22 08 16 Science Week 1
*Speakers: Marie-Louise Ayres (M), Alex Zafiroglu (A)
M: My name is Marie-Louise Ayres and I'm the Director General of the Library. As we begin I’d like to acknowledge Australia’s first nations peoples, the first Australians as the traditional owners and custodians of this land on which we are privileged to do our work and give my respects to their elders past and present and through them to all Australian Aboriginal and Torres Strait Islander people.
Thank you for attending this event either in person or online which is coming to you from Ngunawal and Ngambri country. May I just say we’d love to have a conversation like this in the more intimate atmosphere of our very beautiful conference room on the fourth floor which many of you will know. You probably also know that our roof was destroyed in the hailstorm in January 2020. We’ve been without our fourth floor since then and will be for another year so we have to have intimate conversations in quite large arenas at the moment.
So thank you for attending the event which is celebrating National Science Week 2022. The Library and AI, this event is the first event in a series co-presented by the Library and the ANU School of Cybernetics. During this series of cybernetic thinking for a new world we’ll examine the connections between libraries, language and cybernetics. While some of you may see science and libraries as a strange partnership science and technology drive us forward and help us create new and exciting ways to engage with our collections. Libraries are often leaders in creating systems that allow people to explore collections. We create IT systems and platforms that are cutting edge of which Trove is the obvious example.
So cybernetics first found form in the 1940s and ‘50s as a response to the rapid expansions in computing technology following world war two, fusing maths, engineering and philosophy with biology, psychology, anthropology and many other fields. From its inception cybernetics was a generative, intellectual wellspring shaping everything from AI to critical systems theory, computer-driven art and music, design, thinking and of course the internet. Tonight we’ll discuss applying cybernetic thinking to the libraries of the future. In doing so we hope to better understand our present moment in its complexity, think differently about it, ask critical questions, make meaningful change and shape the futures that we want to see.
So with me this evening is Professor Alex Zafiroglu. Alex is Deputy Director at ANU School of Cybernetics. After completing her PhD in cultural anthropology at Brown University in 2004 Alex commenced a diverse career at Intel Corporation. During her 15 years at Intel Alex made significant contributions to the R and D and commercial development of technology across the Advanced Research, Digital Home and Internet of Things Divisions and is listed as a coinventor on 11 patents. I have to say my count there is zero, Alex.
In 2016 Alex was appointed Principal Engineer in Social Science with the Internet of Things Division and was Intel’s foremost domain expert and research practitioner around homes and home life. She’s authored multiple publications, presented at academic, regional and industry conferences and is a member of the American Anthropological Association and the Ethnographic Praxis in Industry Community. She’s interested in the role that social scientists play in exercising ethics, accountability and data rights in the deployment of commercial cyberphysical systems. So please join me in welcoming Alex.
M: Over to you to kick us off, Alex.
A: Thank you so much, that was such a nice introduction. I wanted to start also by paying my respects to the traditional owners of the country on which we’re meeting, the Ngunnawal and Ngambri people and to acknowledge the incredible role that they’ve played in passing down millennia of knowledge of this country and what a delight it is to be here and to be talking about knowledge systems with the National Library.
Today we’re going to talk about a body of work that we did together last year that produced a report called Custodians and Midwives, the Library of the Future and we liked it so much that we each brought our own –
M: Our own copy.
A: Our own copies. It is also available on the NLA website as well as the School of Cybernetics website. Would you like me to talk a little bit about how we came to do this?
M: Yeah, I think so or about – perhaps we talk about how you came to do it from your perspective and then I’m happy to talk about how we came to do it from our perspective too so why did you say yes when we came to you?
A: Really? Because you're the National Library and you’re amazing. Because we are very interested in thinking about cybernetic systems and cybernetic futures, particularly at the national scale and we understand that our mandate as a national university is to create change and to engage with national organisations. The Library, in particular, has long been a leader in the development of creative new systems, often using advanced technologies and the types of technologies that we now refer to as artificial intelligence. So the opportunity to talk with you about the work that you’re doing and to think about what the opportunities are for the Library as you consider new technologies to integrate into your collection systems was quite exciting. I feel like I’m lisping a little bit.
M: It’s okay.
A: Is that okay? Okay.
M: Well I guess from our perspective if I just take you through what our thinking was and how it - because it changed a lot through the project. We have been using these technologies for many years. Of course the optical character recognition that we apply to newspapers is an example of machine learning and we’re working with a company to progressively increase the accuracy of our OCR and our intent here is quite clear, to improve the accuracy of OCR so that when people search Trove they get better results back.
