Peter Binkley

@pbinkley@code4lib.social
1.1K Followers
802 Following
3.4K Posts
a digital scholarship technologies librarian, interested in stuff like IIIF and minimal computing, also pre-digital history of similar stuff like microfilm and file cards. In Canada.
⚒️ Dig where you stand. ⚒️
#DigitalHumanities #IIIF #libraries #microfilm #RetroTechnology
Githubhttps://pbinkley.github.io/
Research bloghttps://www.wallandbinkley.com/rcb/
ORCID iDhttps://orcid.org/0000-0003-0560-8958
Annals of Clevelandhttps://www.wallandbinkley.com/projects/2019/annals-of-cleveland/
All men tragically great are made so through a certain morbidness. Be sure of this, O young ambition, all mortal greatness is but disease.

OK, Ollama and Magistral and I are getting along better. I've made a model file with

PARAMETER num_ctx 16384
PARAMETER num_predict 16384

to give it more memory for input and output, and I've put a lot of background knowledge for the task in the system prompt. And I've remembered to invoke the new version and not the default one. The development is still experimental, and Magistral's explanation of its reasoning is very helpful so far. We'll see how long the honeymoon lasts.

We’re on a service review meeting with EBSCO and it is so bad I am stepped away for a few minutes because I cannot listen to them and don’t want to say something really bad.
Things LLMs and I have trouble with: 1) I say "write this thing", they hear "give me a structured approach to help me draft this thing". 2) I say "read this data and tell me about it", they hear "hallucinate a bunch more data more or less like this".

Cuts at Ontario colleges leading to 600 programs cut, nearly 10,000 job losses, union says

« Union says it's 'one of the largest mass layoffs in #Ontario's history' »

« The document showed 23 of 24 colleges in Ontario have reported a 48 per cent decrease in first-semester enrolment of international students from September 2023 to September 2024. »

https://www.cbc.ca/news/canada/toronto/ontario-college-layoffs-1.7581037

Cuts at Ontario colleges leading to nearly 10,000 job losses, union says | CBC News

The Ontario Public Service Employees Union says it's 'one of the largest mass layoffs in Ontario's history.'

CBC

Canada needs to reduce its dependence on US tech companies — and cloud infrastructure is a perfect place to start.

For the Canadian Centre for Policy Alternatives, I explain the threat posed by our dependence on US-owned data centres and why we need a public cloud that isn’t shaped by the pressure for shareholder profits.

https://www.policyalternatives.ca/news-research/canada-should-build-public-cloud-infrastructure-rather-than-relying-on-u-s-tech-giants/

#tech #cloud #datacenters #cdnpoli #cdntech #canada #markcarney

Canada should build public cloud infrastructure rather than relying on U.S. tech giants - CCPA

During the 2025 election campaign, prime minister Mark Carney made a striking admission about Canada’s dependence on U.S. tech companies. The Canadian government was in the process of choosing a partner for a 25-year cloud computing contract for the federal government. The contract’s length should have set off red flags of its own, but the…

CCPA -
There was an LLM
Who was an LL gem
And did things you'd think were impossible
When it was right it was very very right
And when it was wrong it was plausible
If I could do that with a prompt, with added intelligence (ids are sequential, dates are sequential within a section, etc.), it would make life easier (I did one volume; there are lots more). Sample output: https://www.wallandbinkley.com/projects/2019/annals-of-cleveland/headings/animals/ ; code: https://github.com/pbinkley/annals-of-cleveland/blob/main/lib/abstract.rb ; a conference poster about my project: https://doi.org/10.7939/r3-5shd-gp89
Animals

No subheadings See also: Abstracts: 3 Abstracts 11 Cleveland Morning Leader, 31 March 1864 (ed; 3 inches) ~ See original p.2, col.4 ~ View at ChronAm 11 - L. Mar. 31; ed:2/4 - Dr. MacGowan, surgeon in the U. S. "army, con- templates the introduction of two cold-enduring animals of northern China, the Thibetan goat and the yak of the Himalayas. They may be naturalized in corresponding climates in this country.

Five years ago (?!) I worked on parsing metadata out of the OCR text of a 1930s newspaper index, with references like this: "11 - L. Mar. 31; ed:2/4", parsed with regexes like this:

([a-zA-Z]+)[\\.,]?\\s # 'H' newspaper
(\\S+)\\s # 'Dec.': month
(#{OCRDIGIT}+)#{OCRCOLON}+\\s? # '2:' day
([a-zA-Z]*)#{OCRCOLON}?\\s? # 'ed' type (ed, adv)
(
(?:
#{OCRDIGIT}+[/\"']#{OCRDIGIT}+ # '2/3' page/column
(?:[,-]#{OCRDIGIT}+)?
(?:,\\s)?
)+
)
(.*)$ # ',2' or '-3': sequence or range

A good example of the kind of AI project I like: the AI is used in a one-time process to populate a traditional reference tool or data set https://www.archipanion.com/en/blog/ai-cataloguing-registry-finding-aid-palatinate/ - people made the original metadata, people make and use the resulting dataset, AI packs the boxes and drives the moving van
Archipanion: Erschliessung und Zugänglichkeit in Deinem Archiv - unterstützt durch Machine Learning.

Archipanion ermöglicht Dir eine automatische Tiefenerschliessung Deiner Multimedia-Inhalte. Deine bisher nur rudimentär erschlossenen und schwierig zugänglichen multimedialen Inhalte werden einfach durchsuchbar. Aus digitalisierter Information wird nützliches Wissen!

×