March 21, 2025 • 59 mins
Guests: Presley Dyer, Metadata Librarian at the Georgia Tech Library; Wilhelmina Randtke, Head of Libraries Systems and Technologies at Georgia Southern University; and Mary Beth Weber, Head of Central Technical Services for Rutgers University Libraries.
First broadcast March 21 2025. Transcript at https://hdl.handle.net/1853/77465
Playlist here
"We promise we won’t let the robots take control of the show."
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
[MUSIC PLAYING]
FRED RASCOE (00:02):
How close do you think
we are to just machinesfilling metadata fields?
MARY BETH WEBER (00:07):
Never because there are certain things
that it takes--
it takes skill and knowledgeand thinking about it.
I just think thereare some things that
are just too nuancedfor a machine to handle.
FRED RASCOE (00:20):
I like that.
That was an instant never.
You didn't even haveto think about that.
[THEME MUSIC]
CHARLIE BENNETT (00:49):
You are listening to WREK Atlanta,
and this is Lost in theStacks, the research library
rock and roll radio show.
How about never?
Does never work for you?
FRED RASCOE (00:58):
Never always works.
CHARLIE BENNETT (01:00):
I'm Charlie Bennett in the studio
with the whole gang--
Alex McGee, Molly Gibbons,Fred Rascoe, oh, and me.
Each week on Lost in theStacks, we pick a theme
and then use it to create amix of music and library talk.
Whichever you're herefor, we hope you dig it.
ALEX MCGEE (01:15):
And our show today is called "This Machine Fills
Metadata Fields."
CHARLIE BENNETT (01:19):
I love a Woody Guthrie reference.
ALEX MCGEE (01:20):
Yeah, it is a good one,
but the Lost in the Stackscrew can't really take credit
for the title this time.
Wait a second, I thought oneof you came up with the title.
Whose idea was it?
The answer to that question is abit of an intellectual property
conundrum.
FRED RASCOE (01:35):
I guess we can say it was suggested by ChatGPT.
CHARLIE BENNETT (01:40):
Oh, no.
FRED RASCOE (01:41):
Yeah.
CHARLIE BENNETT: You let a machine (01:41):
undefined
into the production process.
How did it suggest that title?
Was it through theftor hallucination?
ALEX MCGEE (01:49):
Either way, listeners,
we promise we won't let therobots take control of the show.
FRED RASCOE (01:53):
Even if they do occasionally
help us with some things.
And
MARLEE GIVENS (01:57):
That's something we'll
hear more about from today'sguest, a trio of cataloging
and metadataexperts who recently
led an online discussionabout AI and metadata.
FRED RASCOE (02:06):
And our songs today are about hallucinations,
mysteries behind thecurtain, and the limitations
of technological tools.
Librarians havequestions about using
AI like how much labor arewe willing to hand over
to a technology?
CHARLIE BENNETT (02:22):
Uh oh.
FRED RASCOE (02:23):
In my weaker moments,
I'll admit I wish a machinewould just handle everything.
CHARLIE BENNETT (02:28):
Don't say it out loud, Fred.
FRED RASCOE (02:29):
So let's start with a song
about feelingoverwhelmed by the tasks
of daily professional work.
This is "AI, Do My Work for Me"by the Photocopies right here
on Lost in the Stacks.
[MUSIC PLAYING]
ALEX MCGEE (02:49):
That was "AI, Do My Work for Me" by the Photocopies.
This is Lost in the Stacks ontoday's show is called "This
Machine Fills Metadata Fields."
CHARLIE BENNETT (02:59):
We interviewed three librarians
about the AI issues thatconcern the metadata community.
MARLEE GIVENS (03:05):
Those were Georgia Tech's own metadata
librarian Presley Dyer, GeorgiaSouthern University's head
of library systems andtechnologies Wilhelmina Randtke,
who goes by Mina, and the headof central technical services
for Rutgers UniversityLibraries Mary Beth Weber.
So we were really interestedbecause the three of you
(03:26):
recently led what'sknown as an e-forum.
I'll let you explain thee-forum concept in a second.
But you led an e-forumon AI and metadata.
If you described this toone of your colleagues
what you were aboutto do, how did you
describe what an e-forum is?
MINA RANDTKE (03:47):
It's a two-day directed email conversation
where there are topicsfor two-hour or so windows
with a conversation about it,and so someone can be at work
and catch up on itbetween meetings
or participate in away of meeting people
who work in the same fieldwithout having to travel.
FRED RASCOE (04:09):
Is there any reason that it
is an email exchange of ideas?
I've seen similar kindsof exchanges of ideas
on social media, Iguess Twitter or X
or whatever it's called now.
Is there any reasonwhy it's particularly
targeted towards email?
MARY BETH WEBER (04:30):
Because when it was initially started,
that was the mainway to communicate
with a large group of people andit's just continued that way.
MINA RANDTKE (04:39):
Yeah, there have been similar things.
So ALA had had spectrumTwitter chats, which Twitter--
no, I guess now it's X--
those died off.
And so there have been socialmedia things done by ALA
that are similar.
MARLEE GIVENS (04:55):
And was this anyone's first experience
with the e-forum?
PRESLEY DYER (05:00):
It was my first experience.
ALEX MCGEE (05:02):
So even like as a participant?
PRESLEY DYER (05:04):
Mmm hmm.
So, yes.
Yeah, it was all new to me.
So I had to get used to it fromthe participant side of things
and the moderatorside of things.
ALEX MCGEE (05:17):
How did the three of you find each other?
MINA RANDTKE (05:20):
ALA has e-forum coordinators
and they put out a call andI responded to the call.
But someone proposedthe topic, and then they
put out a call whoelse wants to organize.
And I don't actually know whoproposed the topic of first one
of us three.
PRESLEY DYER (05:34):
Yeah, I--
well, it startedactually with Marlee
which she sent it out to GeorgiaTech Library and was just like,
hey, if anybodywould be interested,
I've done it e-forums before andthey are fun to participate in.
So I was just like, oh,well, maybe I should do this.
(05:58):
So I emailed aboutthe e-forum and was
saying that I'm involvedin cataloging and metadata
and how do I begin.
And so I had to send alist of potential topics.
And, of course, AI was the topicthat piqued the most interest,
(06:27):
and so from there, I didan outline of that topic.
And then they sent it outfor volunteers to help me.
MARLEE GIVENS (06:38):
And so, Mary Beth, you
responded to that call.
MINA RANDTKE (06:42):
Mmm hmm.
Yeah, I respondedto the call, too.
FRED RASCOE (06:47):
Was there a reason that this particular e-forum
resonated with you all?
Presley, you wanted to startan e-forum, Mary Beth and Mina,
you wanted to participate init, and the organizing concept
around it was AI in metadata.
Was it because that particulartopic resonated somehow or is it
(07:10):
important in your work somehow?
MARY BETH WEBER: I would say both. (07:13):
undefined
FRED RASCOE (07:15):
And how so?
MARY BETH WEBER (07:17):
We have done some experimenting with AI here
through our vendoras have other people,
and we're always looking forways to streamline our work.
And everybody'stalking about AI.
So we figured it was agood time to step into it.
FRED RASCOE (07:36):
And, Mina, do you have any association with AI
at your place of workat Georgia Southern?
