[LINK] Machine Learning Was: Re: Robot cars and the fear gap

[David Lochrin]

On 2016-07-28 10:52 JanW  wrote:

> LOL doubt it.
>  
> Rods and cones (something like five different types I think I read at last count) respond to specific wavelengths within the visual range. The brain constructs the combinations to provide the visual image we perceive. Visual perception is an incredibly complex process, involving the visual cortex, the limbic system for emotional reaction and autonomic responses (fight/flight) and the language and pre-frontal cortex to provide the meaning interpretations to behave beyond the reactionary level. Most animals don't have the last two to any degree that we can recognise. "Colour" is a word, just like we use words to describe sound, e.g. high, low, bass, treble, rumble, piercing, screech.

Quite so, it's actually even more complex than that because different aspects of the basic image, such as colour, brightness, form, movement, etc., are processed in different parts of the brain but it all somehow comes together so we _perceive_ a coherent image.  Without thinking much about it, most people assume the brain reproduces the image internally on a sort of mental TV screen, but that just leads to an infinite regress - who or what looks at the mental TV?

But to return to the thought experiment...  We build a brain analogue from a big box of electronic bits, integrated circuits and so on, and equip it with camera-eyes.  The signal from each pixel results in pulses down a wire into this brain where the pulse rate signals intensity.  And the operation of this electronic neural network involves electronic pulses of varying pulse rate running furiously around its circuits.

Now if we assume everything runs in accordance with physics, what would we expect to see?  Lots of electronic activity, certainly.  But perception?  By what mechanism could this device possibly perceive green grass, blue sky, and a red fire engine?

David L.