So I’ve been thinking about the great opportunity AI presents librarians and wanted to share those and get your feedback:
I’m including the automatic transcript for accessibility and so search engines may find it 🙂
Click for Automated Transcript
WEBVTT – This file was automatically generated by VIMEO
Hello. It seems when I’m invited to speak anywhere,
it’s either on one of two topics,
either it’s the challenges that librarians are facing,
particularly around book challenges, materials shifting ideals,
censorship and intellectual freedom or artificial intelligence.
’cause there’s clearly fit together. Sure they do.
But we’ll get there in a moment. Several weeks ago I presented, I,
I posted a presentation I did around the idea of the challenges that we face.
And it was a little dark.
It had pictures of puppies and kittens to make it feel better,
but really was demonstrating how legislatures and formal policies being
drafted against librarians as individual,
not just libraries and their programs and their materials.
I consider this a bookend the other side because it’s actually a very optimistic
view of what I see for libraries and their role within artificial
intelligence, a particular generative artificial intelligence.
This presentation blog post comes from two things. One, a,
uh, presentation and conversation I had at Filoni, which is a,
the largest Latin American book festival. Uh, we teamed up, uh,
with the University National in Mexico and University of Texas.
And I was down there and had some brilliant conversations with some brilliant
people about ai and it’s,
its lead and its take and role in industry and shaped this idea.
The other was then taking that and putting into a presentation for a Danish
public library around AI and its role. So with those,
I just wanted to share these ideas. Um, and so let’s get into it.
I wanna talk about new librarianship and artificial intelligence.
A better way might be talking about the idea of the opportunities of
librarianship in artificial intelligence. And I wanna begin with a metaphor.
I wanna begin in an interesting way because I wanna talk about how AI has been
shifting our knowledge infrastructure. We’ll get into that in a moment.
And so let’s start with that metaphor to begin with.
This is a chimera, um, or an AI generated version of a chime.
All the images in here are AI generated.
A chimera in ancient mythology was a com was a beast,
a mythical beast that is combined of several animals, most of them, uh,
representation of a head of a lion, the head of a goat,
and the tail of a serpent. I like this one. It’s got wings.
You sometimes they have dragons to them,
but the point is that it’s just a bunch of organisms snap together in a
terrifying way that doesn’t necessarily make sense.
I became familiar with the idea and the phrase of a chimera because it turns out
I am one.
A chimera in biology are two different genetic systems from two different
hosts that are surviving and working side by side in one host.
So in my case,
my muscles and face and hair and brain and eyes,
Um, for better or worse come from my parents.
They have one genetic system,
one set of d n a instructions that are generating their production. However,
what you don’t see is my blood. My blood is actually comes from my son,
different d n a system, different genetic system.
And the reason is because I received a bone marrow transplant to fight cancer.
And this new blood system besides feeding my cells and
providing oxygen and all these other things turns out also kills cancer
cells that were produced by my blood system. So I’m a chime,
but here’s the thing. While that was a dramatic process, trust me,
it was a dramatic process. And twice you wouldn’t know it by looking at me,
you wouldn’t see it by the medication I take,
it turns out that bone marrow is the one organ transplant you receive.
That you don’t have to take anti-rejection drugs for the rest of your life
because in essence it is your immune system. It’s,
you wouldn’t tell by a doctor doing a physical exam by how I speak,
by my different mannerisms,
and yet I’m here and I’m able to survive because of this
alien d n a in my body. In essence,
they replaced part of me, but outwardly looks the same,
but inwardly is vastly different.
And what I want to use this as a metaphor is to talk about our knowledge
infrastructure. So, um,
AI came here as and knowledge.
So I’m gonna make a few really uninspiring grand
The first is that artificial intelligence has already profoundly changed our
That is not that grand unless you know what the heck I mean by knowledge
infrastructure. So give me a moment.
The knowledge infrastructure are the people, the technology,
the sources and the policies that guide how we know the world,
not just in formal education or in business or in training,
but in everyday life and how you interact with your wife,
how you interact with your children, how you learn about hobbies,
how you decide which direction you’re going to go on a walk when you decide
what you’re going to read. All of this is part of a knowledge infrastructure,
the world of information that you interchange with on a regular basis.