MINA RANDTKE (07:41):
Well, my background is in metadata,
so that was more ofthe interest for me.
And with all of theartificial intelligence
hype that has come up in thepast three to five years,
one of the things I've seenis people-- or businesses
re-branding existingproducts with AI on them
including structured metadata.
So I like to look at howis this tool working,
(08:04):
how are DOIs being used, how isauthority control being used.
So, for example,Research Rabbit or Tool
Like That that's brandedas AI, the interface
is built on structured metadataalthough a lot of times the way
that metadata was assignedreally is very technical.
(08:25):
So little elves do it, but alsocomputers help little elves.
And they use the-- you know.
FRED RASCOE (08:31):
So AI is not magic.
There's metadatabehind the curtain.
MINA RANDTKE (08:36):
Yeah.
FRED RASCOE (08:38):
And, Presley, what was your I guess interest?
Was it just the fact that AIwas a new and interesting topic
in the world ofcataloging and how
AI was going to be integratedin cataloging practice?
PRESLEY DYER (08:54):
Yes.
Our systems anddifferent platforms
are starting to embed itin the library catalog.
And so it started like,oh, well, let me use--
let me try-- test it out bycreating an original record
(09:14):
and then, oh, well, let me seehow this original record looks
against another ChatGPT.
And so it was, oh, well,it just piqued my interest,
and just because it istaking over professions
(09:38):
and daily lives and stuff.
So it's just I would liketo learn more about it.
MARLEE GIVENS (09:45):
We'll be back with more
from Presley Dyer, MinaRandtke, and Mary Beth
Weber after our music set.
CHARLIE BENNETT (09:51):
File this set under TT390.L853.
[MUSIC PLAYING]
Now, Fred.
FRED RASCOE (09:59):
Yeah, sorry.
MARLEE GIVENS (10:08):
That was "Behind the Curtains" by Jan Bradley.
Before that we heard another"The Letter" by Wire,
and we started our setwith "Beyond the Fringe"
by the Lime Spiders, songs aboutrevealing the mysteries behind
the curtain.
[MUSIC PLAYING]
CHARLIE BENNETT (10:27):
This is Lost in the Stacks,
and we are speaking with PresleyDyer, Mina Randtke, and Mary
Beth Weber aboutAI and metadata.
MARLEE GIVENS (10:35):
Presley, Mina, and Mary Beth
recently led a two-daye-forum on that topic,
and we wanted to find out whatAI tools they were already
using before the e-forum.
PRESLEY DYER (10:47):
We use Exlibris products,
so Exlibris has embeddedAI assistant tool
within the metadata editor.
And so it's startedwith that, and then we
have started using AI pops upin everyday tools like Google.
(11:13):
And Amazon has its own thing.
So it's just like I wasjust like, oh, well, I
know other organizationsare trying to incorporate it
in the library board.
So OCLC is about--
is trying to do a startup tohave their own AI assistant
(11:37):
tool.
So, yeah, I hadknowledge, but I wanted
to learn more about what otherswere doing in their libraries.
MARY BETH WEBER (11:49):
We also are an Alma library,
and we volunteered toparticipate in early testing
of the AI metadata assistant.
But some of my colleaguesin the department
also have usedChatGPT for things
like to generate Python scriptsor to help with projects.
And now we have alicense for Copilot,
(12:10):
and we use it for things.
I haven't used it forany cataloging things,
but, for example, I wrotesomething and the editor--
the publisher asked mefor a list of acronyms.
So I could have gonethrough it, and I was lazy.
I asked Copilot get allthe acronyms for me,
alphabetize them.
FRED RASCOE (12:26):
Oh, there are too many acronyms in librarianship
but especially in cataloging.
MARY BETH WEBER (12:31):
Yeah.
FRED RASCOE (12:31):
Cataloging has way too many acronyms.
So I totally sympathizewith you farming that out
to an artificial intelligence.
MARLEE GIVENS (12:40):
Is Georgia Southern also Ex Libris?
MINA RANDTKE (12:44):
We're Ex Libras, but that's--
I actually didn't playwith the tool in Alma.
Presley and I are both involvedin the same statewide migration,
so I've done a lot ofdata cleaning in Alma.
But in that case, what I'vetried to do is look for outliers
and reduce the complexityof data before moving it,
and it's been much morestatistical analysis.
(13:08):
Most of my work withartificial intelligence
has actually been opticalcharacter recognition,
so we don't think aboutit as cutting edge.
But it really islooking at a picture
and detecting it as text.
And so if you think aboutscanning documents and then
having a person writewhat they're about
(13:30):
versus being able to just havethe text taken out of them
and put in a searchable format,the time and cost savings
is so great.
So I worked indigital libraries,
and we've done work withoptical character recognition.
And then here internally,my role at the library
is somewhat like achief information
(13:54):
officer for the library wherethe kind of things I'm doing
is like setting up anintranet and making sure
things are searchable andshow up to the same people.
So just looking atlittle details like that
and maybe havingsmall technical tools
that are establishedbut are fundamentally
driven by machine learning.
MARLEE GIVENS (14:13):
What are your biggest concerns about using AI
for resource descriptionor for metadata management?
MARY BETH WEBER: We're short staffed, (14:23):
undefined
and as people retire orleave, they're not replaced.
So we have a very, verysmall group of catalogers
for a library system our size.
And my fear is something alongthe lines that, oh, well,
we can use that--we wouldn't say it,
but somebody elsein the organization
might say, well, we can useAI to catalog the books.
(14:45):
We don't really need people.
I know that sounds farfetched, but people
have believed hokier things.
FRED RASCOE (14:52):
I've heard--
definitely heardsimilar conversations.
Alex, I wonder as thearchivist in the room have you
heard similarconversations about AI
being able to describearchival objects?
ALEX MCGEE (15:05):
Oh, yeah.
When I was at MIT actually,there was a project,
and it was 2019, 2020 where wewere picking a small collection.
And basically theidea of the project
was we would demonstratethat the AI had biases
and created baddescription was our goal.
(15:26):
Actually, we didn'tget to see it
all the way throughbecause of the pandemic,
but that was the theorywe were testing out
was that it would notcreate ideal description.
So we were actuallydoing women's collections
and then showingthe gaps that AI has
and how to respectfully describemore complicated material.
MARLEE GIVENS (15:46):
Did any of you-- have any of you
in your own workfound anything that AI
seems to be goodat when it comes
to cataloging and metadata?
MINA RANDTKE (15:55):
Well, it does this-- it does optical character
recognition for me.
So I've never typed atranscript of a text document
in my adult life.
PRESLEY DYER (16:08):
I just recently used
it to help with adigitalization project and just
taking pictures ofhistorical documents.
And then it being able tocreate a spreadsheet of what
we were trying to archive.
(16:31):
I feel like it did a greatjob of being able to just do
the spreadsheet side of things.
So, again, it freed us up tobe able to vacuum and actually
clean the material.
I like the fact that it canhelp with the Excel spreadsheet
stuff.
MARLEE GIVENS (16:53):
Are there tasks that it seems to struggle with?
MINA RANDTKE (16:57):
It doesn't have sensors,
so it's just a computer program.
So if anything exists in thereal world, it can't tell it.
