WORD Christchurch 2023: Will Team Human take the day?

An assorted crowd of booklovers is gathered in Tautoru/TSB Space at Tūranga for what might be the deciding battle in the war between humans and machines. You probably thought that happened in one of the Terminator movies but you are obviously mistaken. I'm not sure what use Sarah Connor would have been in this conflict, to be honest. Is she even much of a reader?

But disses of fictional badasses aside, a good (and informative) time was had by all.

My Tūranga colleague Brian was at the mic guiding us through this AI adventure ably assisted by Max (not Headroom) at the laptop keyboard and the rockstars of readers advisory; Mike, Sarah, Eamonn and Fee (though Fee was for some reason was showing support for machines by wearing a Dalek t-shirt?! - what team are you on, Fee?)

AI vs Librarians
Team Human and Team AI face off at Tūranga, WORD Christchurch 2023.

To start with Brian ran us through a bit of an AI primer pointing the possibilities but also potential problems that AI might cause. Will they replace us, librarians? Well, here's where we find out as we pit human book recommenders against the power of AI, ChatGPT. Brian assured us, whatever the outcome we were in for "a spirited conversation about the future of humanity... or something." We were also informed that the key to the success of the evening was audience participation in the form of supplying book recommendation questions that could be posed to Team Human and Team AI.

By way of introduction we were treated to various and stats and facts relating to the two teams. Team Human has over 70 years of combined experience in recommending books, and ChatGPT has $US 11 billion in funding etc. HR won't let me be too specific, but I can confirm that your average librarian is far less expensive than that.

And so the questions were put forth by the audience and after 1 minute of conferring and thinking time Team Human would make their suggestions, and then the question would be put to ChatGPT and the audience member would judge which options they liked best. One person wanted rom-coms, someone else had a sizeable list of authors they enjoyed including Fleur Beale and Tania Roxborough, but also they expressed an interest in Chinese or Japanese stories. 

Problems emerged with ChatGPT's answers early as some titles suggested were not by indigenous authors, and at least once ChatGPT suggested that we "reach out to a librarian" for more recommendations. Way to shoot yourself in the foot, ChatGPT! 

By the time someone had asked for reading similar to An Unsuitable Boy by Vikram Seth (who promised a sequel but has so far not produced one) a pattern of suggestions predictably lacking in the human touch, but also sometimes a bit obvious was emerging. Before we knew it, the librarians were in the lead 3-0.


Yet more requests came from someone who like enormous fantasy novels "with a map in the front and an index in the back". Map bestrewn titles were suggested by the Fab Four, including the Rivers of London series, The Poppy War trilogy and The Dragonbone Chair by Tad Williams. Fee  (the fantasy fan of the quartet) outrageously attempted to curry favour by declaring said audience member's taste in reading material fantastic and even declaring "you're my best friend", giving a clear advantage over ChatGPT which never complimented anyone the whole time. ChatGPT's regurgitated answers were deemed by the questioner to be "pretty good by normie" and "decent... for a machine". This may have been a fair appraisal but I fancy I heard one of ChatGPT's computer chips fizzle to a crisp at this low-level burn.

In between book recommendation requests, ChatGPT was asked to generate some literary trivia questions and so it was that a number of people left with, not just a bunch of book titles to check out, but also PRIZES.

After several rounds Team Human was in a strong position. Would the last request be their downfall or would it be a (virtual) down-trou for the AI?

A reader asked for a recommendation similar to Perfume by Patrick Süskind - historical, descriptive, atmospheric with a baddie that you might empathise with even despite their badness.

Sarah came in strong with Miss Smilla's feeling for snow and the team was agreed that The master and the margarita was a solid suggestion with its historic setting, engaging story and really interesting characters, with Mike declaring "I read it on the bus and I can't read on the bus without throwing up". High praise, indeed. You'll not get a recommendation like that from ChatGPT (which, as an AI, is not capable of throwing up). ChatGPT overed a couple of suggestions that didn't particularly excite our reader, and somehow also recommended the same book, Perfume, back to her which was not particularly helpful.

And with that, the fate of humanity (and librarianship) was decided.

Will AI replace us? Perhaps. But not today, my friends. NOT TODAY.

Find out more

Want more? Try Fee's list, Librarians vs AI : Twelve of the Best AI stories

List created by fionajay

Just so we're clear, Artificial Intelligence doesn't do emotions, right? Here are some of the best AI stories that challenge our assumptions and push the boundaries - in this, the century of realisation. Written by Hoomans.

View Full List