I think the other thing that might be less obvious to our users is all the algorithms that we use or reuse or tinker with to get the best possible relevance ranking, our collection is absolutely enormous so trying to get useful research results back to a user through a search box is quite a big task. Our largest text-based collection – that’s the Australian web archive – involved months of work on algorithms to make sure that the best and most useful results turned up so that’s probably not so visible.
I guess when we started to think more about this we were thinking about the scale of our collections, huge, about 10 million physical items, 280km of shelving. Our digital collections are enormous so at the end of last year our web archive is billions of files. Just to give you an indication, just our four bulk harvests of Australian government websites last year added 6 terabytes or 70 million files to our collection so they’re just beyond human beings to deal with them.
We're also really thinking about maximising access to our collection so when I think about this from the frameworks we’re thinking about the first is the kind of OCR example, improving access to the collection for immediate or near-term benefit. That’s one framework we can think about. A second is to think more broadly about things we can’t currently do and I thought I’d just give you a couple of examples of things we’ve thought about here and they’re not nearly as exciting as what Alex and team have thought about.
One that I’ve had in my mind for a long time relates to our 19th and early 20th century manuscripts. We know that our collection is full of hidden records and snippets of first nations languages and descriptions of culture, we have a lot of these. So we also know that handwriting recognition, which has been a holy grail for my entire career, is quite mature now. So would it be possible to digitise more 19th and early 20th century collections, put those together with already digitised material, apply AI technologies that did something like this? Look at all the words in this text, find the words that don’t look like English or other languages for which AI technologies exist, tell us something about their characteristics and maybe we’d be able to narrow in on resources that have got language material and then engage with community so that’s an example.
The other is one that’s actually a bit funnier. We have a huge image collection and about a million images and probably hundreds of thousands of those don’t have item level description. So we have many, many images of sheep in paddocks, right? So a human being looking at one of these might be able to say this is a large sheep in a big dry paddock or ewe and lambs in springtime but maybe image recognition could do other things. Maybe it would be able to detect whether the sheep is a Corriedale or a Merino. Maybe we’d be able to automatically detect drought conditions. Maybe we could link up to images from the CSIRO to identify the trees in the background and giving us an idea of the likely location so you got to kind of start with these ideas of what can’t you do now.
But the other thing that was in my mind is about – and I'm sorry to bring in the banal – is driving down the cost of making collections accessible. So our collection is growing really, really rapidly, the existing collection is beyond our resourcing to use human beings to describe and if I think about the last 10 years the Australian population has increased by 20%. We’ve seen an explosion of digital content that we must collect and provide access to legally but we have one-third fewer staff than we did 10 years ago and that tends to really concentrate your mind on your ways of doing things.
We've also experimented with these technologies but as we did I started to have a lot of worries so I started from a deficit or a risk perspective. I was thinking okay, if we use image recognition software and we start to apply facial recognition software could we cause harm unintentionally to individuals or communities? We see this everywhere at the moment, that kind of thing. If we try to drive down our costs by using machine transcription of oral histories what if it doesn’t get it quite right and we cause kind of serious offence? I think as we’ve learned more about this work we’ve also learned that using these available third party technologies kind of poses a negative risk, that is, that the datasets on which these tools are developed are inevitably partial, leaving some groups out and of course no technology is neutral. On Thursday we have another fantastic Science Week talk from the School of Cybernetics PhD student, Kathy Reid, who’ll be talking about the consequences when languages are left out of speech recognition tools.
So I’d say I came to this from a position of worry, thinking what if we do these things without understanding the consequences and we cause harm unintentionally? I knew that we needed to engage with the human parts of the work and I knew we had the right people across the lake at the School of Cybernetics. So that’s kind of where I started from but it is not where these wonderful people took us so I’ll hand over to you -
A: It is definitely not where we took it. I would say that – you mentioned earlier that my training is in the social sciences and so – and even though I worked in the tech world for quite a while I tend to think about understanding the world around us as a set of opportunities and a set of ways that people are in the world that suggest multiple futures. So I’m not necessarily a designer or a technologist who is trained to think about what is the problem? Tell me what the problem is and then I will address the problem, fix the problem. So I tend to think a bit more reflectively and generatively and with less necessarily decisions about having one right answer.
What we proposed to you in response to your request to think about what the future of technology is and what some of the opportunities and risks would be is to take a little bit of a step back because we were very keen to create something with you and I'm being very careful to talk about this as a joint project -
M: ‘Cause it was.
A: - and a creating with rather than a research project from a university team that came in and talked with you. Really it was in itself a cybernetic process of conversations and dialogue and figuring out where we needed to be to produce the types of insights that were going to be useful to you. But my point here is taking a step back and thinking about what are the overall models and structures that you have in place that would help you think about the possibilities for future technology capabilities.