And in the library,it's we have three kinds
of materials in our collection.
So we have thingswe've purchased
from vendors, which usually therecords come from the vendor.
Then we have equipment itemsthat get checked out really
(17:20):
frequently like laptops,and it can't look at it
and say, this serialnumber and barcode
don't match because ithas no eyes to look.
And then we have ourarchival materials
and our specialcollections materials,
and it can't look at those.
So it doesn't have sensorsto detect the real world.
MARY BETH WEBER: Some people have (17:39):
undefined
found that it would generatefake contents notes.
FRED RASCOE (17:43):
Oh, that's interesting.
Just like it-- whenpeople do research
and it hallucinates journalarticles that aren't really
there.
What kind of cataloginghallucinations examples
have you seen?
MARY BETH WEBER: Both cases were-- (17:56):
undefined
where people werecataloging things
in languages other than English,and one was a non-Roman script
as well.
And they said thatit generated contents
notes for the booksthey were cataloging,
and they were not even remotelyclose to what was in the book.
And I don't know where they haddrawn that information from,
(18:18):
where as in the AInot the catalogers.
FRED RASCOE (18:22):
Mina, have you ever used--
in your use of optical characterrecognition tools, have you
ever had the tool that you'reusing hallucinate text that
wasn't actually there?
MINA RANDTKE (18:34):
No, it's--
it'll mis-detectletters, but I don't
think I've been to thepoint of it mis-detecting.
And a lot of it is maybeto do an analysis of things
and then make a decision ofhere's a clean up project to do
or here's a direction to go in.
I'm really looking atwhatever tool I'm using.
MARLEE GIVENS (18:56):
You're listening to Lost in the Stacks,
and we'll be back withmore about AI and metadata
on the left side of the hour.
[MUSIC PLAYING]
MARK RIEDL (19:15):
Hi there.
This is Mark Riedl of GeorgiaTech's computer science
department--
or is it?
Maybe my voice hasbeen deep faked,
and this is justa digital forgery
created by a neural network.
Either way, you're definitelylistening to Lost in the Stacks
on WREK Atlanta.
CHARLIE BENNETT (19:38):
Today's show is called "This Machine Fills
Metadata Fields," and if youdon't know why I called that
a Woody Guthrie reference, Isuggest the Google image search
Woody Guthrie guitar.
MARLEE GIVENS (19:49):
Even if this proverbial metadata machine
never comes to be, it istrue that a machine, ChatGPT,
is responsible for theshow title this week.
ChatGPT suggested that as thetitle of a Lost in the Stacks
episode that never happened,an episode that ChatGPT totally
made up.
CHARLIE BENNETT (20:06):
How dare it?
MARLEE GIVENS (20:07):
It's one of many examples of a generative AI
chatbot presenting itsown hallucination as fact.
After we recordedthis interview,
I attended theLibrary 2.0 Conference
on AI and Libraries, Literacy,Ethics, and Responsible Use,
and during the opening keynote,presenter Crystal Trice
addressed the hallucinationsand inaccuracies
(20:28):
that often come up in herconversations with library
staff.
She says--
CHARLIE BENNETT: Everyone assumes (20:33):
undefined
that those are going toget better over time,
and maybe they will.
But really there needs tobe this basic understanding
that the tools are notdesigned to be factual.
These algorithms aren'tdesigned to be factual.
They are designedto be generative,
and that's a real mindshift for library staff.
MARLEE GIVENS (20:52):
She continues--
CHARLIE BENNETT: It's not a bad thing (20:54):
undefined
when we're using AI indifferent aspects of our work,
but it is important to keepin mind that being factual
is not even in its radar.
MARLEE GIVENS (21:04):
While this is mildly unsettling to me
as a reference andinstruction librarian,
I find the metadata communitiesapproach practical, thoughtful,
and forward thinking.
Even if the machine thatfills metadata fields
remains a fantasy, ourcolleagues in technical services
are embracing the futureone tool at a time.
CHARLIE BENNETT (21:24):
File this set under QA76.99.N38.
[MUSIC PLAYING]
ALEX MCGEE (21:42):
That was hallution--
"Hallucinations" by BakerKnight and the Knightmares, k-m,
and then "Mirage" bySiouxsie and the Banshees.
CHARLIE BENNETT (21:54):
Oh, I just realized the difference
in our ages.
ALEX MCGEE (21:56):
How young I am, yeah.
Songs about things thataren't really there.
[MUSIC PLAYING]
MARLEE GIVENS (22:04):
Welcome back to Lost in the Stacks.
We are speaking withPresley Dyer, Mina Randtke,
and Mary Beth Weber, whorecently facilitated an e-forum
about AI and metadata.
In the final segmentof our interview,
we asked them what they learnedfrom that online discussion.
PRESLEY DYER (22:19):
Everybody is voicing the same concerns
like environmentalclimate change, that
was an issue that came up.
The fact that it couldtake over staff positions
And so everyone, it seemedlike had the same concerns,
(22:43):
and it was just really niceto just be in that environment
to actually hear, hey,I'm not the only one
thinking these same things.
And it was justreally nice, too,
the fact that it was alldifferent types of libraries,
too.
There were specializedacademic, public.
(23:05):
We did get a goodvariety in that regard.
MARLEE GIVENS (23:09):
Mary Beth, what did you take away from it?
MARY BETH WEBER (23:11):
I would have to say
it's very similarto what Presley
said that a lot of people arejust at the beginning point.
They don't really understand.
A lot of them hadfears about their jobs.
There was some commentto there's always
feedback after theEA forms, and someone
said it was like the blindleading the blind because nobody
had a firm grip ora defining answer.
MARLEE GIVENS (23:35):
Did you think that maybe some people came
thinking that you were experts?
MARY BETH WEBER (23:41):
Yes, that always happens.
Yeah.
MARLEE GIVENS (23:45):
Mina, did you pick up anything?
MINA RANDTKE (23:48):
For me, what was most striking
was how much concernparticipants expressed
with environmental concernsand environmental impacts
because it wasn'tsomething we had
asked about in the questions.
The e-forum will ask pre-writtenquestions, and none of them
were about that.
But it kept coming outemergent as something
that participants talked about.
FRED RASCOE (24:10):
It sounds like from all three of you,
the general sense from the--
I think the E - forum was twodays of email conversation.
From the two days,the general sense
is that people are interested,lots of reservations
about the utility of it,about replacing labor, about
environmental concerns.
(24:32):
In those two days,were there any outliers
among the cataloginglibrarian community,
people who stood on theirhill and said AI no way.
This is not something thatwill ever be useful to us,
or on the other hand, AIis the future committing
(24:53):
totally no reservations?
Any interestingoutliers like that?
MINA RANDTKE (24:58):
I actually don't remember
any that were acomplete outlier people
had examples of here'ssomething where a vendor used
it to generatemetadata, and it's
had these problems becausethey sometimes a word has
multiple meanings or something.
Or people would expressconcern with their job,
(25:19):
but there weren't really anyoutliers who were completely
against or completely with.
I think they were just lookingfor what tools can we use.
MARY BETH WEBER (25:26):
All those down to Earth level
headed catalogersin the discussion?
Yeah.
MINA RANDTKE (25:34):
Someone opposed might not
have participated, right?
