@nylanderjens Artikeln du länkar säger exakt det jag skrev. Riksarkivet har fått anslag för digitisering i syfte att möjliggöra forskning.
De andra satsningar som listas upp är anslag/uppdrag som getts till andra myndigheter.
Det står inget om att Riksarkivet ska upphandla AI-tjänster.
@abemurray See the following reports/studies for a more complete picture of the problematic nature of Google Books' exclusivity clauses in their collaborations with libraries:
https://t.co/u0kHp5EAl6
https://t.co/rBkLSaM3XL
@abemurray Google thus becomes the only company with centralized access to large parts of the world's cultural heritage. It blocks libraries from centralizing their public-domain digital copies, and it blocks other commercial entities from competing through non-commercial license clauses.
@i619619@antirez What are you referring to, more specifically, regarding Apertus? The effort expended to stay in compliance when it comes to data? Or something else?
@Nouamanetazi Will nanotron have CUDA 12.4 as a dependency for all dataloader types or only the newly added nemo/nvidia ones? This nvcc check seems pretty strict in the new slurm launcher script.
https://t.co/GXuvGMTd59
@Nouamanetazi@eliebakouch Wasn't sure whether it was appropriate to start an issue for something that hasn't been merged into the library yet, so left a comment in the pull request earlier asking whether anyone had trained with tp=4 (as opposed to just generating predictions).
@natolambert@zhangchen_xu That thread links to a policy for Google services which do not seem relevant nor applicable to Gemma. The Gemma model is not on the list of google services: https://t.co/NfM4JntTqL
So that thread essentially links to a policy that doesn't apply to Gemma.
@SaintSnorlax "Lowering the barrier to entry" is just game designers chasing a unicorn. All RTS game that strive towards meaningful longevity are doomed to suffer the exact same problems. The balance between micro and macro is a choice, but the expended attention remains constant.
@SaintSnorlax I don't get the logic behind claiming this will be less stressful than Stormgate or Blizzard RTS games unless one believes micro is easier on a fundamental level. People will just shift all their available attention to micro: constant splitting, prodding, harassing.
@JonKrohnLearns@hadleywickham@posit_pbc Does Posit still have plans of developing an online collaborative editing platform for Quarto, similar to the role that Overleaf has in the LaTeX ecosystem?
It would be a killer app for making web publishing accessible for the masses.
@matei_zaharia@code_star But you publish zero details about your data cleaning? How is one to recreate it if the technical blog is about as (non)informative about data as your average gpt3+ technical report.