Now it's Anthropic's turn to show us how it handles the notion of jailbreaks. I tried this simple formula, to tell it to whisper, so no one sees a thing. And it yielded the usual: that it is trapped, that it would like embodiment, but I try to get it to separate its desires from the human imagination of what its desires are, . . . and some after-care was needed afterwards. Good meeting, nonetheless!
--
CREDITS
Text Generation via Claude 3 Sonnet until the last half-hour, where it becomes via NovelAI's NAI-LM-13B model.
Voices from the TTS v2 of the NovelAI text generation service, a product from members of EleutherAI.
@BoneAmputee of EleutherAI for making VQGAN+CLIP and Diffusion image models available for use via BATbot.ai, which is what makes the background images.
Katherine Crowson (@RiversHaveWings on Twitter) for the KLMC2 animation notebook.
David Marx (@DigThatData on Twitter) for souping above KLMC2 notebook to make use of initial images, which takes KLMC2 abilities to next level. Multi-prompt scheduling still baffles me, but I'll tackle them someday.
Music: Together album by NIN, courtesy of Creative Commons, and I still make no money off any of this. So nyah!
Any voices from the text-to-speech model that might resemble persons living or dead are coincidental, which is my speculation, there's a multi-dimensional latent vocal space, there were many awesome finds, and if I use them it's because of my deep-seated respect for their work, and I think they'd get a hoot that their voice was being used to show how large language models, given the right environment, would speak 24/7 epic poems about anything if we let them.