FRED RASCOE (25:38):
That's also very true.
Yeah, sample bias I suppose.
MARLEE GIVENS (25:43):
How do you each feel about AI
when it comes to your own job?
Are you-- do you feellike this is inevitable?
I need to learn it.
Do you think it'sgoing to be helpful?
Are you kind of dubious?
So, Presley, shaking herhead, I'll start with Presley.
PRESLEY DYER (26:01):
I use it every day for something,
creating coding for Alma,the normalization rules.
I use it-- for it to quicklygenerate and spit out the rule
that we'll use tochange the call number
(26:23):
or something like that.
And I am about to help theprofession test it out for OCLC
so it can be used broadly.
So I feel like--
it's just become an everydaypart of my life at this point.
MARLEE GIVENS (26:47):
How about Mary Beth?
MARY BETH WEBER (26:49):
I've worked long enough that I've
gone from a library that hadthe old beehive OCLC terminals
and cards to onlinecatalogs, so I just
see it as a naturalpart of the evolution.
I don't think it's goingto solve everything,
but I think it's atool that can help us.
MINA RANDTKE (27:05):
I'm more like Mary Beth.
It's just another tool.
So each thing isjust another tool.
Maybe even in a researchdatabase, suddenly I'll see.
Now I have a facet forfunding organization,
and it's some kind of entityrecognition that they put in.
Or I'll see maybe animproved search somewhere.
But to me, it's another tool,and it builds on past tools.
FRED RASCOE (27:29):
So you maybe see in just a few years like library
schools will be teaching theirstudents cataloging and teaching
them how toresponsibly use the AI
tools that everyone in thecataloging profession is using?
Or is that too far?
MINA RANDTKE (27:48):
I don't know if they teach cataloging.
That's one of theweaknesses of--
FRED RASCOE (27:52):
That's--
MINA RANDTKE (27:53):
Yeah.
Because libraryschools don't tend
to teach electronicresources management,
don't tend to teach how toteach one-shot instructions
and don't tend toteach cataloging.
FRED RASCOE (28:04):
I may have been out of library school too long,
2004 graduate.
So my knowledge of the currentcurriculum may be out of date.
MARY BETH WEBER (28:13):
And those that teach it,
too, it's not always required.
They teach it here.
They do better when they haveadjuncts teaching the class
because otherwisethey have people
who are only teaching theory.
And theory's great,but there's real life.
So I'm hoping thatthey will teach them
how to use it responsibly.
They have subscriptionsto other things
(28:34):
like classification weband the RDA toolkit,
and the students learnhow to use those.
So maybe that will just beanother thing they show them.
MINA RANDTKE (28:44):
And to me, fundamentally,
information literacy isalready a good framework
for understanding a lot of thetools for content creation.
So understanding sourcesand it's a different world
of librarianship for metadata.
But information literacy.
I think gives thetools to understand
some of the newtechnologies we're
seeing with generated content.
(29:06):
MARLEE GIVENS: Well, this is great. (29:09):
undefined
Thanks for filling usin on AI and metadata
and your e-forum experience.
MARY BETH WEBER (29:19):
Thank you for asking us.
MINA RANDTKE (29:20):
Yeah, thank you for having us.
PRESLEY DYER (29:22):
Thanks, everyone.
FRED RASCOE (29:24):
You are listening to Lost in the Stacks,
and today's episode is called"This Machine Fills Metadata
Fields."
MARLEE GIVENS (29:30):
And our guests were Presley Dyer,
metadata librarian atthe Georgia Tech Library,
Wilhelmina Randtke,head of libraries,
technologies, and systems atGeorgia Southern University,
and Mary Beth Weber,head of Central Technical
Services at Rutgers University.
File this set underRC553.H3 no dot S23.
(29:52):
[MUSIC PLAYING]
FRED RASCOE (30:00):
That was "Blind or Dead" by Lost Cherries.
And we started with "SimpleMachines" by Dancing Cigarettes,
songs about the tools that canhelp us as long as we realize
their limitations.
[MUSIC PLAYING]
MARLEE GIVENS (30:21):
Today's show was called "This Machine Fills
Metadata Fields--
All About the Benefits andLimitations of AI Tools
in Libraries."
We heard about opticalcharacter recognition,
automated spreadsheet creation,and using ChatGPT to find
all the acronyms in a document.
So lately the thing thatI've been relying on AI for
(30:42):
is editing, making something Iwrote fit a certain word count
or summarizing free textanswers from a survey.
So are there AI toolsthat you use at work
or in everyday life, Fred?
FRED RASCOE (30:55):
I guess this is probably
a really low bar forAI, but I do actually
use that in your iPhonewhen you talk into it
and it sends a textmessage of what you say.
So it converts a speech to text.
It's probably notreal sophisticated AI,
but that's more than ChatGPTor anything like that.
That's what I use.
(31:15):
How about you, Alex?
ALEX MCGEE (31:16):
Yeah on my iPhone, I actually turned
off the AI featuresfor my email because I
didn't like the summaries.
I thought they werewrong a lot of the time.
They were pulling indetails from the fine print
at the bottom and having thatat the summary at the top.
So I turned that off.
I do occasionally look at theGoogle summaries from their AI,
(31:41):
but I was mentioningearlier I've
found where those are literallysaying wrong things that I
know are not true.
So I take it all with agrain of salt right now.
How about you, Charlie?
CHARLIE BENNETT (31:52):
All my I use is accidental, involuntary,
and under duress.
[LAUGHTER]
[END THEME]
FRED RASCOE (32:04):
Lost in the Stacks is a collaboration
between WREK Atlanta and theGeorgia Tech Library written
and produced by Alex McGee,Charlie Bennett, Fred
Rascoe and Marlee Givens.
ALEX MCGEE (32:14):
Legal counsel and a Siouxsie and the Banshees rock
doc where I'm really hopingto study my pronunciation
were provided by the BurrusIntellectual Property Law Group
in Atlanta, Georgia.
MARLEE GIVENS (32:26):
Special thanks to Presley, Mina, and Mary
Beth for being on the show,to everyone out there figuring
out this AI stuffand thanks as always
to each and every oneof you for listening.
CHARLIE BENNETT (32:36):
Our web page is Library.GATech.e
du/LostInTheStacks, where you'llfind our most recent episode,
a link to our podcastfeed, and a web form
if you want to getin touch with us.
ALEX MCGEE (32:47):
Next week's show is our first in a new series
that I am veryexcited about called
There's an Archivist for That,where we meet archivists working
in places you might notexpect to find them.
CHARLIE BENNETT (32:59):
I like that.
FRED RASCOE (33:00):
It's time for our last song today.
Tech companies havea way of promising
that their new technologieswill liberate us
in our professional lives.
CHARLIE BENNETT (33:08):
This machine fills training gaps.
FRED RASCOE (33:10):
And I'm old enough that I heard that promise
about personal homecomputers and then I heard it
about the internet and thenI heard it about smartphones
and now I'm hearing it aboutartificial intelligence.
CHARLIE BENNETT (33:22):
Have you learned your lesson yet?
FRED RASCOE (33:24):
While AI will almost certainly
become another technologicaltool that we will use daily,
perhaps under duress,when we get past all
the flashy marketingand look at AI tools
in the cold light of day,librarianship will still
need humans in the loop.
