Zoom Logo

Deep Learning for Librarians (Core IG Week) - Shared screen with speaker view
Jenny Levine (Core staff, she/hers)
11:49
Hi, everyone - I'm a Core staff member. Just a quick note that our Code of Conduct can be found at http://www.ala.org/core/about/statement-of-conduct . Please contact me through private chat or at jlevine@ala.org if you experience or see any violations of the Code.
Jenny Levine (Core staff, she/hers)
12:44
If you're not already a member of this interest group, you can join it at https://connect.ala.org/core/communities/community-home?CommunityKey=c6ad80f9-1866-4041-801e-a9bccb7c9a72
Lynne Edgar
13:18
is there sound?
Jenny Levine (Core staff, she/hers)
13:27
Yes, John is speaking right now
Lynne Edgar
13:32
Thanks!
Harry Glass
13:40
will the recording be available afterwards
Jenny Levine (Core staff, she/hers)
13:57
Yes, we'll post links to the recording on the IG Week Program Schedule and in the ALA Connect group
Harry Glass
14:18
great
Thomas San Filippo
26:36
this is awesome
Julie Carmen
31:59
Incredible, archival preservation!
Jay Forrest
34:37
Link to the LC Labs Newspaper Navigator: https://labs.loc.gov/work/experiments/newspaper-navigator/
Lisa Bodenheimer
35:16
Sooo, the aim here is not to rely on library-supplied metadata but the image itself?
Adam DeWitt
35:55
It sounds a lot like the AI deep learning project(s) Google is running.
Jay Forrest
36:26
Lisa - yes, we train a subset of the images, and then apply the model to other images - so that library-supplied metadata is not applied to all 16M images
Adam DeWitt
38:41
The implications for bibliographic work - the network collects monographic data from the publisher and constructs metadata without the cataloger having to do anything (in theory). Wow!
Lisa Bodenheimer
39:15
Yes, wow!
Jay Forrest
39:54
BERT is really promsing
Catherine Coleman
40:24
If you are interested in learning more from LAM perspective, join AI4LAM at ai4lam.org
Julie Carmen
40:44
I can see a service within Makerspaces where patrons can bring in their faded diaries, photos, exc... to help create more readable items. Is this a possibility, or too expensive?
Harry Glass
40:48
what hardware were you using
Glenn Bunton
41:38
Does anyone know of library-specific AI and neural net work being done? As mentioned by others, the idea of bibliographic metadata work, for example.
Catherine Coleman
44:46
@Glen. Here is a “look book” of projects done within libraries archives and museums: https://docs.google.com/presentation/d/1iWG9RpPaMlikUAe8mfVlYQeoCiNH8ct2ILFtbMI7P_o/edit?usp=sharing
Glenn Bunton
45:06
Thanks Catherine!
Bela Gupta
51:09
Have any studies been done as to the accuracy of these projects. What is the accuracy rate when thousands of photographs are analyzed by this method and how accurate would it be for library material cataloging & metadata?
Jenny Levine (Core staff, she/hers)
53:54
You can join the interest group by logging in at https://connect.ala.org/core/communities/community-home?CommunityKey=c6ad80f9-1866-4041-801e-a9bccb7c9a72
Richard Fairfield
57:45
Interested in bibliographic aspects
Tienya Smith
59:34
Interested in customer service chatbots or sensing technology applications
Justine Withers
59:36
I recently attended a presentation by Brian Christian on the alignment problem. I think that is an interesting issue that fits into librarians' values and experience: reducing bias, matching the user's intent (as you touched on with chatbots and relevancy algorithms).
Meredith Goins
01:00:30
Interested in project management aspect of AI as well as chatbot applications.
CAROL ZIESE
01:01:44
Interested in a tech literate society, access to new technology, understanding AI effect on society
Meredith Goins
01:01:48
Also interested in mitigating bias in training.
Adam DeWitt
01:01:51
Interested in the implications of AI evolution on society, ethics, and philosophy.
Lisa Bodenheimer
01:02:37
I'm interested in the applications of AI for library bibliographical applications. Also applications of chatbots.
CAROL ZIESE
01:02:43
Ditto what Adam DeWitt said!
Sai Deng
01:04:43
I’m interesting in learning training datasets to supply subject headings and keywords to new metadata.
Sai Deng
01:05:09
… to new records
Rahman Corr
01:06:28
could you comment on getting a pilot project going in a production environment / workplace vis-a-vis IT staff? given the resource heavy nature of AI training (gpu) would you recommend using personal machine or cloud instance
Harry Glass
01:06:45
great discussion
Meredith Goins
01:07:01
Excellent question!
Bela Gupta
01:07:45
Thanks for the great presentation!
CAROL ZIESE
01:09:40
thank you!
Lucy LT
01:09:41
Thank you very much!!!!
A. Houtkamp
01:09:43
thank you for all this information
Deborah Matthews
01:09:44
Thank you!
Eduards Skvireckis
01:09:46
Thank you!
Tienya Smith
01:09:48
Thank you, enjoyed the presentation. I am excited about the group!
Adam DeWitt
01:09:50
Thank you
Meredith Goins
01:09:51
Thank you all!
Paromita Biswas
01:09:54
thank you!
Ann Kutulas
01:09:54
Thank you!!
john butler
01:09:55
Thank you.
Audra Kackley
01:09:57
Thank you
Wendy Johnson
01:10:00
Thank you!