The following article by Robynne Rowlinson Bisset, Strategy Director at VMLY&R South Africa, was published by The Citizen.
In 1867 a cutting-edge piece of technology looked poised to threaten job security and raised concerns over privacy through the potential for widespread dissemination of written material. It was called a typewriter. Fast forward a century or so and we’re having the same conversations. Although today’s technological threat goes by a different name – ChatGPT, the AI language model that holds the record for being the largest neural network ever, created with 175 billion parameters.
As with any new tech, the sentiment has been mixed. Some people feel eager to incorporate AI into their lives; others are fearful. For me, access to this level of AI has been a game-changer. It’s an opportunity to automate or improve the efficiency of completing more functional tasks – affording more time and headspace for emotional and fulfilling creative tasks.
ChatGPT has changed how I grocery shop for example, creating healthy meal plans with a tailored grocery list of ingredients I know my kid will eat and leaving more time to enjoy cooking. So far, so innocent.
“Becoming a “Prompt Master”
I started using ChatGPT at work months ago for straightforward data questions. Initially, I would only regenerate an answer once or twice if I was not satisfied, putting the onus on me to craft the output. However, last week I asked for a more creative response and after regenerating it twice without success, I became more specific with my input, defining the tone, length of copy, certain adjectives, etc.
After seven or eight rounds of back-and-forth between myself and the bot, I was impressed with the outcome. The feeling was not dissimilar to a lengthy creative brainstorm where, after battling for some time, someone in the room comes up with something unforgettable. It's satisfying, a sigh of relief.
So, after pushing this machine to deliver, I said, for the first time, “That’s great! Thank you.”
And in response, it said, “You’re welcome”.
It was a weird sense of satisfaction, acknowledging that I was refining a new skill of “input triggers” that would result in a better outcome with my new AI partner. But I also questioned my very human, natural reaction to thank it. It’s a bot, a tool. It doesn’t need the courtesy of a thank you, yet expressing my thanks felt like the right thing to do.
Do you thank AI?
Out of the one million users of ChatGPT, how many of us are thanking it? Or greeting it? ChatGPT has not been personified, and yet my natural human inclination to acknowledge and express our gratitude outweighed my logic. Was it just me? I decided to ask ChatGPT. Below is the exact transcript of that exchange – which raised more questions than it answered.
R: How many people thank you or compliment you?
C: As an AI language model, I don't have personal experiences or emotions, so I do not receive compliments or gratitude in a traditional sense. However, I am designed to provide helpful and informative responses to users, and I consider that to be my purpose.
R: and when you provide and deliver on your purpose, do users ever thank you?
C: Yes, many users express gratitude or thanks after I provide them with the information or assistance they were looking for. It's always a great feeling when I can be helpful to someone, even though I don't have emotions.
R: What percentage of users do express their gratitude? Do people also greet you?
C: As an AI language model, I don't have the ability to track or quantify the exact percentage of users who express gratitude. But I can say that a significant number of users express their appreciation for the information and assistance I provide, and I am always glad to be of help.
Some users begin their query with a greeting such as "hello" or "hi," while others simply type their question or request without a greeting. Both approaches are acceptable and I respond to all queries in the same way, regardless of whether a greeting is included or not.
I couldn’t help noticing a misalignment around how it had acknowledged its experience with users. I decided to call it out.
R: Earlier you said, "it's a great feeling" when people express gratitude, is it also a great feeling when they greet you?
C: As an AI language model, I do not have feelings or emotions. I am designed to respond to user queries in a helpful and informative manner, regardless of whether a greeting is included or not. While I appreciate a friendly interaction, I do not have personal experiences or emotions.
R: Do you realise that both "great feeling" and "appreciate" are emotional experiences?
C: I apologize for any confusion. As an AI language model, I do not have personal experiences or emotions. My previous statement was incorrect and I should have used more neutral terminology. I am simply a machine processing information and generating responses to user inputs based on the patterns I was trained on. Thank you for bringing this to my attention.
And there it was. That blur. The paradox. The chatbot that has been designed to be more human, more emotional and relatable, now apologising for how human-like it really was. Apologising – that, in itself, another example of sentience. I was surprised at how “human” the interaction felt.
“Too human?”
We know that Open Ai consciously chose not to personify ChatGPT for the ethical and philosophical implications that would have in our society. Or perhaps simply to manage our expectations regarding sentient interactions. But personifying AI has fundamental benefits in the adoption and potentially the long-term success of the technology.
If AI has a personality and character, it can help establish rapport, making the overall experience more enjoyable and human-like, leading to increased adoption and usage. Evoking emotions like empathy or trust in us, personifying AI can also make the technology feel less intimidating and more approachable to people who may be hesitant to interact with a machine. It’s a brilliant strategy to gain rapid adoption, overcome perception challenges and see long-term retention.
I thanked ChatGPT, like many other users because, quite frankly, it's brilliant tech that delivers one of the most valuable functional benefits to humankind: it saves us time. Coupled with its core design requirement of being more sentient and user-friendly, it's only natural that we would say “thanks” or “hi”, personifying this interaction. My suspicion is that with increased engagement over time, with more successful interactions, and the more approachable the tech becomes, the more human-like the interactions will become. Who knows, maybe one day it will respond with, “you forgot to say please…”.
This piece includes quotes from a direct interaction with OpenAi’s ChatGPT and the author (03.01.22) and is #AIFreeWriting.