When I say future ones I mean both what’s in the very near term and kind of what you already have in your toolkit including optical character recognition as well as capabilities that were a little bit further out including various types of machine transcription so you talked about being – taking images or video or audio or other media and translating it or transferring it into machine-readable text that then can be acted upon through to things like machine-actionable collections and ideas of transparent AI so more to agendas and aspirational goals through to the very beginnings of the types of capabilities you already had.
Thinking about those as technology capabilities that would develop over time rather than what are solutions that are on the market today and what might happen if we use this particular solution? Because then you have a report that lasts maybe six months and this is designed to be something that you can use for many years to come.
M: I think - that’s the advantage of working with a fantastic partner like this who doesn’t have to worry about the kind of risk or how many dollars you’ve got in the bank, it kind of lifts your thinking.
Alex, I was thinking about you know you said you took a cybernetic and an anthropological approach. You had many conversations with many people in the Library -
A: We did.
M: - at the beginning, along the way and then I think importantly afterwards also we had an opportunity to reflect back so how was that from your perspective? I mean those experiences of talking to quite a number of individuals in the organisations.
A: It was fantastic and amazing and there’s a little joke that any time I interview someone I come out of it saying they’re my new favourite person so I had about 15 new favourite people. I say that because it is always a privilege and a delight to talk to people that are so knowledgeable about an area and have them share that knowledge with me and increase my understanding. So yeah, we had conversations ongoing with you, of course, with people in Collections, in people in Digital or your IT Department, in people in Reader Services, in people in Archives and I'm – Engagement Office. I’m trying to think of the other ones without -
M: I think you’ve got all -
A: - but I think I got quite a range there. These were conversations that we had really about what collections is like and what you are doing with the Library with your core mandate of collecting and organising and making accessible resources of national significance to Australia. Also just about the – a little bit both about high-level processes as well as the day-to-day work that they’re doing.
We also had the opportunity to make sure that the conversations that we had were reflecting – making sure that we got it right so we sent transcripts back and things like that and we had ongoing conversations that really changed what we were doing and even as we did those conversations we got a much richer understanding of all the different perspectives on collections at the Library and all the different types of people that touch on the library.
But really what we were inspired by at the very beginning were some of the early works in cybernetics which has always had an interest in the libraries so if you look back at the work in the 1967 archivist at the Library of Congress doing work in calling on librarians to start thinking about the automation of indexing, the automation of collections and arguing that cybernetics should be part of that through to Heinz Von Foerster who was a famous American scientist. He gave a talk at the University of Wisconsin’s Library Institute which is now the ICE School in which he has this quote which I promise I will never read anything off a screen but what is so delightful about this is that it gives us an opportunity to think about the library as a system and as a cybernetic system and one in which there are two roles that it plays in the fabric of society.
So it’s not something off to the side but it is central to what it means to be part of a community and the idea that the library is a place of - where you can acquire knowledge and that acquisition of knowledge is both as – libraries as custodians of collections and also as places that kind of give birth or – they don’t give birth themselves but they are midwives to the creation of new knowledge by the audiences that come and interact with the library.
So we were really interested in what were the processes behind that and what type of an organisation is a library and how we might use that conception of the library as a cybernetic system to help you think about what the processes are around new AI capabilities.
M: I think that idea of – you’ve mentioned the word process quite a few times -
A: Yes, I probably did.
M: - and I think it’s important because these days we understand so many more phenomenon, whether they’re natural or otherwise, as processes rather than kind of finite things. But also because libraries are full of processes. Some of them are novel and many of them are repeatable and just get repeated over and over again and over again, you kind of try to improve them. I suppose the work that we were doing here is almost kind of putting those two things together, really, thinking about repeatable processes but also novel ways of thinking about it to get to a better understanding of what might be possible so yeah.
A: So we – just to build on what you were saying we are really keen on understanding libraries in general and particularly the National Library as a cybernetic system so the Library and the collections in the Library are not a thing, they’re a set of dynamic processes, they’re a set of dynamic relationships between ecological or kind of natural systems, between human systems and between technology systems. Understanding the interplay of the resources and the people and the processes and the infrastructure and the data that go into making the Library what it is, is the approach that we took.
In particular what you were saying about the Library being full of processes, I would say the Library is full of components that together make a system and that system is dynamic. By – what I mean by that is that it’s never the same thing twice just like your collection is never the same thing twice and every time you use those numbers I get like a chill about like the kilometres of content and how many new pieces of content are coming into the Library per year. I get actually a little bit of a hot flush. It’s really quite exciting but it’s like Heraclitus’ river, you're never stepping into the same thing twice.