So let's close with"Cold Light of Day"
by Lewsberg right here on Lostin the Stacks and have a great
(33:45):
weekend, everybody.
[MUSIC PLAYING]
[MUSIC PLAYING]
FRED RASCOE (00:02):
How close do you think
we are to just machinesfilling metadata fields?
MARY BETH WEBER (00:07):
Never because there are certain things
that it takes--
it takes skill and knowledgeand thinking about it.
I just think thereare some things that
are just too nuancedfor a machine to handle.
FRED RASCOE (00:20):
I like that.
That was an instant never.
You didn't even haveto think about that.
[THEME MUSIC]
CHARLIE BENNETT (00:49):
You are listening to WREK Atlanta,
and this is Lost in theStacks, the research library
rock and roll radio show.
How about never?
Does never work for you?
FRED RASCOE (00:58):
Never always works.
CHARLIE BENNETT (01:00):
I'm Charlie Bennett in the studio
with the whole gang--
Alex McGee, Molly Gibbons,Fred Rascoe, oh, and me.
Each week on Lost in theStacks, we pick a theme
and then use it to create amix of music and library talk.
Whichever you're herefor, we hope you dig it.
ALEX MCGEE (01:15):
And our show today is called "This Machine Fills
Metadata Fields."
CHARLIE BENNETT (01:19):
I love a Woody Guthrie reference.
ALEX MCGEE (01:20):
Yeah, it is a good one,
but the Lost in the Stackscrew can't really take credit
for the title this time.
Wait a second, I thought oneof you came up with the title.
Whose idea was it?
The answer to that question is abit of an intellectual property
conundrum.
FRED RASCOE (01:35):
I guess we can say it was suggested by ChatGPT.
CHARLIE BENNETT (01:40):
Oh, no.
FRED RASCOE (01:41):
Yeah.
CHARLIE BENNETT: You let a machine (01:41):
undefined
into the production process.
How did it suggest that title?
Was it through theftor hallucination?
ALEX MCGEE (01:49):
Either way, listeners,
we promise we won't let therobots take control of the show.
FRED RASCOE (01:53):
Even if they do occasionally
help us with some things.
And
MARLEE GIVENS (01:57):
That's something we'll
hear more about from today'sguest, a trio of cataloging
and metadataexperts who recently
led an online discussionabout AI and metadata.
FRED RASCOE (02:06):
And our songs today are about hallucinations,
mysteries behind thecurtain, and the limitations
of technological tools.
Librarians havequestions about using
AI like how much labor arewe willing to hand over
to a technology?
CHARLIE BENNETT (02:22):
Uh oh.
FRED RASCOE (02:23):
In my weaker moments,
I'll admit I wish a machinewould just handle everything.
CHARLIE BENNETT (02:28):
Don't say it out loud, Fred.
FRED RASCOE (02:29):
So let's start with a song
about feelingoverwhelmed by the tasks
of daily professional work.
This is "AI, Do My Work for Me"by the Photocopies right here
on Lost in the Stacks.
[MUSIC PLAYING]
ALEX MCGEE (02:49):
That was "AI, Do My Work for Me" by the Photocopies.
This is Lost in the Stacks ontoday's show is called "This
Machine Fills Metadata Fields."
CHARLIE BENNETT (02:59):
We interviewed three librarians
about the AI issues thatconcern the metadata community.
MARLEE GIVENS (03:05):
Those were Georgia Tech's own metadata
librarian Presley Dyer, GeorgiaSouthern University's head
of library systems andtechnologies Wilhelmina Randtke,
who goes by Mina, and the headof central technical services
for Rutgers UniversityLibraries Mary Beth Weber.
So we were really interestedbecause the three of you
(03:26):
recently led what'sknown as an e-forum.
I'll let you explain thee-forum concept in a second.
But you led an e-forumon AI and metadata.
If you described this toone of your colleagues
what you were aboutto do, how did you
describe what an e-forum is?
MINA RANDTKE (03:47):
It's a two-day directed email conversation
where there are topicsfor two-hour or so windows
with a conversation about it,and so someone can be at work
and catch up on itbetween meetings
or participate in away of meeting people
who work in the same fieldwithout having to travel.
FRED RASCOE (04:09):
Is there any reason that it
is an email exchange of ideas?
I've seen similar kindsof exchanges of ideas
on social media, Iguess Twitter or X
or whatever it's called now.
Is there any reasonwhy it's particularly
targeted towards email?
MARY BETH WEBER (04:30):
Because when it was initially started,
that was the mainway to communicate
with a large group of people andit's just continued that way.
MINA RANDTKE (04:39):
Yeah, there have been similar things.
So ALA had had spectrumTwitter chats, which Twitter--
no, I guess now it's X--
those died off.
And so there have been socialmedia things done by ALA
that are similar.
MARLEE GIVENS (04:55):
And was this anyone's first experience
with the e-forum?
PRESLEY DYER (05:00):
It was my first experience.
ALEX MCGEE (05:02):
So even like as a participant?
PRESLEY DYER (05:04):
Mmm hmm.
So, yes.
Yeah, it was all new to me.
So I had to get used to it fromthe participant side of things
and the moderatorside of things.
ALEX MCGEE (05:17):
How did the three of you find each other?
MINA RANDTKE (05:20):
ALA has e-forum coordinators
and they put out a call andI responded to the call.
But someone proposedthe topic, and then they
put out a call whoelse wants to organize.
And I don't actually know whoproposed the topic of first one
of us three.
PRESLEY DYER (05:34):
Yeah, I--
well, it startedactually with Marlee
which she sent it out to GeorgiaTech Library and was just like,
hey, if anybodywould be interested,
I've done it e-forums before andthey are fun to participate in.
So I was just like, oh,well, maybe I should do this.
(05:58):
So I emailed aboutthe e-forum and was
saying that I'm involvedin cataloging and metadata
and how do I begin.
And so I had to send alist of potential topics.
And, of course, AI was the topicthat piqued the most interest,
(06:27):
and so from there, I didan outline of that topic.
And then they sent it outfor volunteers to help me.
MARLEE GIVENS (06:38):
And so, Mary Beth, you
responded to that call.
MINA RANDTKE (06:42):
Mmm hmm.
Yeah, I respondedto the call, too.
FRED RASCOE (06:47):
Was there a reason that this particular e-forum
resonated with you all?
Presley, you wanted to startan e-forum, Mary Beth and Mina,
you wanted to participate init, and the organizing concept
around it was AI in metadata.
Was it because that particulartopic resonated somehow or is it
(07:10):
important in your work somehow?
MARY BETH WEBER: I would say both. (07:13):
undefined
FRED RASCOE (07:15):
And how so?
MARY BETH WEBER (07:17):
We have done some experimenting with AI here
through our vendoras have other people,
and we're always looking forways to streamline our work.
And everybody'stalking about AI.
So we figured it was agood time to step into it.
FRED RASCOE (07:36):
And, Mina, do you have any association with AI
at your place of workat Georgia Southern?
MINA RANDTKE (07:41):
Well, my background is in metadata,
so that was more ofthe interest for me.
And with all of theartificial intelligence
hype that has come up in thepast three to five years,
one of the things I've seenis people-- or businesses
re-branding existingproducts with AI on them
including structured metadata.
