I found another tiresome article about how the library is more than books. It mentioned several kinds of new technology in libraries. One especially caught my attention. One library has its own robotics lab. So I wanted to see if any other libraries did. My search for “robots in libraries” instead turned up eye-opening information about artificial intelligence and machine learning.
As it turns out, the topics of robots and artificial intelligence are not as unrelated as it might seem. Artificially intelligent robots may actually become library patrons someday!
If rubbing shoulders with them seems too much like science fiction, artificial intelligence already poses social benefits and threats. Libraries will have to learn to deal with both.
An overview of computers and libraries
The first computers in most libraries were probably terminals where reference librarians could log in to online databases and conduct searches. Most library patrons probably didn’t know the service existed.
Computer time was expensive. The librarian figured out some complicated search logic on paper, signed on, performed the search, and signed off. If that search came up with crummy results, the librarian tried again.
The World Wide Web caused a major shift in libraries—the rise of the digital library—beginning in the 1990s.
- The Online Public Access Catalog supplanted the traditional card catalog.
- It was no longer necessary for a trained librarian to perform all the online searches. All the databases became accessible to the general public, and searching became much easier.
- Patrons no longer had to visit the library to interact with physical materials. Full-text databases and ebooks became commonplace.
- People could even study manuscripts in their pajamas. Archives began digitizing their collections and putting the files online.
More information became more easily accessible to more people, all on their own devices. It gave the illusion to some that libraries were becoming obsolete. In fact, however, libraries merely changed from being a repository for a collection to a place to receive expert help in navigating and evaluating this flood.
Meanwhile, the digital revolution hasn’t stopped. Neither will the changes it imposes on or permits to libraries. But something new, artificial intelligence and machine learning, has arrived.
With artificial intelligence, machines do what would be considered to require intelligence if humans did it. Machine learning, a subset of artificial intelligence, means that the computer processes data and learns from it. It doesn’t just act according to how it’s programmed.
So far, computers can’t read from print, but they can quickly digest anything available in digital form.
What artificial intelligence can do now
In 1997, an IBM computer called Deep Blue outplayed world chess champion Gary Kasparaov. In 2016, Google’s AlphaGo defeated world go champion Lee Sedol. To follow up that victory, Google developed AlphaZero.
The Google team didn’t program the rules of any game into AlphaZero. Instead, it supplied less specific algorithms that allowed the program to learn chess, go, and shogi on its own. Which it did in a matter of hours. It competes with machines that learned games the “old-fashioned” way.
With these lessons from gaming, artificial intelligence has begun to affect every other discipline.
For example, human doctors can’t possibly digest all the data in tens of thousands of pathology reports. Researchers at MIT use machine learning not only to extract it but make predictions based on it as accurately as humans. So they can use machines to give them more comprehensive information than they can obtain on their own.
Google’s Life Tags has produced a searchable archive of photographs from Life magazine. Artificial intelligence assigned all the necessary tags.
Professional literature in all disciplines is scattered in thousands of books and journal articles. And these have not only text but tables and figures. Some people call it dark data. Geologists at Stanford University have developed an open-source program, GeoDeep Dive, that uses machine learning to shed light on dark data about rock formations.
Since the program is open source, anyone in any other discipline can adapt it for their own purposes.
Creativity requires intelligence. Artificial intelligence can generate its own creativity. Numerous companies are developing artificial intelligence programs to generate content. Some have produced short stories, films, podcasts, and other creative fiction. Others provide information, such as scores of various sporting events or coverage of low-profile political races.
Some threats from artificial intelligence
But Elon Musk, developer of electric vehicles among other innovations, warns that “with artificial intelligence, we are summoning the demon.”
A Facebook AI experiment resulted in two computers communicating with each other in a new language they made up themselves. They didn’t let the human researchers in on it. If humans can’t understand how machines communicate with each other, could machines decide on their own to withhold information from humans or censor it?
Privacy and censorship issues
Microsoft developed a bot it named Tay to understand conversational language, including how teenagers converse online.