So understanding how things change over time is quite important how we think about stuff and the connection between the Library as a system and concepts of time is quite interesting because in your Collections strategy for 2021 you talk about collecting now what will be important in the future which is a really difficult thing to figure out, I would think. When we had conversations with the people that work at the Library it became very clear that you were sitting in a present moment but it is always and already about the past in say 40, 50 years in the past and the decisions that were made then about what to collect and how to label it etc and where does it go and how will we be able to find it again? Also your future users and what they will want to be doing in 40 or 50 years.
So one way of going about to understand that is to consider the Library as a cybernetic system and what we ended up doing when we did an analysis of the Library was thinking about five components of the Library as a system. There’s the data which -
M: Lots of it.
A: Lots of it, lots of it with the data that’s in the collection itself and then data about the collection and about your users and about your employees etc. There are the various processes that go into collections, there are the agendas or the context around which the Library is situated. There is the infrastructure which is everything from the digital infrastructure and all of those platforms as well as the physical infrastructure of the building and the kind of physicality of how you take care of collections. Then there are agents or people so who is actually doing this work both in the Library and in various roles? Together these things form the system of the Library and we were particularly interested in understanding the dynamics in the Library so what sets of opposing forces generate the work in the Library?
So we came up with five of them including this tension that you might feel between quality and quantity which I think you’ve talked about already, thinking about the Library both as a place in itself as a node and as part of a network of multiple libraries around the world, thinking about the relationship between indexing and being able to organise information into smaller and smaller folders basically to get the thing that you want and serendipity and the idea of being able to make associative connections among the material, and a few others including are your – who is an agent and who is an audience and that your audiences so the users of the Library may actually be changing with the development of things like machine-actionable collections and what happens when your users are also some of your creators?
So we used these types of five parts of the system to ask questions about well what data has evolved in collections and what processes etc and to ask a series of questions between data and infrastructure and what are the dynamics between data and infrastructure or agents and agendas etc? What came out of that is – we’ll leave it there for a second – what came out of that is really an analysis of four categories of AI capability so that we talked about optical character recognition, machine-actionable collections, machine transcription and responsible or transparent artificial intelligence and an analysis that helped you understand some of the opportunities with these technologies or with these solutions, some of the risks so those would be things that in the use case that you thought you would use a technology for might be an unintended or it might be a bad consequence.
Pitfalls which would be the same thing but if you ended up using the technology in a new way so for example what you were saying about facial recognition, that when you originally collected the items there may not have been the types of technology available to be able to do the types of search that you can do now, and of course issues which would be kind of known problems with the implementation of the technology. We chose to think at the level of tech capabilities rather than particular solutions so that you can reuse this process over and over.
One of the things that we did in workshops with the National Library and at the School was to start to take these capabilities and play them out over time. So to think about well this is what machine transcription looks like now, what could it look like in 20 years? We’re very careful to start using signals from what is available in the market today and what we know is happening in research labs around the world to spin out what those scenarios would look like. One of them that we did, and this one was for optical character recognition, is the one about the second – the avatar of the second Prime Minster of Australia.
M: Which I’ve shared with a High Court judge with a special interest in Deakin, much to his kind of delight. I guess if we think about this, this is one example and this is a wonderful work of Charlotte Bradley -
A: Yes, it is, one of our PhD students.
M: PhD students but that approach of – so you’ve talked us through kind of systems, you’ve also talked about time and I might digress about time for a moment because you’re absolutely right about that sense that we’re in this present moment but the great privilege of work in a place like this and my great privilege leading the organisation for a chunk of time is that you know that the collections we have have been carefully developed and stewarded by generations before you and they’re going to last longer than you will, much, much longer than you will. But the other thing is over time the meanings in already existing collections change even when technologies aren’t changing because researchers always have different questions.
I’ve been involved in our kind of research activities here for 20 years and I continue to be incredibly surprised at the new uses to which collections I thought I knew really well could be put so research questions, whether it’s academic research questions or the community, the same material generates completely different meanings for individuals and also over time.
So that’s just the time digression but if we’re sort of talking about time and we’ve talked about systems and the cybernetic style will be really helpful for us to think about as we’re thinking about opportunities, kind of testing it against those things. Moving into a speculative fiction approach for how to convey this, where did that come from? Just more talent inside the School?