So I like to look at howis this tool working,
(08:04):
how are DOIs being used, how isauthority control being used.
So, for example,Research Rabbit or Tool
Like That that's brandedas AI, the interface
is built on structured metadataalthough a lot of times the way
that metadata was assignedreally is very technical.
(08:25):
So little elves do it, but alsocomputers help little elves.
And they use the-- you know.
FRED RASCOE (08:31):
So AI is not magic.
There's metadatabehind the curtain.
MINA RANDTKE (08:36):
Yeah.
FRED RASCOE (08:38):
And, Presley, what was your I guess interest?
Was it just the fact that AIwas a new and interesting topic
in the world ofcataloging and how
AI was going to be integratedin cataloging practice?
PRESLEY DYER (08:54):
Yes.
Our systems anddifferent platforms
are starting to embed itin the library catalog.
And so it started like,oh, well, let me use--
let me try-- test it out bycreating an original record
(09:14):
and then, oh, well, let me seehow this original record looks
against another ChatGPT.
And so it was, oh, well,it just piqued my interest,
and just because it istaking over professions
(09:38):
and daily lives and stuff.
So it's just I would liketo learn more about it.
MARLEE GIVENS (09:45):
We'll be back with more
from Presley Dyer, MinaRandtke, and Mary Beth
Weber after our music set.
CHARLIE BENNETT (09:51):
File this set under TT390.L853.
[MUSIC PLAYING]
Now, Fred.
FRED RASCOE (09:59):
Yeah, sorry.
MARLEE GIVENS (10:08):
That was "Behind the Curtains" by Jan Bradley.
Before that we heard another"The Letter" by Wire,
and we started our setwith "Beyond the Fringe"
by the Lime Spiders, songs aboutrevealing the mysteries behind
the curtain.
[MUSIC PLAYING]
CHARLIE BENNETT (10:27):
This is Lost in the Stacks,
and we are speaking with PresleyDyer, Mina Randtke, and Mary
Beth Weber aboutAI and metadata.
MARLEE GIVENS (10:35):
Presley, Mina, and Mary Beth
recently led a two-daye-forum on that topic,
and we wanted to find out whatAI tools they were already
using before the e-forum.
PRESLEY DYER (10:47):
We use Exlibris products,
so Exlibris has embeddedAI assistant tool
within the metadata editor.
And so it's startedwith that, and then we
have started using AI pops upin everyday tools like Google.
(11:13):
And Amazon has its own thing.
So it's just like I wasjust like, oh, well, I
know other organizationsare trying to incorporate it
in the library board.
So OCLC is about--
is trying to do a startup tohave their own AI assistant
(11:37):
tool.
So, yeah, I hadknowledge, but I wanted
to learn more about what otherswere doing in their libraries.
MARY BETH WEBER (11:49):
We also are an Alma library,
and we volunteered toparticipate in early testing
of the AI metadata assistant.
But some of my colleaguesin the department
also have usedChatGPT for things
like to generate Python scriptsor to help with projects.
And now we have alicense for Copilot,
(12:10):
and we use it for things.
I haven't used it forany cataloging things,
but, for example, I wrotesomething and the editor--
the publisher asked mefor a list of acronyms.
So I could have gonethrough it, and I was lazy.
I asked Copilot get allthe acronyms for me,
alphabetize them.
FRED RASCOE (12:26):
Oh, there are too many acronyms in librarianship
but especially in cataloging.
MARY BETH WEBER (12:31):
Yeah.
FRED RASCOE (12:31):
Cataloging has way too many acronyms.
So I totally sympathizewith you farming that out
to an artificial intelligence.
MARLEE GIVENS (12:40):
Is Georgia Southern also Ex Libris?
MINA RANDTKE (12:44):
We're Ex Libras, but that's--
I actually didn't playwith the tool in Alma.
Presley and I are both involvedin the same statewide migration,
so I've done a lot ofdata cleaning in Alma.
But in that case, what I'vetried to do is look for outliers
and reduce the complexityof data before moving it,
and it's been much morestatistical analysis.
(13:08):
Most of my work withartificial intelligence
has actually been opticalcharacter recognition,
so we don't think aboutit as cutting edge.
But it really islooking at a picture
and detecting it as text.
And so if you think aboutscanning documents and then
having a person writewhat they're about
(13:30):
versus being able to just havethe text taken out of them
and put in a searchable format,the time and cost savings
is so great.
So I worked indigital libraries,
and we've done work withoptical character recognition.
And then here internally,my role at the library
is somewhat like achief information
(13:54):
officer for the library wherethe kind of things I'm doing
is like setting up anintranet and making sure
things are searchable andshow up to the same people.
So just looking atlittle details like that
and maybe havingsmall technical tools
that are establishedbut are fundamentally
driven by machine learning.
MARLEE GIVENS (14:13):
What are your biggest concerns about using AI
for resource descriptionor for metadata management?
MARY BETH WEBER: We're short staffed, (14:23):
undefined
and as people retire orleave, they're not replaced.
So we have a very, verysmall group of catalogers
for a library system our size.
And my fear is something alongthe lines that, oh, well,
we can use that--we wouldn't say it,
but somebody elsein the organization
might say, well, we can useAI to catalog the books.
(14:45):
We don't really need people.
I know that sounds farfetched, but people
have believed hokier things.
FRED RASCOE (14:52):
I've heard--
definitely heardsimilar conversations.
Alex, I wonder as thearchivist in the room have you
heard similarconversations about AI
being able to describearchival objects?
ALEX MCGEE (15:05):
Oh, yeah.
When I was at MIT actually,there was a project,
and it was 2019, 2020 where wewere picking a small collection.
And basically theidea of the project
was we would demonstratethat the AI had biases
and created baddescription was our goal.
(15:26):
Actually, we didn'tget to see it
all the way throughbecause of the pandemic,
but that was the theorywe were testing out
was that it would notcreate ideal description.
So we were actuallydoing women's collections
and then showingthe gaps that AI has
and how to respectfully describemore complicated material.
MARLEE GIVENS (15:46):
Did any of you-- have any of you
in your own workfound anything that AI
seems to be goodat when it comes
to cataloging and metadata?
MINA RANDTKE (15:55):
Well, it does this-- it does optical character
recognition for me.
So I've never typed atranscript of a text document
in my adult life.
PRESLEY DYER (16:08):
I just recently used
it to help with adigitalization project and just
taking pictures ofhistorical documents.
And then it being able tocreate a spreadsheet of what
we were trying to archive.
(16:31):
I feel like it did a greatjob of being able to just do
the spreadsheet side of things.
So, again, it freed us up tobe able to vacuum and actually
clean the material.
I like the fact that it canhelp with the Excel spreadsheet
stuff.
MARLEE GIVENS (16:53):
Are there tasks that it seems to struggle with?
MINA RANDTKE (16:57):
It doesn't have sensors,
so it's just a computer program.
So if anything exists in thereal world, it can't tell it.
And in the library,it's we have three kinds
of materials in our collection.
So we have thingswe've purchased
from vendors, which usually therecords come from the vendor.
Then we have equipment itemsthat get checked out really
(17:20):
frequently like laptops,and it can't look at it
and say, this serialnumber and barcode
don't match because ithas no eyes to look.