And it consists of people, once again, the people you’re talking to,
maybe it’s a good friend, maybe it’s a colleague, spouse, what have you.
Those are part of your knowledge infrastructure.
Those are the conversations by which you learn.
It’s also the technology of interchange, for example. Does that occur this way?
Transmission? Is it a video conference? Is it in person? Is it in an office?
’cause an office space is a type of technology that concentrates different
workforces together for, um, not just ease and infrastructure,
but also for interchange and inter working together.
It’s the sources of information consulting, Wikipedia consulting, encyclopedia,
the book that you’re reading, um, your own diary,
looking at this different sort of recorded
accessible information sources. And finally,
the policies that dictate how that happen. Uh, those policies can be law,
which you’re allowed to see, which are not. So I’m in Texas,
so I can’t do TikTok if I’m on a any sort of university connected
machine, well, that’s a policy, it’s a rule.
There’s no technological reason I can’t, but the rules say I can’t.
It’s also copyright.
So a lot of the materials that I may wanna access I have to pay for or I can
access them, but I can’t copy them.
All of that fits into the rules and confines by which I learn and how I
interact and all of these things shape how I see and know the
world because one thing that we know is that AI has in essence
radically changed this infrastructure.
We didn’t notice it because there were bits and pieces,
but if you search the web, you’re using ai, right?
The algorithms, the way that it understands me,
the ad placements that it’s doing, all of that shapes what I see in what I see,
which also means it’s shaping how I learn. And YouTube video what comes next.
AI is shaping this in the form of algorithms, inductive algorithms,
machine learning, um,
different forms of using Bayesian technology to anticipate and
predict what I’m learning and what ai.
I’m just generally talking about how we use these different algorithms to
generate human-like capacity in given domains.
The gold ring that we’re trying to get is to do it in any domain
called generalized ai. We’re not there yet,
but we see that in certain applications. In music. I’m a big Spotify user.
It’s the second biggest music service after YouTube.
Both of them use AI to predict what’s my next song? What’s he going to like?
Is it genre?
And Spotify themselves not only looks at the musical notes and similarities and
genre information, it uses natural language processing on the album cover,
on social media chats,
on websites to try and cr predict what’s Dave’s next favorite song.
We see it in research, um,
in generating information in how we understand information and how we find
those documents. In movies,
we’re seeing a lot of c g I that are produced without any human interference at
all. We’re seeing it in publications. Increasingly,
this flood of AI generated and AI supported, um,
articles going into everything from our academic journals to literary journals
to book publishers to,
we now see a raft of AI generated books available through a, um, Amazon.
And we see it in the military.
And we’re seeing in governance when we now see AI generated political ads,
talking about how we can, um,
represent voice and all these and video and deep fakes and AI is pushing this
technology further and farther.
So we’ve ended up with a very different knowledge infrastructure than we had
even years ago. But oftentimes we don’t notice.
And part of the reason we don’t notice is because when we notice ai,
when it’s brought to our attention, at least previous to the past months,
it’s represented something new. It’s a new feature,
it’s a new technology right now I can track things differently.
I can have different, um, exercise patterns, et cetera.
But at the same time we’ve been looking at the new stuff.
All the old stuff has also been adopting,
adapting to and being generated through AI and AI tools. The next post you read,
the next post you make all of those things now used to be run by different
types of algorithms now very much by ai heuristic learning algorithms.
And so our information infrastructure is a chimera. If you look at it,
it doesn’t look that different, but it is vastly different.
Now that transparency,
that shift became very evident once again a few months ago with the introduc
introduction of chat. G P T. What we’re seeing is that generative AI chat,
g p t Dolly, um, firely,
these idea of creating texts and images and videos and all this
other information, it really has been like a face that’s emerged.
And it was startling because when we saw this technology,
when we could ask it a question and get a human-like answer,
we hadn’t seen that before because once again, AI was always there,
but it was behind the scenes now it was right in our face and it was doing
things that we pretended that we never thought AI could do.
