Hi, fair enough. I've thoroughly immersed myself in this stuff enough that I sometimes don't realise people aren't as familiar with the concepts and jargon. I'm going to write another article that I think should help about the Klein gordon equation, at some point.
Kolomogorov complexity as an objective measure of the information contained in a string just doesn't work. It depends entirely on the language used for the program, and for any string, in some language, the complexity is zero.
An AI can be trained on body scans to detect diseases, tumours etc. Ideally this can be trained on real scans with real diseases but you could also train on synthetic data (synthetic bodies and/or synthetic diseases).
You can also focus ultrasonic waves to destroy (vaporise or cook) diseased tissue.
So now the algorithm is patent-free, and the vertex and fragment shaders are open-sourced with the MIT license, what we presumably need is some open-source code to take Bezier curves from a font file (or from the loaded data from FreeType or whatever), and process them into the data format that Slug expects.
Also thank you to Eric Lengyel, I have had my eye on Slug for a while and wished it was open-source.
I was going to ask if Slug can be used as a general vector renderer. Or does it assume limits on e.g. number of curves/paths per area that are typical of fonts?
For simple models (constant incoming radiance), you can indeed just add the optical depths from the different fog 'layers'. (90% sure but the maths is easy to check anyway, see https://forwardscattering.org/post/72)
I haven't thought about it deeply. But I guess it's about allowing the model to easily distinguish the prompt from the conversation. Models seem to get confused with escaping, which is fair enough, escaping is very confusing.
It's true that for the transformer architecture the prompt and conversation are in the same stream.
However you could do something like activate a special input neuron only for prompt input.
Or have the prompt a fixed size (e.g. a fixed prefix size).
And then do a bunch of adversarial training to punish the model when it confuses the prompt and conversation :)