And then we have ourarchival materials
and our specialcollections materials,
and it can't look at those.
So it doesn't have sensorsto detect the real world.
MARY BETH WEBER: Some people have (17:39):
undefined
found that it would generatefake contents notes.
FRED RASCOE (17:43):
Oh, that's interesting.
Just like it-- whenpeople do research
and it hallucinates journalarticles that aren't really
there.
What kind of cataloginghallucinations examples
have you seen?
MARY BETH WEBER: Both cases were-- (17:56):
undefined
where people werecataloging things
in languages other than English,and one was a non-Roman script
as well.
And they said thatit generated contents
notes for the booksthey were cataloging,
and they were not even remotelyclose to what was in the book.
And I don't know where they haddrawn that information from,
(18:18):
where as in the AInot the catalogers.
FRED RASCOE (18:22):
Mina, have you ever used--
in your use of optical characterrecognition tools, have you
ever had the tool that you'reusing hallucinate text that
wasn't actually there?
MINA RANDTKE (18:34):
No, it's--
it'll mis-detectletters, but I don't
think I've been to thepoint of it mis-detecting.
And a lot of it is maybeto do an analysis of things
and then make a decision ofhere's a clean up project to do
or here's a direction to go in.
I'm really looking atwhatever tool I'm using.
MARLEE GIVENS (18:56):
You're listening to Lost in the Stacks,
and we'll be back withmore about AI and metadata
on the left side of the hour.
[MUSIC PLAYING]
MARK RIEDL (19:15):
Hi there.
This is Mark Riedl of GeorgiaTech's computer science
department--
or is it?
Maybe my voice hasbeen deep faked,
and this is justa digital forgery
created by a neural network.
Either way, you're definitelylistening to Lost in the Stacks
on WREK Atlanta.
CHARLIE BENNETT (19:38):
Today's show is called "This Machine Fills
Metadata Fields," and if youdon't know why I called that
a Woody Guthrie reference, Isuggest the Google image search
Woody Guthrie guitar.
MARLEE GIVENS (19:49):
Even if this proverbial metadata machine
never comes to be, it istrue that a machine, ChatGPT,
is responsible for theshow title this week.
ChatGPT suggested that as thetitle of a Lost in the Stacks
episode that never happened,an episode that ChatGPT totally
made up.
CHARLIE BENNETT (20:06):
How dare it?
MARLEE GIVENS (20:07):
It's one of many examples of a generative AI
chatbot presenting itsown hallucination as fact.
After we recordedthis interview,
I attended theLibrary 2.0 Conference
on AI and Libraries, Literacy,Ethics, and Responsible Use,
and during the opening keynote,presenter Crystal Trice
addressed the hallucinationsand inaccuracies
(20:28):
that often come up in herconversations with library
staff.
She says--
CHARLIE BENNETT: Everyone assumes (20:33):
undefined
that those are going toget better over time,
and maybe they will.
But really there needs tobe this basic understanding
that the tools are notdesigned to be factual.
These algorithms aren'tdesigned to be factual.
They are designedto be generative,
and that's a real mindshift for library staff.
MARLEE GIVENS (20:52):
She continues--
CHARLIE BENNETT: It's not a bad thing (20:54):
undefined
when we're using AI indifferent aspects of our work,
but it is important to keepin mind that being factual
is not even in its radar.
MARLEE GIVENS (21:04):
While this is mildly unsettling to me
as a reference andinstruction librarian,
I find the metadata communitiesapproach practical, thoughtful,
and forward thinking.
Even if the machine thatfills metadata fields
remains a fantasy, ourcolleagues in technical services
are embracing the futureone tool at a time.
CHARLIE BENNETT (21:24):
File this set under QA76.99.N38.
[MUSIC PLAYING]
ALEX MCGEE (21:42):
That was hallution--
"Hallucinations" by BakerKnight and the Knightmares, k-m,
and then "Mirage" bySiouxsie and the Banshees.
CHARLIE BENNETT (21:54):
Oh, I just realized the difference
in our ages.
ALEX MCGEE (21:56):
How young I am, yeah.
Songs about things thataren't really there.
[MUSIC PLAYING]
MARLEE GIVENS (22:04):
Welcome back to Lost in the Stacks.
We are speaking withPresley Dyer, Mina Randtke,
and Mary Beth Weber, whorecently facilitated an e-forum
about AI and metadata.
In the final segmentof our interview,
we asked them what they learnedfrom that online discussion.
PRESLEY DYER (22:19):
Everybody is voicing the same concerns
like environmentalclimate change, that
was an issue that came up.
The fact that it couldtake over staff positions
And so everyone, it seemedlike had the same concerns,
(22:43):
and it was just really niceto just be in that environment
to actually hear, hey,I'm not the only one
thinking these same things.
And it was justreally nice, too,
the fact that it was alldifferent types of libraries,
too.
There were specializedacademic, public.
(23:05):
We did get a goodvariety in that regard.
MARLEE GIVENS (23:09):
Mary Beth, what did you take away from it?
MARY BETH WEBER (23:11):
I would have to say
it's very similarto what Presley
said that a lot of people arejust at the beginning point.
They don't really understand.
A lot of them hadfears about their jobs.
There was some commentto there's always
feedback after theEA forms, and someone
said it was like the blindleading the blind because nobody
had a firm grip ora defining answer.
MARLEE GIVENS (23:35):
Did you think that maybe some people came
thinking that you were experts?
MARY BETH WEBER (23:41):
Yes, that always happens.
Yeah.
MARLEE GIVENS (23:45):
Mina, did you pick up anything?
MINA RANDTKE (23:48):
For me, what was most striking
was how much concernparticipants expressed
with environmental concernsand environmental impacts
because it wasn'tsomething we had
asked about in the questions.
The e-forum will ask pre-writtenquestions, and none of them
were about that.
But it kept coming outemergent as something
that participants talked about.
FRED RASCOE (24:10):
It sounds like from all three of you,
the general sense from the--
I think the E - forum was twodays of email conversation.
From the two days,the general sense
is that people are interested,lots of reservations
about the utility of it,about replacing labor, about
environmental concerns.
(24:32):
In those two days,were there any outliers
among the cataloginglibrarian community,
people who stood on theirhill and said AI no way.
This is not something thatwill ever be useful to us,
or on the other hand, AIis the future committing
(24:53):
totally no reservations?
Any interestingoutliers like that?
MINA RANDTKE (24:58):
I actually don't remember
any that were acomplete outlier people
had examples of here'ssomething where a vendor used
it to generatemetadata, and it's
had these problems becausethey sometimes a word has
multiple meanings or something.
Or people would expressconcern with their job,
(25:19):
but there weren't really anyoutliers who were completely
against or completely with.
I think they were just lookingfor what tools can we use.
MARY BETH WEBER (25:26):
All those down to Earth level
headed catalogersin the discussion?
Yeah.
MINA RANDTKE (25:34):
Someone opposed might not
have participated, right?
FRED RASCOE (25:38):
That's also very true.
Yeah, sample bias I suppose.
MARLEE GIVENS (25:43):
How do you each feel about AI
when it comes to your own job?
Are you-- do you feellike this is inevitable?