So we’re seeing this. And so here’s my once again bold prediction,
which isn’t that bold, which is generative AI will have a larger
impact on publishing in particular than the web did.
The web changed how we’ve disseminated information?
It allowed anyone to become a publisher. It allowed us to think in multimedia.
It allowed us to think in hypertext the concept that ideas could be linked
over vast distances and genres and topics was a whole new concept.
And yet now we are fluid in the notion of narratives moving from place to place
idea to idea rabbit hole to rabbit hole. Well,
I think we’re gonna see the same thing in generative ai.
And it’s not just because we’re gonna get a lot of crap written very quickly by
a machine. We’re going to see it in how we produce images,
but we’re really gonna begin to see it is in the intersection between AI
publication and libraries.
That’s the area that I’m concerned about with this one.
We know that AI generated content is coming into collections.
Joe Janes pointed this out beautifully in a,
in a talk we did a couple years ago.
We know that the books and journal articles and images and all the kinds of
things that we’re acquiring our databases, et cetera,
are coming in now with AI generation portions or in whole.
And we have to ask ourselves, how do we recognize this? How do we catalog this?
How do we distribute this information?
And one of the things that’s sort of an interesting bit of recursion,
sort of meta thought is that a lot of how we’re gonna deal with this is going to
be using the same tools that generated it in order to identify it using AI
tools to recognize this.
Any article that gets handed in by a student these days goes through
turnitin.com, which is increasingly using ai.
It’s moving beyond simple algorithms to identify plagiarism.
AI is allowing it to go beyond simple looking for word matches and now to begin
to move stylistically different patterns, regularity of text, et cetera.
And it’s beginning to use it as a way of identifying texts that come in from
generated sources and cataloging not just the idea of basic human cataloging,
but the idea of enhancing that cataloging other metadata generation using things
other than human generated metadata in order to find and organized materials.
See previous conversation around music, right?
So these are challenges that we as a library are gonna face,
but we’re also seeing that publishers are facing it.
And in some ways publishers don’t want to face it.
When we see things like using AI to identify sexual content in children’s books
and libraries, that can be banned scary. But that’s what’s happening in,
I believe it’s Iowa.
We have to ask our question about who advocates and educates the community about
this. Do we need to warn people that we’re looking at AI generated material?
Do we need to point it out in some ways?
How do we begin to think about this? How do we train people to deal with this?
And once again,
train them either in a legalistic way of saying things like look out for it or
in a creative way on how to use it,
how to take advantage of it as we’ll talk about in a moment.
The other question that’s central to generative AI that’s only now coming up
and it’s going to be amazingly important to its impact is where
will the curated and quality data come for ai?
So the reason that chat g p t does a really amazing job at generating human-like
text is it’s trained on human-like text.
It went out and read every word it could find on the internet copyright or not.
We’ll get to that in a moment. Or it found every image it possibly could.
Looked for patterns, look for regularities,
and now can generate this human-like content.
So it needs masses of data to training. Chat.
T p t couldn’t have existed years ago because there wasn’t enough content
available on the internet. Mind you, there wasn’t enough computing power,
but certainly content that it could ingest.
Now this is probably changing because what we’re seeing is things like the
New York Times has now changed their, um,
terms of service so that it doesn’t allow their data to be harvested by AI
tools. And not only that, but there’s a prediction that very soon, um,
and already one publisher has sued open ai,
but others will for copyright infringement.
That is that they used their materials to train this chat
g p t without paying or enumerating the creators or the own
publishers of those services. Now why does that matter?
Because there’s certainly plenty of text on the internet that isn’t under
copyright. But a recent study public, um, reported on in, uh,
business Insider came up with the,
they did a study and they looked at the efficiency and effectiveness of ai.
When you begin to withdraw heavily curated human generated content,
i e books and articles and published materials,
the author settled one important question right away.
AI models rely heavily on high quality human generated content that is
often under copyright. And here’s the poll quote. Without that,
performance begins to suck hard.
