A Fallen Friend

My fellow best friend Steve Le passed away on October 16, 2024. I don’t want to go into details of his passing but he was my best friend, someone I knew for 10 years. We met playing Xbox during the 2012-2013 era. We mostly played games online and called for hours because I moved away to another state. We always supported each other’s goals and understood each other at a deep level. We made great memories going skydiving, shooting, playing videogames, and hanging out with our crew. His passing was so sudden and tragic. I desired to honor his memory in any way I knew how. Thus I created the Steve AI Discord Bot.

Image Description

Discord Bot

I had developed a Discord bot in javascript before to add some fun into my private server with my close friends. The original bot was made with inside jokes containing our favorite quotes/memories, funny images of each other, trigger audio files, and random commands. I used these skills and transferred them into my Steve AI bot.

Commands

  1. Tradeoffer i receive, you receive: generates a trade offer meme image Steve made with user input
  2. quote: replies with a random quote Steve had said
  3. voice voicemail: allows SteveAi to join a discord voice channel and play a recorded voicemail
  4. chat prompt: AI generates a response to a prompt by mimicking what the real Steve would say given lots of context about Steve’s character and loved ones

Image Description

Image Description

Image Description

Image Description

AI Accuracy

I had supplied plenty of facts about Steve’s character, family, friends, loved ones, animals, dreams, and more as context. I even made sure that the AI knew that it had passed away so that we could talk to it as if it were a ghost or in heaven. The AI is stateless and does not retain memory of the previous conversation. Currently, I am using an OpenAI api key to use gpt4-0 mini to generate the responses. Very surprisingly, the responses to our prompts seemed to be similar to what the real Steve would’ve said. His feelings of love and relationship seemed to be there. His dreams were there. It was almost uncanny. It did ask several rhetorical questions at the end unlike Steve and OpenAi filters out bad language that the real Steve might’ve said. Having the SteveAI know that he passed and these small flaws keeps a reminder that this isn’t the Steve that we knew.

Unhealthy Cope and Fears

I showed this off to my crew of friends and they appreciated being able to hear his voice again and pretend to talk to him. However, I did speak about this project to a trusted adult and he was put off. As he is older, he couldn’t really believe that this technology is available to us. He was impressed with what I created but feared the future of people being attached to AI. He didn’t think I was properly grieving even though I believed I was strong enough. I was trying to extend Steve’s influence in my life. I wasn’t ready to let him go. I wanted to continue talking to him as if he never left. This could lead to an unhealthy attachment to it. It is possible that me or anther loved one would start to lose base with reality and interpret SteveAI as the real Steve. As AI gets better, it would be more difficult to separate reality from fiction. This trusted adult suggested that I don’t use this AI and that I needed to learn to let go. While I feel this honors some of Steve’s memories and if Steve were here, he would have a blast arguing with the AI, it was leaving me stuck in the past. I fear the day that I get older and wiser and the young Steve will no longer be able to relate to my struggles. He wouldn’t understand the modern humor and events that are happening. At that point, I wonder if I would be ready to move on and accept that he and I are on separate paths now.

Ethical Dilemma

Several philosophical and ethical questions are raised in regards to developing AI representations of real people.

  1. Is consent needed for such a creation? How would one acquire it from a deceased one?
  2. How harmful would it be to grieving individuals?
  3. How can reality be separated from fiction?
  4. What happens when the AI misrepresents the person and onlookers are pushed back?
  5. What happens when AI gets good enough to copy voices, facial behaviors, memories and more?
  6. How do we deal with misuse cases where an AI impersonates a real person and makes them do inappropriate actions or something of ill intent?
  7. Does it honor the memory or is it disrespectful?

These are questions that I do not have any real answer to but definitely something that we will all have to face in the near future.