We are all in on creative AI at Mars this year. We’re excited about the rise of large language models and generative text-to-image systems like GPT-3, Stable Diffusion, DALL-E 2, and others. We’re busy building Eden, which will make it easier than ever for Martians to build on top of these transformative tools.
Generative models let you near-instantly generate images or text of almost anything, through natural language interface. This has given rise to a new guild of “prompt engineers,” people practicing the art and alchemy of prompting, or composing the text for these models. Budding prompt engineers gather on Twitter, Reddit, and Discord to share tips and tricks for getting these models to do what they want, faster and more accurately, while startups are forming to serve them.
Sounds easy, right? Wrong. Early prompters began to discover surprising tricks like for example how stylistic modifiers (e.g. “cyberpunk”, “HD wallpaper”, “rendered in unreal engine” etc) gave them very fine-grained control over the content and character of their outputs. By the time techniques like inpainting, outpainting, and interpolation extended text-to-image by another dimension, prompting had become an experimental science. More recently, techniques like finetuning and textual inversion give even more control and precision.
Prompt engineering may be mistaken for an artistic fad, but it represents more. We think it is driving the evolution of software development. Systems like Replit and Copilot hint at this. Prompts can compile to modern code, or lower, enabling people to develop software without coding. This is the future of human computer interaction.
Major in AI at Mars College to sharpen your prompting skills.
Calling all artists, prompters, and developers at Mars this year to build awesome new applications on our foundation.
We’ve been enthusiastic about AI since our first year. The Abraham project was incubated here, and we were early makers of GPT-3-powered chatbots on our Discord. Some Martians have been collaborating on a project called Eden, which hosts and serves models like GPT-3 and Stable Diffusion on-demand, as well as scaffolding to make web applications and Marsbots, our Discord bots. We have our own GPUs and are developing an API on top of them.
What we are planning for the AI art major:
- Regular prompting workshops, where we make images and text, tell stories, make up games, and other fun stuff revolving around generated content.
- Application development for the more technical, help us build second layers on top of these models through interactive web applications or other media frameworks like TouchDesigner.
- Working together to imbue The Simulation with AI.
If you are interested, come join us!