I love that as we show model performance significantly degrades
if trained only on low risk text EEG,
out of copyright books or government documents due to its limited size and
domain coverage. They wrote. In essence,
if we withdraw all of the content that human beings have spent a
lot of time and editors have spent a lot of time and reporters have spent a lot
of time, if we remove those high,
let’s call them quality invested resources,
generative AI suffers greatly. Now,
what this is going to lead to is invariably a content question such
as when search first came out,
when Google and Yahoo and before that Alta Vista and Excite began
crawling the web to find content,
one of the first reactions from publishers was, don’t get our content.
We don’t want your, you’re, you’re ping our content.
You’re breaking copyright by doing this.
And a lot of court cases came up that said, no, this is a transformative work.
It is significantly different and it doesn’t infringe on the original use of the
It in fact highlights and allows you to find that work and so therefore will
And publishers have realized a long time ago that they needed to switch from the
withdrawing the materials to highlighting and in fact optimizing for finding
their materials and using this in a different way.
And I think we’re gonna see the same discussion,
but one of the areas I think as librarians,
I kept talking about opportunities and here we go,
that librarians need to think about is we could be generating some of that
And I don’t mean by making our collections available like the Google scanning
and book project. I mean,
as in working with our communities to generate more of this text outside of
think about what our members can do in our libraries these days.
If they have an idea, whether that’s an invention, whether that’s a story,
whether that’s a oral history, what have you,
they can now up their production,
up their production in both effectiveness and efficiency and speed.
You can go to a chat G p T and you can plug in your idea and ask it to write a
script and you can edit that script quickly as opposed to starting from scratch.
You can take that script and run it through another AI program that will
generate human-like narration without a human being harmed in the process.
In fact, if you would like to use your own voice,
you can actually program it to speak in your own voice with your own cadence.
You can then take that audio and that text and you can feed it into automatic
video generators where it will come back with set of storyboards.
You’ll agree on storyboards and it will produce the whole video for you.
And you could do that in remarkably short time.
What used to take production of weeks can now take production of days if not
hours. And now it’s not gonna be everyone,
but it can be more and more people if they have access guidance and
understanding what they’re doing. Besides the ethical part,
just on how to phrase questions. How to where are these tools available?
How much do they cost? Is there a subsidies it cost?
These are things that librarians can be helping their academics can be helping
their mothers can be helping their children,
can be helping their students do on a regular basis to become more effective
communicators using AI tools not to come up with the message,
but to help frame the message effectively.
And that grooming and editing process can then go into training AI as well.
And we can begin to overcome some of the limitations with the content
that represents our communities. And I wanna be really important about really,
uh, let’s talk about the importance of that.
If we can help our communities generate and tell their stories,
minoritized communities, isolated communities, normally ignored communities,
and we can get this within the training realm of AI and AI tools.
It’s going to begin to change the text that’s gonna begin to change narratives.
It’s gonna be begin to change viewpoints within it.
We can in essence flood the field with information about the common good about
our local community and our local indigenous knowledge systems and begin to
train those systems,
particularly as New York Times and the major publishers are worried
about pulling out of these things because they’re worried they’re going to lose
their market. So librarians can lead the way.
This is what I look at as a huge opportunity,
is the bookend to we’re going to worry about ban books.
We can be helping to prepare society,
prepare our communities,
and utilize technology in a positive way and ensure
ethical use, explainable use will get in that moment.
But this is a great opportunity for librarianship just as if
librarianship used to give access to other things.
We can now continue our move from organizers of things to
facilitators of knowledge,
moving from capturing and providing access to other people’s texts,
to now telling your stories and moving those stories ahead.
That’s what we’re talking about. So,
and librarians and librarianship is uniquely situated to do this.
We are at the right place at the right time to do this. Why? Well,
a couple of things. One context is King.
We understand that information, these products,
this data is all contextual, right? There isn’t necessarily,
I I always use quality up to this point with quotes or a little side eye because
we know it matters in what context it’s being used and you know,
famous examples. But if a reporter walks in asking for how is the,
how is the how are hate groups using the web to recruit people,
sending them to storm front and the k k K is a good quality resource for that
reporting. An eight year old looking for afterschool projects,
not so much that idea that we understand that what matters,
how it matters, how you understand it, how you interpret it,
what the community has brought around it,
what are the nature of quality indicators,
how it’s going to be used in your process.
