Unmaking Sense

By leveraging the power of AI to tailor itself to the requirements of users we outline a revolution in education.

More on whether noise is really noise, and whether an AI or another person or species might be able to see more in it than we can. And separately whether the power AI could confer on those who control it represents a serious existential threat.

Stable diffusion: the noising and denoising processes; latent space; text-prompts, embeddings and the common language they share with diffusion models. Illustrating diffusion with intermediate images that show the whole process from seeding image to final product.

Information, Acculturation, Socialisation, Activation, Exploration: how the latter four hinge on the first; and what happens to them if and when the first no longer needs to be delivered through institutions that teach collectively.

We extend earlier discussion of latent space to encompass joy and misery, optimism and pessimism by considering how human minds can react sympathetically to their perception of the latent spaces intimated by their experiences.

The third in our internal mini series based around denoising latent space: that we cannot see anything in “what is directly in front of us” does not imply that there is nothing there to be seen. By having faith in our intuitions, doggedly putting one foot in front the another, and watching what emerges we may yet find ourselves rich beyond our wildest dreams.

The second in a three-episode linked sequence exploring meanings emerging from latent space.

Episodes 11.54, 55 and 56 were all recorded in one thirty-minute walk. They start from nowhere and end up somewhere, just like the metaphorical denoising of latent space that creates a new image in and of the world.

How diffusion models afford us the ability to add technical, material support to intuitive cognitive claims such as tacit knowledge, in dwelling, and the ability to intuit fecundity even where there is no apparent pattern or meaning.

Details of the way diffusion models noise and denoise images and text to different levels and what that tells us about the sense we can make of the world. In particular, that human intelligence cannot discern a pattern does not imply that there is no pattern there to be discerned; noisy images can and do embody the legacy of the parent images and text whether we can detect it or not.