Day 1 – Tuesday 12 February 2019
Genevieve Bell – Wonder in the age of AI: art, creativity and possibility
SIRAC, the first computer stored memory, began its life at Sydney, but then most of its life at Melbourne University. It taught an entire generation about computers and it was used to process data about weather, Melbourne Cup odds, mortgage rates, calculating odds and much more for many decades. It was also taught to sing – because it made noise – and it did so for five years. It was also the first computer to play games.
The computer was not designed for music, but people were creative and made it happen. And when it was turned off, there was grieving, because it gave so much to the work and the enjoyment of those who worked with it for so long.
This definition was created in 1956 and was referring to machines. Up until then, computer meant women who did maths. However, this definition was based on humans being based on intelligence (referenced by BF Skinner) and was financed by the US Government. The government wanted a translation machine. It failed because there was no sense making, just pure translation, which led to interesting translations.
Return to Square was the first art work created using Fortran and ended up being part of an exhibit comprising totally of poetry, art and music created by computers in the US in 1967. It was across the road from another exhibit which demonstrated the mouse and the internet, so has been widely forgotten.
“Tame the computers appealing transcendent charm”. How do we use the data we have to ask uncomfortable questions that haven’t been asked before and if the results are also uncomfortable, what do you do with them?
Algorithms are just automated processes. But who decides what that process is, the steps in the process and can they be tweaked? The challenge may be to add surprise – create an algorithm for discovery, not just familiarity. Eg. want coffee, but map shows you public art first. What might it mean to also show discomfort – it may be different, but it may be a sign and it is therefore interesting – not noise (which is what most algorithms consider it to be). NOTE from me: Google Search will deliver what it thinks we want…….
AI is learning by increased data, learning from communication with other AIs, looking for patterns. They are creating their own ways of being intelligent, not our ways.
A lot of our world building is based on ideals, which were often created in the 1800s eg. ideal height range, ideal photo quality for white people etc.
But moving forward, questions need to be asked and then build the thing that reflects those questions and the answers – even if they are uncomfortable.
AIs sit everywhere already. They mostly don’t talk to each other.
Why are you collecting data? Why is that AI being enabled and what is the problem it is solving? Many of those are limited, but the exploration of the potential of AI is starting to combine data from many AIs, for new experiences and new data.
Shared her experience of the Tempest and Stratford on the 400th anniversary, where people, flew and floated and CGI was utilised on the fly and live and so much more, which changed the experience. But it used technology that was developed NOT for this purpose.
As we develop technology, the questions are what will it do, not just for purpose but for what it will do with us, through us and for us.
Contemporary collecting: collecting Instagram for local studies
Searched out the list of hashtags that were already being used by people taking photos and posting them on Instagram.
Not long after they stared, a flooding event added to their hashtags and their collection.
Harvesting now results in local history for reference now and as history in future.
The website was to be representative of their area, not just of photos. People, every day, life, culture, business, events and more.
Used BaseCamp for project tracking, Google Drive for storage and communication tools.
Project is called LENTIL and is open source created
Tags and tag sets defined by your local area. Tag sets can be turned on and off as required for Harvesting. Eg. Lunar New Year. Helps you keep control over what you are looking for.
Images are harvested using the tags list and compared to those on Instagram and then store the images and videos on the website. They are also backed up to Amazon S3. The images are moderated before they go live, as hashtags are not always reflective of content.
Hashtags can be fraught – interacting with user data can be unpredictable, can’t make assumptions, best to react to the changing needs of the app as you go – in order to keep it user friendly.
Limitations – rely on Instagram not changing the way that other apps talk to it and dealing with large datasets is difficult.
Future possibilities: other social media platforms to be harvested. Not just images, but tweets, etc.
Reproduced. Atomized & Deconstructed: the future of scholarly communication – Daniel Hook
Reproducibility as a target, is going to need to be so, automatically. Elegance is a key thing to strive for.
Science has popularized its content and brought everything to the common level. Which means the other important content – the things not media grabbing, is left unknown. And the lay people no longer see a need for scientists as everything can be understood by anyone.
Datasets under research is more important as it might be the foundation for many other lines of research.
It’s not just about the data, but the context of the data – although this is discussed in the context of scientific research – to ensure that it makes the research reproducible, it is applicable to all data collection.
Tinker time: developing digital literacies with the growth mindset
Used two projects to help develop a growth mindset. Lifelong learning requires an understanding of how you learn and putting it into context. It is also reliant on the culture of your workplace.
