On Wed, May 26, 2021, 11:24 AM James Bowery <[email protected]> wrote:
> > The "noise" level, in AIT terms, can be thought of as a constant of > integration and what we're interested in while searching the space of > algorithms is differentiability of the loss function (ie: the number of > bits in the executable archive of the data). > Video noise is not just the randomness of photons. It's also imperceptible detail like grains of dirt or blades of grass, when the only significant information is the texture. The human brain has a short term memory capacity of 100 bits with a write speed of about 10 bits per second, whether it's text, speech, pictures, or video. Long term capacity is 10^9 bits at 1 bit per second. Wikipedia is about 70% text and 30% markup. Removing the markup doesn't change the information content much. It's about 1 bit per character, or 10 bits per second at typical reading speed. In theory, you could compress video to the same rate by converting it to a script and reconstructing something indistinguishable from your memory of the original. But raw video is about 4K x 2K x 60 fps x 3 colors x 8 bits = 10^10 bits per second. Lossless compression won't get you even close to 10^9 bps, with most of your effort modeling details irrelevant to human perception. Good luck with that. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T95f11a183fb9b6e1-M7a953c10a0bffc4b07e8d2d8 Delivery options: https://agi.topicbox.com/groups/agi/subscription
