child meeting robot

When I’m asked what I think about ChatGPT I find myself conflicted by what I know as an engineer of chatbots, and what I know as a human being, writes Amy Lomax of LINK Mobility.

My overwhelming thought is that ChatGPT is clever. Ask it to write a poem about bitcoin in the style of Shakespeare and you’ll receive one in the time it takes to say ‘Macbeth’. The indexing and retrieval of information to create such a capable machine is to be applauded. I spend considerable time ‘training’ chatbots used by retailers to help customers with problems. Programming appropriate answers to questions takes effort and thought.

Every business, and every person, is different, and the language used is all variable, so every scenario needs to be carefully considered. How a grocer’s chatbot understands and answers a question about a substitute in a delivery, will be different to how a shoe shop chatbot answers questions about exchanging items.

The difficulty of machine learning and nuance 

Though similar, they are very different queries, and both can be asked in different ways. For instance, take the example of someone receiving Jaffa Cakes instead of Jaffa oranges. The chatbot could be asked ‘why did you substitute oranges for Jaffa Cakes?’, or even ‘why did I get biscuits instead of fruit?’. The nuances need to be examined and managed, as do the points at which a conversation is handed over to a human to help.

ChatGPT has also transcended boundaries to make chatbots a talking point among young and old, technically minded or not. It’s best illustrated by the headlines that draw attention to the way ChatGPT tempts students into plagiarism. It’s made the front pages in a way chatbots never have before.

AI helps us solve problems quickly

I’d argue that’s a good thing. Firstly, it’s made people think about the value chatbots can give to society. It opens another form of accessible communication. It helps save time. It can help share knowledge.

In a work context, it’s meant that business leaders are asking me how chatbots in general could change how they do business. The most obvious way is to use them to give customers straightforward help quickly. This might be along the lines of where’s my order or when will you have more items in stock? These are the things bots can be trained to ‘look up’ in a delivery scheduling tool, or a stock inventory system.

ChatGPT convincing but full of ‘hallucinations’

However, I noticed a shift in enthusiasm for using ChatGPT specifically to replace people and write content for websites, when the media ran stories from Google about warnings of ‘hallucinations’. This is the term used to define when the machine gives a highly convincing but fictious answer to a question.

It underlined that few people had acknowledged that a chatbot is only as good as the information it’s fed. If resources are old and wrong, then the output can’t be relied upon. A day can be a very long time in politics. A thesis written about energy policy today could be inaccurate tomorrow. Chatbots must be maintained and kept up to date and factually true to be valuable.

How to make the machines more human

But it’s not just about accuracy and timeliness. It’s also about the humanness. Can a robot be empathetic to a person who is upset about receiving Jaffa Cakes? It might seem like a glib statement, but it’s of great importance that the interaction someone has with a chatbot strikes the right tone and doesn’t make matters worse.

Sadly, it’s become abundantly evident just how important this is when it comes to helping people struggling financially. My first-hand experience of developing for these specific situations for financial institutions, tells me that chatbots encourage people to seek help when they would otherwise avoid picking up the phone.

Behind a keyboard, people can set aside feelings they might have, like embarrassment, and ask if they can put in place payment plans to manage their financial load.

The chatbot goes beyond human assistance

What I have learnt in studying for qualifications and day to day engineering of chatbots, is that training the chatbot to help people in this way goes far beyond providing factual assistance. It takes an understanding of what drives someone to ask for help. What are the situations someone may have found themselves in through no fault of their own? Inflation driving up the cost of essentials and high energy bills are just two of the many reasons why people will be falling behind on repayments for purchases made in the past.

Tackling these scenarios, with a team of talented people, has taught me that a collective intelligence needs to be applied to make chatbots successful. There’s more life experience as much as technical experience to draw on. This is especially evident in creating chatbots capable of handling more sensitive topics. Being overwhelmed by grief might be why someone will choose to use a chatbot to ask about closing a loved one’s account. Picking the right language for the chatbot to use is crucial.

Women can make an enormous contribution to empathetic chatbots

It’s for this reason then, that my biggest learning is that women can make an enormous contribution to empathetic chatbot design. In my experience, when there are more women involved, the design has a different lens applied.

Furthermore, I believe it will progress equality. How could a recruiter use a chatbot to find great candidates without any bias of gender, race, disability, or religion? When you pose questions like this to an HR audience, the reaction is powerful; women can play an extraordinary role in helping other women, and people of all backgrounds progress their careers.

Women need to get more involved in AI development

I’d like to encourage more women to think about how they could get involved in artificial intelligence. Having a technical background is not essential – I did a degree in International Business and French. What is essential, is being able to make human connections, see the bigger picture, apply life experience, and translate it into language and interactions that are warm and engaging.  We know that having a more diverse workforce is important for innovation and profitability, and this applies to chatbots too. The more diverse the engineering pool, the better chatbots will be at helping people.

Chatbots are here to stay because there’s huge potential to create machines that positively add value. But I really do believe that the more involved women are in their creation the better the outcomes will be for society.