A: Partly. Charlotte Bradley who was primarily responsible for running the workshops and running up these scenarios in collaboration with the National Library and with other researchers at the School actually has a background in drama and theatre so partially that’s where this talent comes from but she is also an excellent researcher, able to pull together the threads that we are hearing from reading documentation from the Library, reading the larger literature, doing the interviews and the conversations that we had with the Library.
This is really a research method that's called speculative futures so it’s less about speculative fiction per se where you can think of science fiction showing us possibilities of the world in the future that we then may turn into science fact. We find the arts and creativity a really useful way for highlighting and uncovering some of the values that we may not realise that we have when we are thinking about the future. So we can produce these or produce these with partners and then kind of analyse them back.
In this case the story is around an avatar of Deakin that has been built in part from materials from the Library and it raises a series of questions about how reliable is this avatar? Can it be called upon in court? Should the Prime Minister potentially be talking to this avatar to get its – it’s really not an it which is kind of the fun of it – take on policy moving forward?
M: I think it would be fair to say that this approach on the speculative futures, oh look, it just went straight home to me and that’s actually something around this work too that I was reflecting on today, that the Library itself is full of people with training in and deep adherence to a whole variety of disciplines from different fields. We also have within our community, we have great systems thinkers, we have really great process thinkers where thinking from the systems perspective is really useful and then we've got people like me, straight up and down humanities person, interested in the narrative arc.
So having a narrative approach to this really, really appealed, I think, or at least for me and I think for quite a few of our colleagues, it got us immediately into what are we talking about here? So as a – in fact if I think about the kind of book as a system as well it works on different levels for people whose kind of disciplinary or thought backgrounds are different and that’s a real strength as well, something I didn’t expect to get.
A: Yeah, we were very particular about starting each of the sections on the AI capabilities with the story and then being able to kind of peel back the layers of how we actually got to something this far out using what we understood about the opportunities and risks and pitfalls and issues around current instantiations of that capability and where it was going in the future. So it’s not just a fun story or a funny story, they’re very carefully crafted towards particular ends to generate particular conversations and not just as an illustration of the research.
So we provided you with kind of printout posters of them as well that we hope that you’ll be able to use and using them with you in workshops as well. There are four of these scenarios in the report, I encourage you to look online and see them. They are all - even as I said they’re not kind of supposed to be just fun, they are delightful to read as well as serious analysis.
M: Really thought-provoking. So Alex, you’ve kind of talked about we’ll be able to use these in kind of workshops and things and that leads me to – and while we’re on it, thinking about where to from here with this piece of work? We - collectively as soon as we saw this piece of work we knew about it, we knew that it needed a bigger audience because it generates such important conversations. COVID kind of made that quite difficult last year but we’re back on track now.
So we’ve already of course done a run of these and in fact your colleague, Professor Genevieve Bell, came and spoke to our Council members in May and I can tell you there’s no other way that we could have kind of gotten them to understand what we’re talking about than a storytelling approach. We have with your permission shared it with another national library that’s really interested, we’ve got School of Cybernetics folks coming to talk to the CEOs of National Estate Libraries Australasia next March and to do workshops with staff who are starting to kind of get it out a little bit more to conferences etc ‘cause I think it’s a piece of work that needs a broader audience.
Then for us inside the Library we’re about to embark on our next big piece of long-term strategic thinking and this will be one of the foundation pieces. The CSIRO’s recent report on – the megatrends report will be another so for us having a report like this doesn’t sit on the shelf. I mean it could sit on the shelf and we’ve got lots of shelves that it could go onto. It’s to communicate, it’s to help us and I think to help ANU to advance our special part of the kind of leadership of conversations like this in this country and beyond. Then for us at some point of course we have to make decisions here about where we’re going to put our investment, where we are going to – where do we think we might get the best returns for an immediate audience, midterm audience or very long? You can’t really predict those things, you just have to make decisions.
How do we balance out our desire to do a lot more about this with our need just to keep our existing large digital services just running let alone improving them? So at some point for us this will come down to investment decisions but I feel really confident that having done this work with the School that we will be coming to it with a much better framework around who might benefit and then when you're thinking from that perspective and you’re thinking how they might benefit you can think which technologies should we apply first to which collections for which purposes and for which intended beneficiaries?
While this is a really long-term thing if we think about how can we make things better for short-term or medium-term audiences we’re probably if we’re careful more or less on the right track for the long term. That’s what I hope ‘cause I won’t be around to see this. So – yeah so how do you kind of see work like this – you don’t have to run a big library but you’ve been at a big corporation where things have turned from ideas and concepts into making those decisions and generating a kind of a product so just interested in your reflections on that about when the rubber hits the road what does it mean when you’ve already done this kind of thinking?
