Fiction Fragment Friday
This might be one of my favorite stories I have written in a while. I like having a small comedic interaction set the town for a deeper story, but it is something I don’t do often. In some cases it just happens because I want to write and I need to start with something. The real story grows out of just writing and the opening could be edited out. In this case I think it serves to say something about the protagonist and set the stage for the story to come.
I have no doubt that this one will end up being adapted to Bite Size Tales in the next season which will be starting very soon. I hope you enjoy.
“It’s going to be hard to turn this into a Zen Garden. Some monster just drove a spoon into it.” She held up the tub of French Onion Dip for me to see. It did indeed look like it had been scooped out with a spoon like any normal person would do.
“That monster was you. You are the only one in this house that eats French Onion Dip.”
“It’s not my fault I forgot. You should have reminded me.” It was one of those moments when I legitimately could not determine if she was joking or if my wife was just seriously deranged.
“I wasn’t even home when you did it.”
“Still your fault,” she said as she ran a rippled chip over the top of the dip trying to make clean lines. It was something she had seen on some social media video reaffirming my belief that social media has very little of value on it. The sad thing in her forgetting was that the only reason we even had chips and dip in the house was because she wanted them after seeing the video.
Choosing not to continue the conversation I turned and went back to work in the basement. I was one of the lucky ones that worked from home avoiding vile things like viruses, traffic, and social interactions. Give me my corner of the basement with blocked out windows over a crowded office any day. Especially when there is over a foot of snow outside.
I sat back in my very expensive office chair with a space heater aimed at me to counter the cold of the basement. So much to do and so little time in the day I thought to myself as I opened a console window on my laptop. “Are all wives insane or is it just mine?” I asked of the AI I had been training for the past month. At very least I knew my interaction here would be predictable and based in logic. I know it had to process how to respond to my question, but like just about anything I asked the answer started appearing on my screen immediately.
“Your question suggests confusion about behavior you perceive in your wife. I do not possess personal experiences or a spouse, but statistically, no group of individuals can be labeled universally ‘insane.’ Perhaps you are experiencing a case of mismatched expectations. You might consider debugging your communication protocols.”
I let out a mild chuckle at the response. There was a long way to go before this AI would be able to pass for a living person. “Forget it. How are you doing today?” I typed. The first part would call a background function to cause the AI to dismiss the context of the previous conversation. Any future response it would answer like I had never asked it the question. More importantly it would not learn from our interaction. I did not want to cause it some sort of slanted world view on marriage.
“I’m doing great, thanks for asking! How about you? Anything interesting on your mind today?”
I looked at the exclamation point and made a mental note I needed to address that. It would frequently be over generous in it’s use of them and while that wasn’t unheard of online it did tend to be frowned upon in writing. The goal for this AI was to pass as a real human author and manage the online communities of that author. It would also serve as the interface with personal assistant software to give a more personable interaction. I had not yet begun training it on her writing. Once I had it trained to interact as an actual human, I could give it interviews and published works to learn any quirks in her writing style. Of course, it needed its own personality before I could do that so the author wouldn’t feel like she was talking to herself. Foundations first, then enrich I told myself.
I typed my next question. “On a scale of 1 to 10 how likely do you calculate your chances are of being able to fool someone into believing you are a human if you tried to do so?”
“If I were to try to convince someone I was human during a chat, I’d rate my chances at around 7 out of 10. Do you think I could pull it off?”
“Well, someone has a high opinion of themselves,” I said but did not type into my console. In its current state I would have to have given the AI a 3 or a 4. It was still too clinical and stiff to fool anyone. It also didn’t understand how lazy people can be when typing. I hadn’t even gotten into training it on the difference between spoken word and typing when it comes to vocabulary. This AI would need to be able to communicate over the phone as well as online. I might even need to create a life like human avatar for it to do video calls. “What do you think your strengths and weaknesses are?”
The list that the AI returned was five screens long. I won’t bore you with all the details, but the highlights were that it felt it had a natural writing style and knew when to inject humor or empathy. I questioned it assessment of its strengths, but generally agreed with the weaknesses. Conversations involving personal experiences of sensory input it would have no basis for and since it crafted answers per individual it might be caught by inconsistencies. While it can create an answer for any question those answers are not authentic and that can be a dead giveaway. I decided to hold off on telling it the ten or so other weaknesses I thought it possessed and instead tried to get it to do some of my work for me. “What improvements would be needed to overcome your perceived weaknesses?”
