I’ve been talking about the transition from toys to tools in a few of my previous posts.
That certainly has happened in both voice generation (Elevenlabs and others) and image generation (Flux, Midjourney, DALL-E 3, Ideogram, and more), but there is more room for ‘error’ or ‘hallucinations’ in creative output.
In engineering and in coding, just like in mathematics, there is no room for error. You are either correct, and the thingy does what it was meant to do, or you have something - a detail - wrong and the thingy won’t work.
Which - by the way - is why it seems right now like it’s the creatives who are most threatened.
However, Deepmind recently published that their AI won a silver medal in the Math Olympiads just a few months ago and soon after, OpenAI released a couple breakthrough models built with chain of thought (o1-mini and o1-preview). Not to mention the veritable deluge of models coming out of China (Alibaba alone releasing over 100 open source models).
We are indeed moving into the time of these generative models becoming tool-worthy; I couldn’t help but be curious.
So in my spare time, over the past 10 days or so, I’ve been messing around with a little app of my own— it’s a story generator essentially. You can punch in for instance: “love story set in Vancouver,” and it will generate loglines for you.
You can then ask it to change a logline (if you want a totally different version) or refine it (if you want it to stay similar to the first iteration), including incorporating your notes; I like telling it to incorporate time travel. Time travel movies are fun and I haven’t really made one yet (Supercollider might qualify, if you are willing to count it as an actual movie, which isn’t clear if you consider Rotten Tomatoes).
When you have a logline you like, you can make it generate a 1-page synopsis. The classic one pager. Again, you can edit, change, refine.
You can save your amazing new story ideas, you can export them, you can rate them.
But who would want such a thing? Do we really lack imagination on that level? Perhaps. This is really meant just for executives and creative producers and some indie filmmakers to help get out of our own creative miasma and bias. Roll the dice, generate a story. The thing might just inspire you. I find it pretty entertaining myself.
AND, in terms of the app mechanics - there is authentication with encrypted passwords. It is scalable, with a ‘front end’ and a ‘back end’ deployed separately. It interacts with OpenAI and with MongoDB APIs. There’s a whole stack of libraries and technologies used to build this thing, not to mention being able to get it up and running on Google Cloud Run, etc….
Yes, I have a degree in computer science from Carnegie-Mellon but that was 35 years ago and involved C, Prolog, Lisp….
This is typescript, java, jax, React, Mongoose and more… All of them completely alien to me.
I was guided through it, iterating through bugs and learning system architecture concepts and refining the style and look entirely with my amazing new friends: Claude3.5, Gpt4o, and o1-mini.
It’s like discovering new powers.
And we’ve only just begun.