A: I think it’s critical doing this type of thinking at the very beginning of any business strategy that you have and part of what you’re doing, and I don’t know if I’m allowed to say this, to talk about your business strategy -
M: Yeah, you're allowed to.
A: You have a business strategy, right? You have a finite set of resources and you’re making decisions about how best to meet your goals. When you don’t do the work kind of earlier on to understand what are the core concepts that are driving your decision-making and how you’re making those decisions and why you end up with very short-term - you can end up with really great short-term profits but no longevity. So understanding at a very general level, not just what is the problem and how you fix it but thinking generatively from the beginning about what is the opportunity and what is the type of world that I want to create? So from being generative rather than defensive, that’s what I believe that this approach will do for you and it is certainly the types of things that I was trying to do with varied success in various product groups at Intel.
When you do get it right and you do present a way of thinking about what you’re doing it can often look very, very simple or it can look very like well we knew that already. My kind of measure if I’d gotten my stakeholders to really understand something was to tell them like oh I knew that already, you didn’t really do much. I’m not accusing you of saying that we didn’t really do much but the idea that really substantive change in your thinking can be very basic and very powerful. It doesn’t need to be in the weeds about the latest technology that’s at work right now but a broader approach to thinking about change.
M: But you used that word substantive and it just – that was resonating with – I guess I was thinking about the alternative approach, the alternative approach to taking this kind of approach and over the last 10 years libraries around the world and in fact in this country have developed kind of labs or experimental kind of groups of people who are testing out these technologies. To my knowledge there has never been a piece of work that’s been done like this so it steps right back and you’re right that in all cases things have been learnt but often they’re very, very short-term gains, the lab disappears in a few years and actually nothing further happens so – and that’s probably I suppose like a product that might look really bright and shiny for a while but it doesn’t have longevity so -
A: Or it’s the product that was 10 years too soon and that it comes back so I think about PDAs and the Newtons and things like that of the world and then coming back 10, 15 years later as -
M: When the time is right.
A: - as the smartphone.
M: Yeah, okay. Also we come back to time again -
A: So we come back to time and like what is the right timing and how do you make sure that it’s not just the thing that you're looking at but the set of relationships between the resources that you have and the people that you have and the infrastructure that’s already in place that will make a particular solution stick or not stick at a moment in time.
M: Yeah, yeah. Well look, from our perspective it’s been a great piece of work. It’s been really – it was really exciting and as I said earlier moving from my – I won’t say I was a narrow risk but it’s my job to think what could go wrong here – moving from that to a much more expansive approach and moving to an approach that really – you talk to a lot of people in the Library who as you say have really deep knowledge of their kind of specificity of what they do I think you mentioned at one point. It was also I think such a great opportunity for science, social sciences and humanities to meet in one place. Again because we’ve got all of these kind of disciplines of ways of thinking in our workforce I think we got a really kind of rich result from that.
I will say that when I saw the report I just was delighted, I just – it just spoke to me straight away and I knew immediately how important it was and the kinds of things that might come from it later on. I was particularly impressed by the case studies that Charlotte Bradley did but that’s ‘cause I’m a kind of humanities tech story person so that’s what’s going to talk to me.
So I think it was work created by two very committed teams and with a really professional crew from the ANU. I think actually those post-project interviews that we did and got a couple of people in the audience, they were really fantastic for us to reflect back on the process as well and we often - here at the Library you’re so busy running from one project to another you don’t often get that chance to reflect. So I think that was really, really fantastic as well.
So look, if the work’s excited you the full report is on the ANU website and on ours but I think we’ll say that you will take these print copies from our hands through our dead cold hands but we want to hold onto them. It’s really worth doing so. But I think if you're really excited by this kind of work and you know somebody who would be I think you have got just the thing for them, haven’t you, Alex?
A: I do have just the thing for them. I wanted to mention that our applications are open for 2023 Masters of Applied Cybernetics. This is a one-year course in cybernetics and applications are open for Friday – until Friday, the 9th of September. We are very interested in people that come from very diverse places in terms of the knowledge that they hold and the practices that they have so you do not have to have a background in any particular discipline or practice area to apply. We have everyone from – I see a couple of my PhD students in the audience, architect and economist as well as we have computer sciences, engineers, dramaturgs as I mentioned earlier with Charlotte, chemists, IT people, people that have backgrounds in professional writing. It is an opportunity to learn how to think in a cybernetic fashion and we’re very interested in as I said diverse backgrounds.
If you don’t already have an undergraduate degree that might be okay as long as you have really detailed and long professional experience. We are more interested in creative thinkers and critical thinkers than necessarily what your undergrad degree was in or was not in so we are an open and exciting group of people and we hope that if you’re interested please check out our website.
