Artificial intelligence (AI) systematically tops popular lists of the most important emerging technologies. With a mix of fear and excitement, commentators seem to agree that it will shape the future, although not always on how.
To borrow a definition, AI is ‘programming computers to do things, which, if done by humans, would be said to require “intelligence’ (MIT Press Essential Knowledge Book on Machine Learning, cited by Chris Bourg).
To do this, it draws on analysis of (often large amounts of) data in order to find connections and draw conclusions. In this way , it can search for information, detect medical conditions, interpret from one language to another, or write books, for example.
Just as previous information-related technological revolutions have marked libraries – printing, electricity, the Internet – it’s worth asking how AI might do so. This blog runs through a number of the dimensions already highlighted by others.
It covers both those that directly impact libraries, and those affecting the broader information – and social – environment in which libraries work. Finally, it asks what scope there is for libraries to bring their resources – and values – to bear to shape this new world.
Robot Librarians, Robot Books?
Perhaps the most regularly voiced concern is that AI will accelerate the evolution in the way people look for information, at the expense of librarians.
Search engines have already gone a long way to replicating the traditional role of libraries in helping find basic information – capitals, dates etc. As AI becomes smarter, the argument goes, it becomes better at some of the more nuanced, smarter searching that librarians have long performed for users.
It is true that computers can ‘read’ literature far faster and more comprehensively than any individual can. Nonetheless, it remains the case that a search is only as good as the search terms put in. A good librarian, through working with a user, can provide a much better tailored service, potentially using up time freed up by using AI.
Moreover, of course, it opens up some truly exciting possibilities to do more with works already in collections (as long as they are digitised, open access, and ideally have the right metadata to be used across institutions). In effect, it can make libraries, and librarians, more valuable, rather than less, as Catherine Nicole Coleman underlines.
A second, and quite different issue, relates to the materials libraries stock. With AI already capable of writing novels, newspaper articles and research papers, there is a question as to the copyright status of these works.
As the monkey selfie case has underlined (more or less, and at least in the case of US law), it appears for now that only human beings can claim copyright. Potentially, the programmer behind the AI could claim an interest, but as highlighted above, may not be able to explain the process that led to the work in question.
Legal clarity on this point will become more and more pressing as time goes on, and AI-generated works take a larger and larger part in the record of works produced.
AI and the Library Environment
Clearly most discussion of AI doesn’t look at libraries, but on the impact on society as a whole. However, as institutions implanted within societies – and with a mission to serve them – libraries will need to take account of, and even deal with, the consequences.
A first issue is what AI will do to our lives and jobs. On the positive side, it does seem likely to free up time and effort for other activities. Given the impact of ‘attention scarcity’ on reading habits, this may be a positive.
Nonetheless, it also seems set to destroy many jobs, and create further divides in the labour market between high- and low-skilled jobs, a point highlighted by Ben Johnson. While it is comforting to think that things like creativity or entrepreneurship are still only possible for humans, there will still need to be support for individuals in realising this.
A second question concerns information itself. As highlighted in the IFLA FAIFE blog on data ethics, there is uncertainty as to how algorithms and artificial intelligence comes up with its answers. When searching for, or making use of, information, it may not be possible to explain how this happens.
While there never was a perfect search, it was at least possible to analyse the process followed. The risk now is that there is a black box, with little possibility to explain the results.
The one thing we do know is that there is a risk of bias and discrimination. Given that AI feeds on data about the present, it is liable simply to repeat this present into the future. Moreover, commercially-run AI brings with it risks of prioritisation (in the case of search in particular), or of course use of personal data (in general), with all the implications this brings for privacy and Intellectual Freedom.
In short, AI seems to be making two key functions of libraries in their work with communities – support for personal development, and information literacy – as important as ever, if not necessarily easier.
Yet libraries are not powerless. Our community’s skills and values, as well as our institutions’ collections, have the potential to impact the development of AI. And by getting involved, we also have a chance to underline the importance of libraries to one of the most significant technological developments of today.
The most immediate contribution comes through library collections. Artificial intelligence works by looking at existing materials and drawing new connections and conclusions. Libraries, collectively, contain the richest imaginable resource for the development of AI. Indeed, the GoogleBooks project, for all its critics, made a major contribution to getting AI to where it is today.
The implication of this is, of course, that digitisation projects (at a quality that allows for machine reading), open access, and linked data are more important than ever. So too are efforts to ensure that copyright law does not drag down efforts to advance AI.
Linked to the use of library collections should be the application of library values. As highlighted above, AI (and big data in general) comes with risks of bias and discrimination.
By opening up whole library collections (in particular from a wider variety of sources), libraries can help provide greater balance in the material with which AI works. Moreover, by cooperating with programmers, there is an opportunity also to identify and combat biases that could lead to unbalanced results. Chris Bourg underlines this opportunity – and the challenges it implies, in her blog.
It remains relatively early days for artificial intelligence. But there are already a number of areas where libraries will need to be active. It is worth exploring further the library angle on these, in order not just to be ready for, but to shape the changes AI will bring.
Bourg, Chris (2017), What happens to libraries and librarians when machines can read all the books, blog, 16 March 2017
Coleman, Catherine Nicole (2017), Artificial intelligence and the library of the future, revisited, blog, 3 November 2017
International Telecommunications Union, AI for Good
Jacknis, Norman (2017), The AI-Enhanced Library, blog, 21 June 2017
Johnson, Ben (2018), Libraries in the Age of Artificial Intelligence, InfoToday, January/February
Mulgan, Geoff (2018), AI for Good: Is It for Real?, nesta blog, 23 July 2018
Snow, Jackie (2017), Bias already exists in search engine results and it’s only going to get worse, MIT Technology Review, 26 February 2018
Tay, Aaron (2017), How libraries might change when AI, machine learning, open data, block chain & other technologies are the norm, blog, 9 April 2017