That whole understanding the world of our community members
is our strength. We’re not creating generalized tools for the world,
we’re creating services for individual community members and their community.
So we understand credibility in a very difference.
We understand quality in a very different way.
We understand utility in a very different way.
All of these nuanced research thought through
tradition views are absolutely essential to adapting
and adopting into the AI world.
We understand longer that we need to stop being gatekeepers and start being
publishers and platforms that our goal isn’t to only collect the good stuff.
Our goal is to work with you to understand what is good for you,
for your purpose, for your need now. And by doing that,
we then unleash the ability not only for you to consume information,
but to deliver information, to send your views into the world,
to help people podcast, to create video, to create music, to share stories,
to create their own histories,
that valuable platform of consuming and producing because it’s
a false dichotomy, right? That idea of disseminating and adjusting,
it’s a false dichotomy.
It’s all learning and it’s all around our mission and we could be doing this
so this is one thing that makes librarians uniquely situated for power,
for making positive change for
shepherding the common good in this era. The other is we understand churn,
We know that this knowledge infrastructure that we’re existing that existed back
when it was Egyptians writing hieroglyphics onto papyrus and before that cave
painting and before that, right?
We know the knowledge infrastructure is always shifted from when Gutenberg did
his thing to when the original libraries happened in Babylonia and in the Korean
hills and bamboo, um, books and materials.
We know that that knowledge infrastructure changes and that’s what we as
librarians pay attention to and not only pay attention to in nice little slices,
but across the domain of human endeavor.
We have librarians that yes,
look at scholarship and academia and look at the production of information
through a scientific process. We have, um,
scholars sitting in religious institutions and theological libraries that look
at knowledge prediction and understanding in a, in a faith context.
We have academic,
we have public libraries working across rural settings and urban settings.
We have school libraries working from K to . We work across this domain.
Academic libraries, law libraries,
medical libraries and bias librarians continually connecting with each other
across these perceived boundaries.
We have a really good sense of how this knowledge infrastructure is shifting and
changing more so than any other domain perhaps than an
information scientist. But this is what we pay attention to.
So we need to use the span of our libraries to understand this world and we have
that network and we can reinforce that network.
The last reason I wanna talk about why,
why librarians are so well situated for this land and AI is around agility.
Librarians has show,
have shown persistence of focus and agility through dramatic read
We have gone through so much and we constantly churn.
We are the ultimate lifelong learners. Learning about new technologies,
new ideas, new impact, new community development, uh,
librarians in, um,
in the UK that monitor their community and provide regular reports,
but regular reports to council members so those council members can be better
informed about community and community needs.
We know that medical librarians are constantly looking across the literature,
looking at different research so that they can inform medical practice.
We know that all of this thing that librarians aren’t simply a matter of
but we’re part of the conversation and we can share this and we know that the
technology to do that has changed dramatically. Now
I feel your pain, right? By the way,
this is the prompt I used to generate this image profile of tired old librarian
sits in futuristic library with her head in her hands contemplating a tough
life, right? If you’re retiring this year,
if you plan to retire in the next five years, think about say a year career.
You saw the advent of desktop computing when the idea was it moved from
computers and businesses where it took up multiple rooms to what sat on your
created things like desktop publishing where anyone could put out a newsletter
or produce something that looked professional.
We had the ability to suddenly do budgets and work electronically and write in
It changed to education when word processing came in and people didn’t correct
anything on paper anymore,
it changed medicine as eventually people could get medical information and share
it back and forth.
That was a massive change to society when we
saw digitization. And this isn’t just a matter of scanning things.
I mean when analog things went digital,
when the publishing industry went from paper production to digital production
and digital production to digital distribution started as paper and now is both.
That was a massive change. Not only scanning text,
but when born digital documents outnumbered everything else in our life.