Growth mindset is permission to learn and do things, to have negative feelings, but then get on with it. Growth mindset was taken into Tinker time which was about staff capacity building. They understood that the skills they expected the students to have, should also be present in staff.
Learning was about being a little uncomfortable, would stretch them and encourage involvement. The program ran over a semester and involved staff voted skills, monthly workshops, lesson plans, and more. Staff chose their own projects to do as part of this process.
Language focused around learning not results. The power of YET. I’m not good at that YET. Being a semester long, people left things to the very end. Needed more scheduling to help them achieve the goals.
Tinker Kits were a collection of electronic kits for digital literacy, including Arduino, Little Bits and robots. The kits were designed to be entry level but could be scaffolded up. Kits were accompanied by learning materials. All kits had to be applicable to multiple faculties and were available to all students, even those in different faculties.
Kits were developed with a collaborative mindset, with faculties, students, cataloguers and more. Feedback was enthusiastically taken on board and everyone was encouraged to explore, make mistakes and refine the processes and kits to ensure the right tools were made available.
Did have unsuccessful sessions, but they were a good foundation for programs improvements. Started with education and engineering faculties and are now moving to business, design and dance faculties, who are using technology is new and exciting ways.
It was a pilot which everyone accepted but has quietly transformed and now it is a part of the program and library work.
“I didn’t get good, but I did get better”. This program is about improving, not about perfection.
A Culture of Learning is a great start, but you also need to start small and work up, given permission to make mistakes, to encourage always and lead by example. Learning culture is caught, not taught.
- Claim the space – do it, before someone else does
- Reputation building – build the reputation and then show it off and watch it develop and grow. And they now get invited to seminars, working groups and more.
- Shared language – the whole department has this around a growth mindset and digital helping to define library digital literacy policy.
Beyond Time and Space: using AI to solve client service challenges now and into the future
One of our biggest challenges is serving 24/7. Why AI? 1/3 of queries occur after hours. AI can be 24/7, is quality control, sustainable and flexible.
They built their first viable Chat Bot with the student team, within 10 weeks. If the Bot couldn’t answer the question, it would give the student the option to refer to a librarian and forward on details if they agreed to do so.
Version 1 had 77% accuracy, whereas Siri has 78%. It is multiple platform so could be used on any device and towards the end of the project, was considered a prototype, with the view to expansion.
They couldn’t go live with it, so it has been tested and demonstrated, but they don’t have the technical expertise and support to make it go live.
The team were unable to get grant funding for the project, so it was the students as part of their course, that got the project across the line. They also had a lot of difficult getting their chat session data from Springshare into a useable format, so they ended up getting staff to add more content to the knowledge base that the chat bot could use. There is no AI strategy at UOW, so they had to tread carefully and had to bypass some hurdles.
Even though there have been restrictions, they will still push ahead with the Chat Bot, but they know it is possible and know it works, plus they have gained invaluable information and have shown the university what is possible with AI – to the point that it is now on the university agenda.
Innovation is tough – it takes time, tenacity and courage.
Relationality and the unrealized potential of digital collections: Mike Jones
Australian Museum – Westpac Long Gallery links relevant content through cascading slides and touch points, without the need for searching. The relational museum brings multiplexity and more in a growing number of museums.
These connections however are not widely seen in collection management and library catalogues. Francis Bacon said that knowledge comes from interconnected systems, including libraries, zoos, laboratories and more.
Often catalogues are great descriptions and give wonderful information, but don’t give you other paths to explore.
But the basis for an automated question answering system can be traced back decade. Thomas Marill – 1963 Libraries Question Answering Systems. Libraries of the future – JCR Licklider in 1965 – although he severely underestimated the amount of data in libraries in 2000.
A flip happened between 2000 and 2001, where the number of visitors became more and more digital and less physical. The pattern continues.
Wikipedia is not perfect but is a good representation of connections and new pathways through their mandatory citations.
Besides the technical barriers, there are also process based barriers. How do we cross connect between institutions for example? There are also a wide range of standards being used, only some of which have outward facing relational capabilities.
Aggregation is one way that some institutions deal with it – one search function over the top of all collections. Eg. Smithsonian has done that. However relational connections may still not exist, because they used different catalogue standards, different keywords and more.
If humanity is as deeply intertwined as is continually demonstrated, then why isn’t out data?
Knowledge and collections should be like rugs – where we weave things together. We need to think about this when creating and building our online collections and spaces.
Relational systems are changing all the time – like rivers, the flow changes as does the banks of the river. We have been so focused on the banks, we lost sight of what is happening with the river.
Other questions to consider – automated maintenance or even user communities to help with this.