FACT symposium at ACMI, view from audience to speakers on the stage.
FACT symposium at ACMI, view from audience to speakers on the stage, people holding their arms up.
FACT symposium at ACMI, attendees talking in the break, in front of screen displays.
FACT symposium at ACMI, view from audience to speakers on the stage.
👉 Register for the FACT (Future of Art, Culture & Technology) Symposium, Workshops & Site Visits, 12-14 Feb 2025
#FACT2025 #Melbourne @acmimuseum.bsky.social
www.acmi.net.au/whats-on/fac...
The Scottish Sunday Herald TV guide featured this preview of the Trump inauguration.
It’s fucking brilliant.
Governments bending over backwards to find ways to allow AI companies to use copyrighted material for free in order to make corporate profits - but libraries have to pay millions to access the same material for educational purposes? Something’s broken here, for sure. 📚
The myriad ways in which far too many humanities scholars run roughshod over anything like GLAM (galleries, libraries, archives, museums) disciplinary expertise is something that routinely drives me, a humanities / cultural studies person who also works in GLAM, absolutely up the wall.
A Navajo rug with a complex pattern with muted reds, pinks and blues. The pattern consists of various vertical and horizontal rectangles with stripes. Around the border are small alternating black and colored rectangles. The weaving is mounted in a wooden frame and hanging on the museum wall.
I recently saw an amazing Navajo rug at the National Gallery of Art. It looks abstract at first, but it is a detailed representation of the Intel Pentium processor. Called "Replica of a Chip", it was created in 1994 by Marilou Schultz, a Navajo/Diné weaver and math teacher. 1/n
Agree. Yet to find an example of a library (or other GLAM) that has a collection strategy or policy that states this outright. Most collection statements could've been written 40 years ago with "digital" sprinkled through, still very traditional format centric.
A key point here about libraries. One of the investment cases for libraries/archives/museums is that forms of knowledge are changing - the special collections we should be collecting now for humanities & social science research are digital: web/social media/personal digital collns, highly ephemeral.
“The novel is a form which deals with change in human personality and human society, bringing to surface those values, patterns of conduct and dilemmas, psychological and technological, which abide within the human predicament.”
bsky.app/profile/math...
Not that Google was ever good, but this cartoon is a good summary of now.
I love that Apple is trying to do privacy-related services, but this just appeared at the bottom of my Settings screen over the holiday break when I wasn’t paying attention. It sends data about my private photos to Apple.
Season’s greetings everyone 🍒 the Fantastic Futures 2024 conference videos are up.
www.nfsa.gov.au/fantastic-fu...
An array of bright lights to explore. 🌟 ✨
The transcriptions are up too and some have errors (yet to be corrected). Feel free to point these out! #AI4LAM #DigitalHumanities
An incredible collection! Congratulations to everyone!
#dh
#digitalhumanities
#glam
On Dec 3, I'll be a panelist at the Turing Institute on AI & Filmmaking for the Clapham International Film Festival. They'll be showing Because of You (w/ @evijit.io) & I'll discuss the ramifications of AI usage in art with Jennifer Ding and Leo Crane. Free (in person & on Zoom) w/ registration!
#DHmakes 🔥
How money to pay lawyers changes your experience of work in this world.
www.theguardian.com/music/2024/n...
🗣️ Session 6 Institutional Adoption, Lizabeth Johnson (LANL), Barbara Shubinski (RC), Kara Kennedy, Karen Thompson (UMELB), Neil Fitzgerald (BL), hosted by Jess Herrington (ANU) #ff2024 #ai4lam
https://buff.ly/4h3fgI8
https://buff.ly/3NpMnZi
https://buff.ly/403YiDA
https://buff.ly/3NkyMCq
🗣️ Session 6 AI and Institutional Adoption, talks by Laura McGuinness (LANL), Kate Follington & Asa Letourneau (PROV), Rachel Senese Myers (GSU), hosted by Jess Herrington (ANU) #ff2024 #ai4lam
https://buff.ly/3zLTLel
https://buff.ly/3zWAJBQ
https://buff.ly/4h2J8nU
🗣️ Session 6 AI and Institutional Adoption, talks by Neil Fitzgerald (BL) and Francis Crimmins (NLA), hosted by Jess Herrington (ANU) #ff2024 #ai4lam
https://buff.ly/3XYppNA
https://buff.ly/3U8xqye
🗣️ Session 5 AI and Culture & Interpretation, talks by Morgan Strong (QAGOMA), Miguel Escobar (NUS), Katrina Grant (USYD) & Svetlana Koroteeva (NLNZ), with Rebecca Coronel (NFSA) #ff2024 #ai4lam
https://buff.ly/404bU1w
https://buff.ly/487iqXf
https://buff.ly/3ZX3Nnr
https://buff.ly/3NpbjzZ
🗣️ Session 5 AI and Culture & Interpretation, panel session with Lauren Booker, Jason de Santolo (UTS), Robin Wright (DPC), facilitated by Kirsten Thorpe #ff2024 #ai4lam
Ready for a big week of #GLAMR & #DH discussion on using AI in heritage collection and HASS research
https://buff.ly/401ABeU
Keep up by using #ff2024 and #ai4lam on socials -
https://buff.ly/406cls7
https://buff.ly/403FnbL
Plus, #ff2024 channel in AI4LAM slack
https://buff.ly/3zMrYuf
Tickets for the dinner on Thursday evening (17 Oct) following day 1 of the Fantastic Futures conference have reopened. Spaces available for more conference goers and +1.
From this point forward, we concentrate on our most effective strategy: a classifier trained on intermediate LLM representations within the exact answer tokens, referred to as 'probing classifiers' (Belinkov, 2021). This approach helps us explore what these representations reveal about LLMs. Our demonstration that a trained probing classifier can predict errors suggests that LLMs encode information related to their own truthfulness. However, we find that probing classifiers do not generalize across different tasks (Section 4). Instead, generalization occurs only within tasks requiring similar skills (e.g., factual retrieval), indicating "skill-specific" truthfulness features. For tasks involving different skills, e.g., sentiment analysis, these classifiers are no better-or worse-than logit-based uncertainty predictors, challenging the idea of a "universal truthfulness" encoding proposed in previous work (Marks & Tegmark, 2023; Slobodkin et al., 2023). Instead, our results indicate that
It’s not a simple fix for hallucination, though, because different types of untruthfulness are associated with different internal states. Or as the authors put it “LLMs encode multiple distinct notions of truth.”
Against AI Narratives Wed 16 Oct 2-4pm - hosted by Rasa Bočytė
https://buff.ly/4dAJEXr
Traces and Places: empowering GLAM voices to shape AI futures Wed 16 12-2pm - hosted by the ANU School of Cybernetics
https://buff.ly/4dAbt2a
Hands on with IIIF & Glycerine Workshops Wed 16 9am-4pm - hosted by Katrina Grant
https://buff.ly/3YbuIuc
NFSA Acton Digitisation Tour Wed 16 Oct 11am-12pm - hosted by Gordon MacPhail
https://buff.ly/4dseTnu
Countdown to the week of the Fantastic Futures 2024 conference. There are still a few spaces left on tours, workshops and satellite events to pack into your week.
Futile Futures Tue 15 Oct 5-8pm - hosted by Adam Bell
Hands on with IIIF & Glycerine Workshops Wed 16 9am-4pm - hosted by Katrina Grant
https://buff.ly/3YbuIuc
NFSA Acton Digitisation Tour Wed 16 Oct 11am-12pm - hosted by Gordon MacPhail
https://buff.ly/4dseTnu