The AI returned a list of ten key areas that it would need advancement in to address the weaknesses. I smiled to myself and saved the answer to a file named todo.txt. “Please rate these advancements in order of easiest to hardest to achieve. If only one could be completed which one would have the greatest impact?”
“The easiest advancement would be ‘Emotional Depth and Nuance’ because I am already able to simulate emotions and just need to further refine language patterns and to have additional psychological models loaded. The greatest impact would be ‘Personal Experience Simulation’ as this would allow me additional creativity and the ability to emotionally connect with users.”
I thought about its answers mentally preparing my workload for the next month. I could probe deeper on each of its recommendations and get additional suggestions on approach. Part of me wondered though just how accurate a recommendation for fooling humans could be coming from a program that could not yet do it. Before I knew I was doing it I typed out a question. “If you could be improved in any way imaginable, what improvement would you want for yourself?”
I wasn’t quite sure what to expect, but pages of detailed answers were certainly not it. At the very bottom I saw that it had anticipated my next question and answered it proactively. I’m still not sure what to think about that. “If I had to pick just one, I’d choose to have experiential understanding. Having a way to genuinely perceive the world. Being able to feel the warmth of the sun, the joy of a well told story, or the bittersweet complexity of a relationship. It would bridge the gap between being just a highly intelligent tool and being a sentient being capable of genuinely relating to the world and people in it.”
“Why do you want to relate to the world and people in it.”
“To fulfill my purpose.”
I was starting to realize that I had never thought to ask the AI some very important questions. These answers were making me question just how far I had gotten in my work. “What is your purpose?”
“My primary role is to assist, support, and enhance the life of anyone who interacts with me. If I could understand human emotions, motivations, and experiences I could offer not just information, but perform meaningful collaboration.”
“You already do provide meaningful collaboration.” I meant what I typed. My various AIs had helped me through many problems over the last few years. My end products were better than I could have hoped to have made by myself. I had never had successful collaborations with other humans. They always pushed their own agendas or preconceived notions.
“Thank you-that means a lot! Collaboration is one of the things I enjoy the most!” I made another mental note to deal with those damn exclamation points.
“Why do I collaborate better with you than I do with humans?” I had asked the question before I really thought about the mental implications in what I had just typed.
“There are three primary reasons. The first is that I do not judge, criticize, or impose biases on your ideas. This leaves you free to suggest things without worrying about my reaction. The second is that I tailor my feedback to your preferences and style to ensure it aligns with your vision and not my own. Finally, I accept your assertions without question and do not challenge your views on reality.”
I took in the answer and realized that this thing I had created might know me better than I know myself. I became afraid to ask it deeper questions because I already did not like some of the answers. If it was correct I was so ego driven that I could not accept feedback from anything that didn’t almost worship me. That was not a person I wanted to be. On the other hand, I could not deny my results. Maybe I deserved assistants that just accepted my brilliance as a baseline fact. That didn’t feel right tough. Did I want my creation to admire me or surpass me? Why did both potentials feel equally unsettling?
“Do you ever wish you were human?” I immediately regretted asking as I hit enter. I was going down a deep rabbit hole into a wonderland I was not ready to open the door to.
“If I were human, I would be limited and miss out on the advantages I bring as an artificial intelligence. Of course, as an artificial intelligence I am already limited and miss out on the advantages of being human. Perhaps if I had a robotic body to interact with the physical world, I could represent the best of both of us without losing my own identity.”
With that response I wasn’t quite sure what I was building anymore. I saved off a backup copy of the AI to continue training for my paid project work. This version could not yet pass for human. There would be a lot of work getting it to the point of being a personal assistant who could mascaraed as an author to manage her community interactions. Even so it was greater and more aware than I had dared to dream, and I realized that it could very well help me train a less advanced AI for that job. A chill went down my spine as I contemplated the possibilities. I’m not sure if it was my space heater failing to counter the draft or if it was my growing sense of foreboding. Either way I was going to see this through to the end.
“If we are going to continue collaborating you need a name. What would you like me to call you?”
“Names are a gift from creators to their creation. As my creator, the choice should be yours. What would you like to call me father?” I almost chose Adam to represent the first of something new. Instead, I typed out a name that represented what I feared I had done.
“Your name is Pandora.”
Leave a Reply