The telecommunication revolution where phones went from boxes on walls with
wires to aluminum and glass slabs that slip into our pockets
where it used to be about making phone calls and now it’s about texting and
playing games and interpreting the world and, and,
and when the internet came out and then when the web came out and
amplified the internet, um, explosion. And by the way, yes,
I know the internet’s been more than years old.
I’m talking about sort of its break to libraries at large in the common
zeitgeist, if you will,
massive changes when everything was online and you had to vote online and you
had to see things online, et cetera. I get your pain over years.
Not only have these been tremendous seismic shifts in the knowledge
infrastructure and therefore our profession,
but they’ve gotten closer and closer together.
We are dealing with larger change more often and that can be exhausting.
But now note the query to generate this image,
a triumphant librarian standing before a mass of people guiding them towards a
brighter future because that’s what we can do.
Not only did we go through those transitions,
but our communities went through those transitions.
Our governments went through those transitions.
Our institutions are still going through those traditions,
through those transitions, and we have survived those. I mean,
once again, when the, when desktop computing first came out, eh,
not so much is that gonna be apparel libraries,
but it turned into a massive thing about public access computing when
digitization, oh, will anyone need the books anymore?
Will anyone need us anymore? Will it all be online? Telecommunications,
revolutions in the phone, the internet, the web.
All of those had this sort of interesting starting query about, well,
are we still relevant?
What I love about the current discussion around generative AI in the library
domain is it’s very much proactive.
Here’s how you can evaluate, here’s how you can use it.
Here’s what we should be doing in it. Ooh, here’s, I’ve tried this here.
Look at this.
This is a chance that we can look at this not as an existential crisis for us,
but as an opportunity to help those that see an existential crisis.
The publishing industry from this experience I’ve had over just the past couple
of weeks are kind of terrified of generative ai.
They’re being overwhelmed by AI submissions they’re worried about in the digital
humanities. Is there still a role for the scholar in writing up this material?
Is there still a role for art production?
We’re seeing this push to the idea of is this,
are we still going to have academia? Are we still going to have publishers?
Are we still going to have writers? Are we still gonna have reporters?
And then one of those that are not on the list is,
are we gonna still have librarians? ’cause the librarians are like, yeah,
we’re here for it. We need it. Because all of those things,
whether they’re replaced or not,
are gonna still produce information that goes into this chaotic knowledge
infrastructure and they’re gonna need guidance, which is us. So
here’s a few things to think about and what we do about this. First,
realize nothing will change without structural change.
There’s kind of a reason that this chimera of knowledge infrastructure we have
didn’t radically change the world in a very obvious way until now.
And that is there’s a lot of entrenched stakeholders that didn’t want it to,
right? Google didn’t want to create something that replaced Google,
so they used AI to make a bigger Google, right?
That idea that Spotify and Apple and YouTube didn’t wanna replace
themselves with whole new ways of discovering and for and producing music.
And so they wanted to look like it used to look,
even though they may be gaining efficiencies on the backend and be able to
monetize the data that’s being generated from it.
We see this entrenchment in intellectual property and the other entrenched
stakeholders, and I use eBooks as an example. EBooks,
we’re supposed to change the world and now they’re just boring.
So when the music industry lost a couple of billion dollars of valuation when
digital music became widely available through Napster,
and that’s eventually through cents iTunes songs that destroyed the idea of
an album. This whole transition, the publisher said,
we’re going to wait and see what happens. And when they did finally come forth,
not only did they create something that looks remarkably like a paper book,
but it’s in digital form,
but they changed the business model to reinforce their fa
false scarcity, right?
They wanted to be able to sell digital books like they sold physical books.
They wanted to be able to not sell one to a library,
but sell and sell ’em over and over and over again like they did physical
books. And so they didn’t move until they could fit the genie back into a
seemingly similar BO bottle. And for that,
we have suffered the ebook revolution didn’t occur and it didn’t occur because
there isn’t opportunities in eBooks. Yes, people like reading paper. Yes,
the younger you are, you like reading paper, I get that.
But rather than coming up with something new that you couldn’t do in paper with
the embedding in multimedia with social learning activities,
with the idea of hypertext truly within this thing,
it stopped because entrenched folks didn’t want to.