I need to learn it.
Do you think it'sgoing to be helpful?
Are you kind of dubious?
So, Presley, shaking herhead, I'll start with Presley.
PRESLEY DYER (26:01):
I use it every day for something,
creating coding for Alma,the normalization rules.
I use it-- for it to quicklygenerate and spit out the rule
that we'll use tochange the call number
(26:23):
or something like that.
And I am about to help theprofession test it out for OCLC
so it can be used broadly.
So I feel like--
it's just become an everydaypart of my life at this point.
MARLEE GIVENS (26:47):
How about Mary Beth?
MARY BETH WEBER (26:49):
I've worked long enough that I've
gone from a library that hadthe old beehive OCLC terminals
and cards to onlinecatalogs, so I just
see it as a naturalpart of the evolution.
I don't think it's goingto solve everything,
but I think it's atool that can help us.
MINA RANDTKE (27:05):
I'm more like Mary Beth.
It's just another tool.
So each thing isjust another tool.
Maybe even in a researchdatabase, suddenly I'll see.
Now I have a facet forfunding organization,
and it's some kind of entityrecognition that they put in.
Or I'll see maybe animproved search somewhere.
But to me, it's another tool,and it builds on past tools.
FRED RASCOE (27:29):
So you maybe see in just a few years like library
schools will be teaching theirstudents cataloging and teaching
them how toresponsibly use the AI
tools that everyone in thecataloging profession is using?
Or is that too far?
MINA RANDTKE (27:48):
I don't know if they teach cataloging.
That's one of theweaknesses of--
FRED RASCOE (27:52):
That's--
MINA RANDTKE (27:53):
Yeah.
Because libraryschools don't tend
to teach electronicresources management,
don't tend to teach how toteach one-shot instructions
and don't tend toteach cataloging.
FRED RASCOE (28:04):
I may have been out of library school too long,
2004 graduate.
So my knowledge of the currentcurriculum may be out of date.
MARY BETH WEBER (28:13):
And those that teach it,
too, it's not always required.
They teach it here.
They do better when they haveadjuncts teaching the class
because otherwisethey have people
who are only teaching theory.
And theory's great,but there's real life.
So I'm hoping thatthey will teach them
how to use it responsibly.
They have subscriptionsto other things
(28:34):
like classification weband the RDA toolkit,
and the students learnhow to use those.
So maybe that will just beanother thing they show them.
MINA RANDTKE (28:44):
And to me, fundamentally,
information literacy isalready a good framework
for understanding a lot of thetools for content creation.
So understanding sourcesand it's a different world
of librarianship for metadata.
But information literacy.
I think gives thetools to understand
some of the newtechnologies we're
seeing with generated content.
(29:06):
MARLEE GIVENS: Well, this is great. (29:09):
undefined
Thanks for filling usin on AI and metadata
and your e-forum experience.
MARY BETH WEBER (29:19):
Thank you for asking us.
MINA RANDTKE (29:20):
Yeah, thank you for having us.
PRESLEY DYER (29:22):
Thanks, everyone.
FRED RASCOE (29:24):
You are listening to Lost in the Stacks,
and today's episode is called"This Machine Fills Metadata
Fields."
MARLEE GIVENS (29:30):
And our guests were Presley Dyer,
metadata librarian atthe Georgia Tech Library,
Wilhelmina Randtke,head of libraries,
technologies, and systems atGeorgia Southern University,
and Mary Beth Weber,head of Central Technical
Services at Rutgers University.
File this set underRC553.H3 no dot S23.
(29:52):
[MUSIC PLAYING]
FRED RASCOE (30:00):
That was "Blind or Dead" by Lost Cherries.
And we started with "SimpleMachines" by Dancing Cigarettes,
songs about the tools that canhelp us as long as we realize
their limitations.
[MUSIC PLAYING]
MARLEE GIVENS (30:21):
Today's show was called "This Machine Fills
Metadata Fields--
All About the Benefits andLimitations of AI Tools
in Libraries."
We heard about opticalcharacter recognition,
automated spreadsheet creation,and using ChatGPT to find
all the acronyms in a document.
So lately the thing thatI've been relying on AI for
(30:42):
is editing, making something Iwrote fit a certain word count
or summarizing free textanswers from a survey.
So are there AI toolsthat you use at work
or in everyday life, Fred?
FRED RASCOE (30:55):
I guess this is probably
a really low bar forAI, but I do actually
use that in your iPhonewhen you talk into it
and it sends a textmessage of what you say.
So it converts a speech to text.
It's probably notreal sophisticated AI,
but that's more than ChatGPTor anything like that.
That's what I use.
(31:15):
How about you, Alex?
ALEX MCGEE (31:16):
Yeah on my iPhone, I actually turned
off the AI featuresfor my email because I
didn't like the summaries.
I thought they werewrong a lot of the time.
They were pulling indetails from the fine print
at the bottom and having thatat the summary at the top.
So I turned that off.
I do occasionally look at theGoogle summaries from their AI,
(31:41):
but I was mentioningearlier I've
found where those are literallysaying wrong things that I
know are not true.
So I take it all with agrain of salt right now.
How about you, Charlie?
CHARLIE BENNETT (31:52):
All my I use is accidental, involuntary,
and under duress.
[LAUGHTER]
[END THEME]
FRED RASCOE (32:04):
Lost in the Stacks is a collaboration
between WREK Atlanta and theGeorgia Tech Library written
and produced by Alex McGee,Charlie Bennett, Fred
Rascoe and Marlee Givens.
ALEX MCGEE (32:14):
Legal counsel and a Siouxsie and the Banshees rock
doc where I'm really hopingto study my pronunciation
were provided by the BurrusIntellectual Property Law Group
in Atlanta, Georgia.
MARLEE GIVENS (32:26):
Special thanks to Presley, Mina, and Mary
Beth for being on the show,to everyone out there figuring
out this AI stuffand thanks as always
to each and every oneof you for listening.
CHARLIE BENNETT (32:36):
Our web page is Library.GATech.e
du/LostInTheStacks, where you'llfind our most recent episode,
a link to our podcastfeed, and a web form
if you want to getin touch with us.
ALEX MCGEE (32:47):
Next week's show is our first in a new series
that I am veryexcited about called
There's an Archivist for That,where we meet archivists working
in places you might notexpect to find them.
CHARLIE BENNETT (32:59):
I like that.
FRED RASCOE (33:00):
It's time for our last song today.
Tech companies havea way of promising
that their new technologieswill liberate us
in our professional lives.
CHARLIE BENNETT (33:08):
This machine fills training gaps.
FRED RASCOE (33:10):
And I'm old enough that I heard that promise
about personal homecomputers and then I heard it
about the internet and thenI heard it about smartphones
and now I'm hearing it aboutartificial intelligence.
CHARLIE BENNETT (33:22):
Have you learned your lesson yet?
FRED RASCOE (33:24):
While AI will almost certainly
become another technologicaltool that we will use daily,
perhaps under duress,when we get past all
the flashy marketingand look at AI tools
in the cold light of day,librarianship will still
need humans in the loop.
So let's close with"Cold Light of Day"
by Lewsberg right here on Lostin the Stacks and have a great
(33:45):
weekend, everybody.
[MUSIC PLAYING]
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.