Jörg Lehmann

Written together with @amsichani.bsky.social …

A colleague from the UK and I bring in our two cents on a highly divisive issue, written from the perspective of European CHIs and research libraries.

A Position Paper on AI and Copyrights in Cultural Heritage and Research (EU and UK)

doi.org/10.5334/johd.290

#genAI #commons #openness

Two more blog posts on #openness #GLAMs and #opensource

Openness & its shades (of grey)
mmk.sbb.berlin/2024/06/21/o...

Openness & closed systems
mmk.sbb.berlin/2024/06/25/o...

Thus forming a trio of reflections on redefining openness in the 21st century

Brewster Kahle vs. HF reminded me of WorldCat vs. Anna's Archive, one month ago:

www.infodocket.com/2024/02/07/r...
Mass scraping of bibliographic metadata from WorldCat ...

... obviously, we have (again) to become more clear of what is "open", "public domain", CC0 etc.

Sigh. This topic #openness, intellectual property rights #IPR, #genAI is getting really complicated for #GLAM institutions.
Wrote a blogpost to chart what's up in the EU and what we currently need:

mmk.sbb.berlin/2024/03/13/o...

Currently, there is no technical solution to implement an opt out...

New post on the "power hungry magic" of contemporary artificial intelligence published on the blog of the HumanMachineCulture project:
Energy, CO2 intensity and sustainability as mostly overlooked issues in the deployment of GPTs.

mmk.sbb.berlin/2024/01/26/p...

#LLMs #ChatGPT #metaverse

Copyright is but one indicator of the value of digital texts, which have gone through a quality filter called ‚publishing houses‘. The same can apply to texts in the public domain, and GLAM institutions should reflect on this. Texts in open access are as well valuable, see
doi.org/10.54900/zg9...

Dutch National Library restricts access for commercial AI
Blocking is done via the robots.txt. Crawlers are thus excluded regardless of copyright. Consequently, public domain material is not accessible to the crawlers. Restriction is selective: Googlebot-image, dataforseo.com, GPTBot, ChatGPT-User

New post "Feeding the cuckoo" published on the blog of the MMK project, focusing on privacy issues in large language models, especially Google's Bard (my friend, the poet).

mmk.sbb.berlin/2024/01/12/f...

#LLMs #privacy #ChatGPT #ethics #elsi

Power Hungry Processing: Watts Driving the Cost of AI Deployment? Recent years have seen a surge in the popularity of commercial AI products based on generative, multi-purpose AI systems promising a unified approach to building machine learning (ML) models into...

Power Hungry Processing

Luccioni, Jernite & Strubell, November 2023

"the most efficient text generation model uses as much energy as 16% of a full smartphone charge for 1,000 inferences, whereas the least efficient image generation model uses as much energy as 950 smartphone charges (11.49 kWh)"

I wrote a blogpost on LLMs and anthropomorphism for the blog of our project:

mmk.sbb.berlin/2023/12/20/h...

People who are a bit lonely before Christmas may want to read it…

Datasheets for Digital Cultural Heritage Datasets:

doi.org/10.5334/johd.124

What are the characteristics of digital cultural heritage datasets? How would dataset documentation look like?
We formulate a series of recommendations and propose a datasheet template, see:
doi.org/10.5281/ZENODO.8375033