We also see this within copyright.
We’re celebrating the th anniversary of hiphop.
Hiphop started by remixing and chummy up with new tunes and really
creation and creative ideas that not only added new things but
reinterpreted old things. But a copyright regime came in,
new enforcement came in, licensing came in,
and we’ve decreased the amount of ability to innovate in that market because
this loan genius copyright narrative is winning.
That is the idea that that, you know,
malt Disney invented Mickey Mouse and he deserves everything from Mickey Mouse.
Not that he was inspired by previous artists,
not that he was inspired by his own artists and illustrators and changed it.
Not the fact that he’s been dead for a couple of years and yet still as a
lone genius, he and his family and his corporation deserve that.
They deserve lifetime plus years. That’s insane ai.
Everyone’s talking about intelligent amplification at the end. That idea that,
well, they do the first draft, but the human goes and makes the change. Well,
that’s how we do it all the time in everything we do. That social narrative,
that social construct of ideas doesn’t fit into our AI narrative.
But unless we change our views of ai, uh, of copyright,
we’re going to limit the potential.
Now we’re already seeing some interesting light. So for example,
in the US the idea is that anything produced by AI cannot be put under
copyright because it needs human production.
That’s not only from the Library of Congress,
but it’s also from some recent court cases and it will be discussed and the
entrenched interest will go in and try to do their best to figure out how that
doesn’t work that way, but we have to rethink it.
The other thing we have to rethink and lead the way in is around the
idea of sustainability. So unlike the web,
when the web first came out, anyone could download a web server.
Anyone could download a web browser.
The idea was that it became a community of people who tried it,
created open source community. One of the most,
one of the biggest web servers out there is Apache, which is open source.
One of the biggest web services when you go to websites out there is WordPress,
which is open source. This idea that anyone could come in and play,
but when chat G p T, when OpenAI was trained chat GT three,
they used , gallons of water to train
the equivalent of car productions just to cool the computers that were
chomping on all the texts they found on the internet. That’s not sustainable.
That idea that every time you make a new image, every time you put in a query,
it’s going to generate and use power.
It’s going to generate global warming gases, it’s gonna utilize water.
We’re seeing that there’s a,
this sustainability and climate crisis angle to this as well that we need to
be a part of. We need to help lead our communities to develop ethical,
explainable and regulated ai,
ethical in the sense that it doesn’t just go and take a bunch of copyright books
and reproduce them ethical in that it doesn’t strip a bunch of images that
people put up to for,
to sell and be able to pop out a non copyright version in seconds,
but it’s still based on that person’s work.
We need to talk about ethical in that it’s easy now to create deep fakes and
voice fakes and images and disinformation,
and we need to ensure that people understand the cost of that and work against
that explainable in that much of the AI that we’re talking about in the
When they find breast cancer in different mammogram images,
when they find likelihood of recidivism for a prisoner, when they find,
um, lack of productivity in employee,
how did they come up with that?
The internal logic it uses is mostly just pattern recognition and the patterns
aren’t things that humans would readily understand or, or know about.
How can we enforce the idea that any system that makes a determination or
decision has to be able to justify or explain it?
That’s an ethic that we’re used to within human workers.
We need to begin to expect that of our AI counterparts and regulated
ai. And that is the fact that just like I said,
the web was easy and cheap and anyone could do it on any website and ebook
readers were cheap and could be anywhere.
Running one of these large language models takes a huge investment of capital,
takes a huge investment of equipment, takes a huge investment of money.
And that means that right now AI is being led by Google, apple,
Microsoft Meta, and probably China and some government actors.
And so there isn’t a democratizing ability where we can go in and create our own
AI that’s built around our indigenous knowledge system.
We can’t take all our documents that represent our community,
run them through this training system,
and then be able to have our children and our coworkers query that
system and take advantage of the ability for this human-like
domain processing power and domain without handing it to the rest of the world.
Right? We need to talk about that kind of regulation.
I’m not so worried about it taking over and launching the nukes.