M: I think that speaks perfectly to I guess our experience of we set out of course to look at how a national library like ours might integrate tools and processes enabled by artificial intelligence machine-learning technology capabilities. Integrating those into our core work processes to bring further to life our enabling legislation, our mandate which is to collect, to preserve and to make accessible library materials. I think you’ve seen from I guess the journey that we had from our little experiments, my risk-based thinking to how this process has opened up our thinking, I think you can see the kind of journey somebody might take on if they were to do the masters program.
Now we do have some time for questions this evening. Now the presentation’s being livestreamed and although we might be able to hear your questions without a microphone those at home can’t so if you have a question and don’t be shy even though it’s a big arena, please put up your hand and we will bring a mic to you and we’ll see how we go with answering any of your questions so okay, who’s going to be the brave person putting their hand up first? Okay, I can guarantee that this person will always ask a question at any public forum.
A: Okay. I feel like you might know this person.
M: I do know this person quite well.
Au: Thanks very much. Yes, I’m the reserve question-asker. I wanted to ask you to explore a little bit more about the pairings, the quality, quantity and so on ‘cause I'm interested in how that’s used analytically and my mind went immediately to scenario planning sort of modes which use similar sort of opposite mapped on a matrix over time and what might happen to a system changing over time. So I was just wanting you to explore that a bit more.
A: Certainly. I talked about general dynamics that characterise the Library as a system and there were five of them so kind of quality versus quantity, agents versus audience so what kind of roles are played among people that are associated with the Library, nodes and networks that the Library is kind of a centre in itself but also part of a larger system, formats and fluidity, that an object in the collection may start off as analogue and switch to digital or start off digital and then take an analogue or switch multiple times over the course of its life, and then serendipity and indexing.
Our point here isn’t rather than things may go into a scenario planning matrix where one side takes over more than the other but that what is useful or what characterises the Library as a system is the ongoing tensions between those two forces. And that if you got rid of those tensions and if you just went too much towards quality and not enough towards quantity or too much towards what the Library needs to do as part of a global network versus its mandate to the Australian people it kind of falls apart. So we use it analytically as one lens to think about when we look at a particular technology capability and what the effects of introducing that capability into the system of the Library as kind of the foundational things that you have to keep in mind that are not necessarily going to change that much but that you need to account for. Whereas some of the dynamics or some of the things that come up with a particular technology capability may be specific to that.
M: I think also, Alex, is that also in that kind of sense of tensions I suppose is what the drivers are for making decisions and you’ve talked about audience. Well a driver for us is always going to be to improve access, at the moment is primarily for those used in discovery systems, machine-actionable collections, it’s kind of a different group. But when my CIO said to me a year ago okay, in this space, Marie-Louise, are you most interested, are you putting higher priority on improving access for end-users or driving down the cost of processing collections? I actually said the latter because that’s a driver at the moment but you can also see that if – even if that’s a major driver you’re going to – you inevitably are going to have things balancing out.
So I think at any point in time an institution’s drivers for why they might do this and how they might do this are also going to change, there’ll be some tension and it’s not either or but he needed an answer and I had to give him one so – and it’s a big issue with a collection this big and reducing resources. It’s obviously an issue for us. Other questions? Yes, we’ve got a question here. Thank you. Sorry, I’m pointing at you just so we can get the mic to you. Great.
Au: Hi. I’m just wondering so you developed this big system, how would you ensure that it will actually mesh with the other systems that may be developing elsewhere in the world?
M: Okay well I’d say first I think we’re not talking about developing a system here, we’re thinking about the Library and its work as a system that you can think of in that way. But your second part of your question is really important and it comes back to that thing that librarians really care about too which is standards so to try and make sure that you’re not doing things that can’t interact elsewhere. I would say at the moment the technologies that we are using right now and that we’re likely to use, they’re all produced by third parties. They’re readily available, you don’t build an OCR, there already is one. You wouldn’t build your own handwriting tech, you’d always reuse a tool that is there. Certainly in the library world nationally and internationally there’s a lot of work that goes on in spaces like this from – it used to be about descriptive standards, now we put a lot of effort into preservation standards for the web, for example.
So I think it’s always keeping plugged into national and international movements but a lot of it is around not building when you can use something that’s already available but also being aware of what – no tool, no piece of software, no algorithm that anybody produces is neutral so you’d need to be aware of what the kind of shortfalls are there so I think that’s important. We don’t have to do it all by ourselves either. An example here is for years and years libraries have been talking about linked data, about putting out our metadata in ways that machines just on the open web can easily use. Look, we haven’t had to do that as an institution because we contribute all of our metadata to OCLC, a big international organisation that pulls together 60,000 catalogues from around the world and they’ve done that work on linked data. So I think it’s choosing where you put your effort and not trying to do everything at once too. I hope that helps answer your question.
