In late September of 2001, an office manager at the publication I worked at quietly placed a stack of business cards on my desk. “I didn’t know what to do with them,” she whispered. I placed my hand over 12 cardboard rectangles and burst into tears. A big, ugly cry, right at my desk, amid the empty Diet Coke cans and second-hand smoke.
It seems an odd memory to bring up as a result of a throwaway comment made by Kate Royse, managing director at Hartree Centre for Digital Innovation, at IBM’s recent AI & Innovation Media Event. The gathering was meant to showcase IBM’s work with artificial intelligence and listen to industry leaders.
The panel — which also included Dr Nicola Hodson, CEO at IBM UK&I; Lord Ed Vaizey, member of the House of Lords and former minister in the Department for Digital, Culture, Media & Sport; Paul Weller, head of customer propositions and conversational AI at NatWest; and Michael Conway, partner and AI transformation leader at IBM Consulting UK&I — touched on all the major talking points about AI.
Those points included the need to watch out for the data used in AI models that result in hallucinations, or the possibility a generative AI chatbot will make false statements in a confident manner. There is also the risk that AI could be used as part of political misinformation campaigns. Mr Weller reiterated that conversational chatbots aren’t meant to remove human contact from customer service, but to support those roles.
The uniting point was assurance that AI would automate and take away all the mundane, repetitive and boring work from us humans. With a wry chuckle, Ms Royse commented that she “hadn’t written her own blog in ages”, implying that AI now did this ‘mundane, repetitive and boring work’ for her.
I gasped.
In the past I have refused to ‘write’ someone else’s blog post for them, because I consider posts of this nature to represent someone’s opinion. Writing has always felt personal to me — especially when I am sharing my views — whether it be a teenage diary filled with bad spelling, even worse grammar and hormonal emotion, or a post in The Banker that people share on LinkedIn.
But to others, a blog post is just content marketing, a search engine-optimised, written advertisement designed to lure people to a website to learn about what they really do. There is nothing wrong with that. However, it makes me question how we decide what is meaningful, what is mundane and in what context, especially as we are entering an age where those decisions will be coded into systems that will interact with our lives.
That publication I worked for 22 years ago, back when you could smoke in the office, was creating a central database of contacts. Every time we came back from an event or a meeting with a stack of business cards, we handed it over to our office manager to collate in a giant spreadsheet. Exactly the type of job most would argue could be done by a robot or AI today.
The database was meant to make it easier to contact the head of comms at banks or tech companies, without searching through the little slips of paper, napkins, and half-filled notebooks that most journalists relied on.
At the end of September 2001, this person noticed there were 12 cards, 12 printed names of people, in my pile that would never be contacted again. The annoying PR girl who always called five seconds after sending a press release. The excellent product guy, who knew banking inside and out and could always pinpoint the best place to grab a martini in any city. The old boss, who I had lunch with months earlier, and had invited me to the very event at the top of the World Trade Centre in New York that he and the 11 others on those cards would walk into and would never walk out of.
What would an AI have done with those 12 cards? Would they have quickly removed them from the database, never alerting me to their existence, thus denying me a moment of cathartic grief? Would they have created a soulless, yet accurate, category to place these names, simply labelled ‘deceased’?
More importantly, who would get to decide, who would teach, this AI model in which context 12 worthless scraps of paper were not worthless at all?
Mr Weller made a good point at the IBM event. Whatever AI is, he said, what it is not is “empathic”. There are some who see this as a benefit — times when emotions cloud judgment and using quality data is the better choice to determine outcomes.
But empathy lies at the centre of storytelling. We as humans see the greater truth in what is not said, the meaning in mundane objects that let us understand the ‘why’ of certain decisions, in addition to the ‘how’.
The shortest story ever written was created at the start of the 20th century. It is often attributed to Ernest Hemingway, but Wikipedia says otherwise. I’ve seen this story used to talk about how we teach AI and how those models should learn nuance and context, and to read between the lines and what is not said. It is an important thing to consider as we know the meaningfulness of an object or activity is dependent on which human is interacting with it, not the basic facts of its existence.
Where one robot sees a classified ad, others see a tragedy. These distinctions lie at the heart of where AI will sit within human lives.
For sale: baby shoes, never worn.
You can connect with Liz on LinkedIn, or follow her on Bluesky @lizlum.bsky.social.
Register to sign up to Digital journeys, our new weekly newsletter, as well as the latest analysis and features from The Banker.
