Death is not a happy subject but.. it happens. The average age of death continues to rise due to medical advancements and increases to quality of life. Generally, people are living longer but tragedies happen and sometimes will die long before that average age expectancy. Death is something that affects us all and in a world where technology is getting better, how can AI be used to remember the dead and is using such technology to remember the dead an ethical dilemma?
Dreambooth is the ability to fine-tune an image AI model. You can do this by providing just a few images of that person, dog, cat, car, or object. It can be anything. The image AI can then be instructed to create further images of that subject. The results are surprisingly good. This can be a fun tool to make yourself look like Iron Man or Spiderman or see what you look like in the Game of Thrones universe but can it be used for something more? Should it be used for something more?
Losing a close friend or a family member can be pretty tough. Everyone deals with grief in different ways and often you can find yourself looking at the same videos or images of that person which helps with remembering them. Using Dreambooth on a collection of images of someone who has died will enable you to see images of them that have never existed before and as video technology improves, ultimately bring you a glimpse of them again that was never possible up until now.
Grief from an unexpected loss can be pretty destructive. It can lead people down a path that is hard to recover from so if something could exist that gives people some comfort in an unexpected death, should it be built? I am of a pretty libertarian mindset and for free choice as long as it doesn't impede on the rights of others which then begs the question, who gives you the right to use images of the person in such a way and will be something that people will need to opt-in / opt-out of in future wills?
Dreambooth is just one of the many technologies and advances in the AI field. It requires generally 10+ images to get good results, new technology is already being worked on to be able to create pretty good consistency across images from just a single image. You can read about that here. With models like Cogvideo adding movement to images, it won't be long before you can take an image and create a full video from it.
The way that audio can be trained is also getting better and better. You can clone your own voice relatively easily with services like this so how far away are we from taking an image of someone who has sadly passed, turning that image into a video, and adding audio of their voice on top of that video. It sounds crazy, it sounds like sci-fi. It doesn't sound possible but it will be. It will be sooner than you think. It's most likely going to be this decade. Should AI be used in such a way? Well, if someone wants to do it, let them do it. If you don't, then don't. It is a bit of a can of worms but everyone should have the freedom to choose.
I wrote this blog post as I have found these questions in my own head. I have trained a model dealing with these issues and I have got some surprisingly good images from AI. Images that make me smile, that make me remember, that make me sad, that make me feel all sorts of things all at once. I'm a big believer in making this technology as accessible as possible to everyone and we're still at the very early adopter stage of things. Is this use case any more controversial than using AI to make adult content? I don't think so but this is also an incredibly private matter which is why I don't feel comfortable sharing such images here.
Here at Riku, we're all about bringing all of the best large language models together under one roof giving you the ability to experiment, learn, and deploy them for your needs. Our educational platform makes these models and technologies accessible and we recently added Dreambooth into the mix for you to try out for yourself. If you are interested in getting involved in AI and don't know where to start, Riku is a great first step.
