 It's very exciting for me to be on the stage to talk a little bit about the history, but also In the audience and watch Josh give this delivery It's hard to tell but Josh is only been in the position for about seven months and while he was triple AF aware He was not an active participant So I think it would for those of you who have been in the community and for those of you who would like to come Into the community just a note of recognition for Josh is our still relatively recent, but it's hard to show managing director so I am the associate university librarian and the chief technology strategist from Stanford University and The founder of triple AF so I have a lot of history and most of that comes from the pain of an individual institution That needed interoperable solutions and then the thrill of working with a tremendous number of Technically savvy and user savvy individuals and institutions to frame this Environment so what I'm going to do is give a little bit of the background on some of the great developments that Josh just showed But also some of the pain and some of the promise that we still see at Stanford as part of this so From Stanford's perspective triple AF all started with Parker dot Stanford dot edu and if you're not familiar with this it's a Digital library site that is beautiful and was arguably best in class before we replaced it So now what's even better and then as many of us who have built these sites? No, it was it was hard to build the image delivery in particular was particularly challenging. This was around 2006 2007 And it turns out that we produced this wonderful site with all these technical challenges And it was such a success that its users wanted more they wanted to be able to export the images They wanted to be able to annotate the images. They wanted to be able to run custom transcription software They wanted to be able to compare the images with those of other sites and What we realized Is or what we what we put ourselves in the position is thinking about the wwmd or what would a medievalist do? So we stopped and we thought and we said well who else has the same problems that we have What collectively are we trying to do? What are the core scholarly use cases that we were trying to address with parker dot Stanford edu? And what could we do with others who had similar content, but the same overall sets of problems? And it turns out there were a number of fellow travelers at the same time and many of these came out of the rich digital humanities tradition Many of these but not all of them were funded with the help of the Mellon Foundation Many of them were leading projects from National libraries and universities so in about 2010 we started to have a series of discussions with not just representatives from parker, but the romanda la rose project from Johns Hopkins University diam Which was an independent project at the time, but as images of medieval manuscripts Musical manuscripts the ecotasies project from Switzerland Gallica from the bibliotech national and Oxford and we got together because each of them had built their own version of parker and had tried to solve many of the same problems and We realized that effectively we had built a set of silos we were all doing the same thing, but we had all done it in very different ways and Not by design, but just by the force of independent operation We had built these separate stacks or the silos of both technology and content and while we wanted to cooperate or co-op Essentially what we were doing is standing up different technology stacks and content silos next to each other and Forcing individual users to navigate the across the top as shown by that catwalk so at one of these Dinners we said hey, there are a lot of promises around doing So these meetings we had several people from those different projects coming to Stanford over the course of About a year a year and a half and one night at dinner We sat down and we said look medieval manuscripts are very exciting and they're unquestionably scholarly They're rich and they're beautiful But we all have way more content than just medieval manuscripts and maybe we're thinking about this at too small a scale Can we expand out and let's think about all of the image-based content that we have not only to our research institutions But worldwide so this is we went to a Cuban restaurant. This is the apocryphal story was true We went to a Cuban restaurant. It's not a napkin It was actually paper tablecloth and we started doodling on the tablecloth And tried to figure out what would it look like if we went to global scale and about four months later Triple AF was born and so the idea as Josh described and is quite right is can we come up with a global framework? For sharing all image-based resources in a way that's mutually useful or some Pataco Reducing the friction of sharing materials in a framework that includes not just apis, but software content and community Because again while manuscripts are important part of our collections. It's only part of our collections at the time We had some very specific things that we were trying to do One is just deep zoom was technically hard It's no longer as technically hard in part because of the advances that triple AF had made This is still I think the best example This is from Princeton University libraries and they were an early implementer of open-seed Reagan, which has a lot of overlap We had very large images that we were struggling to deliver over the web This is a gigapixel image, but this is now standard technology within our repository and for many other sites We wanted to compare images and compare images both from the same site and from different sites This is a great example from the Yale Center for British Art and as part of their provenance and conservation history looking at the origin of different types of portraiture Stanford's also a relatively young institution and what we realized is as many medieval manuscripts as we had being founded in 1891 there were others that had many many more and This is an early example of some of the work that my colleague Ben Albrighton who's going to speak later today about manuscripts are But we basically found ourselves undoing the work that others had done a century earlier So Otto Ege is a biblioclast Which is a person who purposefully destroys or tears apart books Why did he do that? He was a manuscript dealer and he found he could make more money by disbanding manuscripts and selling individual pieces Which is great for near-term profit, but harder for long-term scholarship Ben through his research had found that while we held many leaves of a particular Manuscript at Stanford He found through tracing the actual history and descriptive metadata in the community that other institutions in South Southern California, California, North Carolina and Mississippi