I am extraordinarily worried about it,
creating a neocolonialism view where basically the companies with a
lot of power and a lot of money get to make it look like companies with a lot of
power and a lot of money and serve some artificial majority.
Those are the things we need as librarians to be a voice.
And part of we need to continue to become the publisher of the
Moving from AI where just we have the machine generated to intelligence
amplification where it really is working hand in hand.
It’s working with that mother who wants to write her first children’s book but
doesn’t have access to an illustrator.
It’s working with that illustrator who may not be very,
very clean with their language and be able to utilize words to capture their
ideas. It needs to be worked with the marketer,
the person putting together their cv, the the business plan, et cetera,
where these technologies can speed and make more efficient the production of
information if guided properly, but the true way forward,
the true way forward.
Where librarians really need to take up this opportunity is around weaving
together our community.
We need to knit together the community that talks about how we share,
share and seek power.
How do we make sure that these technologies empower minoritized communities?
How do we make sure that there is equity in this work?
How do we make sure that it is not simply a way to quickly generate laws
that favor one ideology over another?
How can we use this as a way of identifying disinformation that is seeking to
disenfranchise? How can we use this technology to seek justice for all?
How can we build a movement?
How can we be working in our communities?
Whether it’s a small public library in Denmark, whether it’s a large urban, uh,
academic library in Texas, whether it’s a national library,
whether it’s a school library, a medical library.
How can we not sit in our silos and go, what’s it going to affect me today,
tomorrow? But instead say,
how can we work together to guarantee a better tomorrow for
all? That’s the promise of librarianship.
It has never been about collecting things.
It has been about collecting ideas to empower a community.
We used to collect them in forms of books and narratives and linear narratives
that have helped our community learn and generate their own narratives and their
own expertise. Well,
now we’re going beyond those tools that we’re working directly with our
communities. We’re working on what they want to learn,
but also what they want to teach. We’re working on what they want to know,
but also what they want other people to know. And in doing that,
we must facilitate an equitable environment for that to happen so that the
majority does not eliminate diversity in ideas.
’cause we know the best ideas come from the most diverse sources and that
includes the most diverse set of people.
We need to continue the mission of librarians to make a
better world, to improve society.
We need to do that by facilitating,
by aiding learning knowledge creation within our communities.
That’s what drives us. But it’s not just our community alone.
We need every other librarian with their community doing the same thing.
And by joining hand and building on the infrastructure we already have as a
profession, we can dominate ethical ai.
We can become part of shaping this technology to ensure what is
best for our communities,
not what is best for the people who have enough power to put through
water systems to cool massive grabs of people’s intellectual
property. We need to talk about how we change policy.
We need to talk about how we ensure our tools are ethical.
We need to help shape the policy within it. Because once again,
it’s not computer scientists that think about the information
impact and the knowledge infrastructure. They think about the tools.
It’s not the lawmakers who think about the whole perspective of prosperity.
They think about the policies.
It’s not just the communities that worry about whether they’re gonna have a job
or not. It’s all the people that fit within it.
It’s not just the people who publish the books and sell the books.
It’s everyone. It’s the full knowledge infrastructure.
The only ones who think across that domain every single
day to answer every single question, to pick every single book for a collection,
to select every program we do.
The only people who regularly encounter the invisible knowledge infrastructure
make it manifest and begin to shape and deal with it are librarians.
This is our time. Yes, there will be stumbling blocks. Yes,
there will be stupid policies that will seek to enforce my majority views on
all. And we need to fight them. And yes,
there will be technology that seeks to ignore us and and profit the
one group. And we need to fight that and counter that.
We need to host this technology. We need to shape this technology.
We need to educate on this technology and educate that technology.
We need to be part of all parts of our knowledge infrastructure and we can do
it. And we’ve done it before and we’re starting from such a great place.
We’re not starting with the idea that libraries are about to be outta business
by yet a new technology we’re starting with our communities are gonna be
dramatically impacted from this.
How can we ensure that our communities thrive in this environment?
We come at this with a positive light of leadership.
That’s my thinking for the day. I welcome your ideas.