Sunday, June 25, 2006

A conversation with Danny

"Danny showed me his video parser. I gave him a .3gp video of my son, newly arrived back from snowboarding in Canada. The video, from my camera-phone, was tiny and grainy. Danny's program read frame after frame, displaying each magnified pixel map. It was incomprehensibly to my eyes.

"The program does standard things like edge-detection, object-inference and finally scene-reconstruction. It's based on the detailed models of people and objects stored in its object-database. The parsed result is a high-resolution 3D animation model similar to those we use to generate video games. We can then invert the process and render this model at any resolution we choose."

Danny showed me his processed video based on my original postage-stamp movie, but it was now rendered on a high-definition TV screen. And it was perfect.

"You do realise the implications, don't you?" Danny said. This cinematic version of your video is a reconstructed reality, a kind of 'informed hallucination' based on your original coarse and tiny video. And that too was a kind of hallucination of reality, formed from pixels illuminated by your camera lens.

"And as you sit there looking at me, your brain is doing just the same thing" he continued. "You have no access to the 'real me'. Your 'view' is entirely reconstructed from retinal images on the back of your two eyes: tiny, noisy and upside-down neuronal 'bit-maps'. It looks like you're seeing the world, but you're hallucinating a view based on retinal image processing."

He grinned. "Intellectually you know it, but you don't really believe it, do you?"

Miffed, I replied. "You're sitting there looking at me and being amused. Consider this - the 'you' being amused is just the activity of 100 billion neurons doing their thing behind your eyes. 'You' are no more than the sum of their activity. How does that work, d'you think? And do you really believe it?""