held other pieces and through the power of AAAF was able to unite those things Virtually together while they were physically dispersed So Josh described this as well as this this notion of the silos and the duplication and that API's enable reuse So taking another look at this from a more a practical or grounded issue is we found that we had use cases where We had a single server and a single application and we wanted to be able to deliver that content So we could have drawn a straight line or we could go through triple IF as a set of API's the advantage of going through a triple AF as a state of set of API's as is in that last example we could Unite content from two different servers and present that in the same application for conjoined study Or we could have a single server with Multiple different applications sitting on top So in this case we might want deep image delivery But also the ability to annotate images in a different application that we hadn't developed ourselves Or third we might want to expose some of our content to external web services for the automated Description or writing of bounding boxes to facilitate transcription and What we had found effectively is that triple IF effectively separates out and Augments separates out the concerns of research institutions with repositories of scholars and of software developers who want to build a set of tools Research institutions hold the materials, but don't want to and shouldn't have to develop all the software That to expose their digital materials Software developers are really good at developing tools But shouldn't be forced to write tools to a specific repository or a specific content type and scholars Really are less concerned about where the content is and what the software is They want the best tools and the best content to pursue pursue their research at any given time and I think the fundamental advantage of triple IF and where I keep seeing the value is through this interoperability And it's interoperability across many different vectors So technology is the most obvious one But it's also the ability to remix and reuse content and tools across different contexts across different geographies Across physical space across time and across different types of users So That's why triple IF has been very exciting and meeting the distinct needs of an institution like Stanford in the past One of the exciting things is what it may offer us going forward and one of the things that we're increasingly interested in is Leveraging triple IF with machine learning and artificial intelligence So in a crude rendering What if instead of writing content from your repository exposing content of your repository to applications? What if you could open it up to the wealth the growing wealth of artificial intelligence services that are out on the open web? Both as to consume your content, but also to enrich your content And we think triple IF has a privileged position in this ecosystem Because as a way to expose your content through a standard API for large-scale consumption But also as a way for these AI enriching services to express the enrichments back in a standard model or a standard language That we can work into our environments. So basically annotations that we can reintegrate into our services So some early work that we've done at Stanford is focused on Extracting road networks and looking at the growth of road networks over time So we basically came up with a computer computer vision pipeline to identify road networks Over a series of years with road atlases for the same road networks in the United States And there's a whole separate presentation that you can look on that The exciting thing is is we found this core technology was not only useful for looking at road networks But for feature extraction from other types of items So here is an example of looking at a net an initial from a medieval manuscript with a folio. So identifying a The key characteristics of that basically doing some on-the-fly parameter setting to let the machine learning know How similar it should look and on the fly. This is the same algorithm that was used for the Extracting and finding similar road networks to find similar medieval initials This doesn't need to just be restricted to cultural heritage and memory institution material These are Images of blood cells. Can I count the number of blood cells or find similar blood cells from stem imagery? And then perhaps More whimsically, but I think that it demonstrates the power where is Waldo most of you are probably familiar It turns out by training the machine with only five Waldo's It's possible to run through and find similar Waldo's So think about the hundred million or billions or I'm gonna go one up tens of billions of materials that are available via triple If we can't actually count so that we know it's over a billion Think about all of the materials where we might want to find similar items across our combined Corpora, this is very exciting to find all of the Waldo's So in some Triple IF from our perspective at Stanford and I think in a historical perspective is important for two things One is interoperability reduces the friction of delivering and using in digital content on the web So it's better faster deeper and cheaper delivery of your assets You can plug and play different types of software and you can publish content once and reuse it in ways that you had never imagined At the same time. It's not just viewing images. It's really interacting with them. So whether that's Deep zoom panning and manipulating whether it's mixing up materials from different collections Whether it's the ability to site to share or to annotate or it's the ability for some of these Up-and-coming machine learning and artificial intelligence algorithms It really offers promise that would be far beyond the capabilities of any single institution to imagine when we started this By convening a set of meetings with people of shared concerns. No one at the room at that time could have imagined where triple IF would go today So here are some here are some I think pithy quotes There are a little bit old but still sum up where triple IF is so Francesca Frey from Harvard Said in a meeting once it's like going from the 18th century to the 21st century in a single click Glenn Robeson who's now the triple AF technology coordinator But was at the National Library of Wales when he said this is triple AF is good for the National Library of Wales but it's also good for sharing content one of the mandates for for that institution and Richard Higgins from Durham University said triple AF doesn't just keep you out of silos It keeps you out of dead ends by making sure that you're not developing a set of technologies that only you are going to use So thank you. That's just a little bit of context And we'll get into some showcases