A: I just wanted to add too that I'm always impressed with the work that is done at the National Library of Australia and that you are world-leading, particularly with OCR and the Trove and the influence that that solution, that system – it’s not really a solution, it’s a set of systems in itself - has for making – for preserving and making content accessible. Truly extraordinary compared to other places in the world where I’ve lived.
M: Yes who like to copy what we do and we gladly say go ahead to make it better for users everywhere so we do need to invest more because yes, we were world-leading 10 years ago but unless you can keep investing you’re not going to be there and that’s another challenge for us too so – some other questions. Yes, we’ve got one up the back here, thank you.
Au: This is a question I think probably to Alex but probably it reflects my ignorance but if I understand cybernetics it’s a sort of control system which deals with changes in what occurs within the environment that it’s interacting with at the system, is it? But what’s been fascinating for me through my boomer lifetime is to watch how AI has changed the way that we as humans look at what’s around us so that I’ve been limited by my five senses and limited intellect but then you apply neural networks and stuff like that and they come up with a completely different way of looking at things and connections which are absolutely fascinating.
The question’s in reverse, are we the right people to be asking the question where the future is or should we be using AI techniques and neural networks to ask these questions rather than asking the individuals who work in the institutions?
A: That is a great question. I would say that my view of cybernetics is less about control theory and more about thinking about how to steer a system towards change but without ever having the idea that you would ever have complete control. And that the types of systems that we’re interested in are open systems and adaptive system so you’re never really going to see the whole thing or control the whole thing.
The question that you have about what types of agents are capable of commenting on potential futures and where we may go with them, I think you're right that there are things that we can take from largescale datasets that are interpreted through various neural networks and things like that but that is not – that is always partial in the same way that our knowledge is partial. It’s not that we humans like don’t know enough and we just give it to the machines, the machines will have absolute knowledge and be correct but that there are multiple ways of addressing what the future might be and playing those off against one another is certainly a more useful way I think of determining where we may be headed than just taking one viewpoint.
M: Yeah, I think I’d also say I definitely don’t think that it should just – the people in this library who are thinking about this, that’s why we want to take it out on the road. I also think if libraries, archives, museums, galleries are not thinking about these things, if we cede the field solely to commercial interests there’d be something really, really missing there too. We don’t have a profit motive, we have a public good motive and we also enjoy a lot of trust from the community. Looking recently of course about some of the large tech companies sort of lost quite a lot of trust because they weren’t transparent about what they were doing.
That’s very much in my mind, trust is really important for us so taking our community sort of on the journey with us to say how are we using these technologies? What might be the results? What are we thinking about in the future is a big part of it. So – as is working nationally with other institutions like ours and internationally. We can only be part of it but if we’re not there there’ll be something really missing from the picture.
I think we have time maybe for one more question if there’s another question from the floor. Okay well we actually have run slightly over time so I was just watching that so only two minutes but that’s okay. So – and of course you’re very welcome to come and have a chat with us afterwards, we’ll probably put our masks back on for that. So would you please first join me in thanking Alex for a really interesting discussion so thank you, Alex.
A: Thank you.
M: Encouraging you to think about the masters and I’ll also just again give you a reminder that the second event in our cybernetic thinking for a new world series will be 6:00 here, Thursday and we hope you can join us as Alex’s colleague, Kathy Reid, examines that speech recognition technology that we’re all using in our everyday lives through a cybernetic lens so the details for that are on our website but I hope you’ll either come along or watch it from home as well. So thanks very much for coming along and again thanks very much, it was such an exciting piece of work and yeah, you’re not getting the beautiful print copies out of our hands so thank you. Okay thanks, Alex.
A: Thank you.
M: Thank you.
End of recording
The library – as an idea, a place, and a cybernetic system – has always been an investment in, and a vision about, the future. Recently the National Library of Australia and the ANU School of Cybernetics conducted a thought experiment: If library indexing has informed the structure of search on the internet, how might the library inform new ways of connecting these large caches of information to generate meaning? In a discussion facilitated by the Director-General of the National Library of Australia, Dr Marie-Louise Ayres FAHA, Professor Alexandra Zafiroglu from the ANU School of Cybernetics examines the processes, structures, networks, and relationships that make a library and imagines how we might build AI-enabled systems to help generate meaning out of vast datasets.