subtle misunderstanding in questions steering a discussion off the rails.
github.com/petergpt/bul...
New working paper out: "Open Maps: New Research Directions and Workflows for Digitized Historical Cartographic Material" led by Vincent Baptist & Jules Schoonman (TU Delft) #dh #maps
openmapsmeeting.nl/publications...
Making progress in my efforts to parse all of the articles from the first 8 editions of the Encyclopaedia Britannica.
120,039,792 words; 114,359 articles; 133 volumes.
jburnford.github.io/early_encycl...
There a big problems with all of these companies, but OpenAI is competing hard to rank as the worst AI lab.
www.nytimes.com/2026/02/27/t...
I remember when Canada couldn’t afford to act because of China…
The long-term concern here, to be clear, is not just about researchers, but about enclosure of the knowledge commons.
Schools are paying tech companies for tools that were trained on their own books—while tying their own hands to prevent internal development of similar tools. It's self-destructive.
Building benchmarks is only one way scholars can help steer AI development. We can also measure the effects of AI on students, build better datasets, or tune new open models. Openness itself could be our most important contribution. Universities have huge libraries, and the legal doctrine of fair use should protect models trained on those collections for a nonprofit educational purpose. At the moment, we are not pressing this advantage. Higher education has been so cautious about fair use that the private sector can now train more freely on our libraries (via Google Books) than is possible for academic AI researchers. We need to be bolder: It is our duty to ensure library collections remain open to the public in a form that empowers 21st-century readers. If our intellectual heritage gets enclosed in proprietary tools, we will find ourselves making the same bad bargain we made with scientific publishers, who sell our own research back to us at a steep markup.
We're in a strange situation rn where Google can train freely on books from university libraries—but researchers *at* universities have limited access. I'm optimistic this can be fixed, but if you're in admin or working at a foundation, please know: univs are failing here & resources are needed.
Bitterness and determination to prove that grade 7 teacher wrong powered me through a PhD in history. Have you read Spellbound by Phil Hanley? It is a powerful book (but not appropriate for children as his depression got pretty dark).
For ex., one great “Critical AI” course in lit. would be “Can AI Write a Suspenseful Story?”
Students work together to answer the question. Likely “no,” but they need to get there rigorously.
They’ll learn to use AI (through an API!), and also learn skepticism that lasts because it was earned. +
My grade 7 teacher suggested I should “try harder”. Spelling doesn’t matter in 2026. Keep fighting.
I think AI having mostly (not entirely) very bad critics is a real problem because it means we’ll get political action focused on things that probably don’t matter that much in deferring it’s very real harms.
Ezra is very good here. He calls bullshit and really pushes the Anthropic employee. TLDR: the political system is broken and it is hard to imagine good public policy to confront AI safety or job market disruption.
We also need to think about the implications of AI agents loose on the internet: youtu.be/3n_jKx6v6qU?...
Claude Code has improved a lot in the past few months. I think Ezra's framing is correct here. We need to start thinking about public policy for major disruptions in the job market. And universities are not ready for the disruptions coming out way.
youtu.be/lIJelwO8yHQ?...
As a centrist…
Just starting Beckert’s new book
Google is mad at Chinese AI labs for stealing their intellectual property to build their Gen AI. (Completely different from Google stealing all human knowledge to train AI systems.)
youtu.be/W9WB2xbM5sc?...
Yes, I used Claude code to figure out the batch process. I normally test on ai studio until I’m happy with the prompt and then switch to the api.