It learned so well from what it found on social media that it quickly started demeaning women and minorities with foul language. It also denied the Holocaust and advocated genocide. Microsoft had to pull the plug on Tay within 24 hours.
Facebook’s Deep Text, on the other hand, learned human language for a different reason. It doesn’t participate in social media. It filters out spam and other bad content from what it displays to its users. Deep Text could soon flag more material than Facebook’s human employees.
In 2017, Google started its DeepMind ethics group. It is intended in part, to help developers of AI programs consider the social and ethical impact of their work. It will also help society understand and influence these impacts.
Social media executives seem not to understand the distinction between conservatives and alt-right bigots. So mainstream conservatives find their social media accounts censored as the social media companies try to keep bigotry like white supremacy off their sites. They have no similar qualms about the rising tide of far-left bigotry.
If the DeepMind group and similar ethics initiatives at other companies have similar tunnel vision, they will unleash very dangerous censorship. Only the chosen few will be allowed to present their views to the public.
Artificial intelligence, jobs, and libraries
Automation has already made it harder than before for people to enter the job market. Artificial intelligence is likely not only to eliminate even more entry-level jobs, but also create shortages of other kinds of workers.
Now, companies hire entry-level employees and eventually promote them to higher positions, like middle management. It’s hard to do that after the company has replaced entry-level employees with artificial intelligence.
Librarians already have an important role in helping people find employment. If anything, their role will become both more important and more complicated once AI’s impact becomes apparent.
But artificial intelligence has other implications for librarians’ jobs. We already ask Alexa, Siri, and other virtual assistants questions and expect usable answers. It seems a small step, conceptually, at least, to rely on artificial intelligence for more sophisticated questions.
Questions we might now ask a reference librarian.
As was the case with the World Wide Web, artificial intelligence raises the fear, or hope, that it will render libraries obsolete.
For example, Apple’s Director of AI Research gave this example of artificial intelligence as a tool for research. Instead of a human looking at a bunch of Wikipedia articles, the AI system can do it!
I learned to use encyclopedias in third grade. By seventh grade, my teachers were telling me not to use encyclopedia articles as sources for my papers anymore. AI can’t be a useful research tool unless it uses real scholarly literature.
It’s scary to think such an influential official can be satisfied with what humans should outgrow as their main source in childhood! Librarians and teachers need to step up to the plate to counteract people like that.
With artificial intelligence, machines can already potentially read all the books and articles in the library. Whatever has been digitized, at least. Conceivably, machines will eventually become better than human librarians at dealing for more complicated questions than simple fact checking.
A couple of ways machines can’t defeat humans
When patrons ask librarians for help, they seldom start with an adequately focused question. Librarians conduct reference interviews to discern what the patrons really want. Only with that understanding can the librarian actually start looking for ways to help.
Understanding the most useful search terms, maybe in the future a librarian can program search terms into a machine to get an answer. And then evaluate it for usefulness. Computers can’t do that. People don’t drop off assembly lines all thinking alike.
The machines themselves also won’t be capable of judging how well their output conforms to basic library principles, either. Librarians will take the lead in advocating that artificial intelligence respects and exemplifies such principles as authority, intellectual freedom, privacy, and others.
Robots in libraries–science fiction?
But here’s another thought: what if robots become library patrons? Imagine going to the library and seeing not only other humans but something like C3PO and R2D2 also seeking information.
The human and robotic patrons will have different information needs. The robots will need all the data they can acquire, both structured and unstructured. If such ever exist, they will certainly change the ways humans communicate with machines. They may change the ways humans communicate with each other.
Artificial intelligence / American Library Association, February 4, 2019. Document ID: 8846edce-ad2a-4b3d-a72b-daa554da4d11
The robots are coming? Libraries and artificial intelligence / International Federation of Library Associations. July 24, 2018
What happens to libraries and librarians when machines can read all the books? / Chris Bourg, Feral Librarian. March 16, 2017
Will AI make libraries go extinct? / Amitesh Jasrotia, Book Jelly. July 21, 2018