0:00
In this fourth lesson, I'm going to try and
introduce you to another way of thinking about how these problems,
how this phenomenology, how these discrepancies between what we see and
what's out there physically can be explained.
And this is an empirical explanation that depends on
using our experience, not over just our lifetimes, but
our experience with information translated from the world
onto retinal images over evolutionary time, as well as our lifetimes,
to generate a method, a way of strategy of getting around the inverse problem.
So let me first of all remind you what the problem is.
We talked about this before in some detail.
But, let me tell you again, that the inverse problem refers to the fact that
letting the stimuli images on the retina entail an entanglement
of all of the information that comes from the real world.
The illumination, the reluctance of surfaces, transmittance are all entangled,
or what comes to the eye as a stimulus and falls on the retina.
So, we have no way of getting back logically from the stimulus to
the contribution of the illuminant, the surfaces that we need to understand
in terms of their properties, transmittance of the atmosphere.
All of these things are conflated in the retinal image,
and they get very hard for us to understand how
it is that we're able to succeed in the world through the ages of vision.
And my point in this exercise is to try and explain how we
can get around the inverse problem, and in the course of doing so,
understand why it is that patches look different from their physical reality and
our perceptual domain of seeing light and dark.
And the thing to keep in mind as I do this, is that basically we have
learned what the relationships are between the physical world and
are internal world of perceived information.
2:58
And you might think, well okay this is an unsolvable problem.
How are we going to know what our human experience over millions
of years of human evolution have been with natural images?
And how can we use that experience to try and
sort out the means by which we get around the inverse problem?
Well, it's not as complicated as you might think in the sense that the world
hasn't changed very much over the three or four million years of human evolution.
And we can of course get a large database of natural scenes
by simply taking photographs like these four examples.
But you can use thousands of examples, and sample them millions and
millions of times, to understand what our human experience has always been,
what have been the prevalences of images falling on
the retina from the nature of a natural world.
So one thing you have to be aware of right away is that it's
no good to think of images in their entirety like these.
This is of course what we are aware of seeing.
But we're talking about learning from experience.
And the fact is that for any image like this, any natural images like these
four examples or any other, you're, in effect, only going to see that image once.
Remember there are millions of photoreceptors on each of our retinas, and
we're only going to have had those photoreceptors activated in exactly
the same way once in relation to all these images like this, why is this simple?
it's because the detail in the activation mode of
photoreceptors via natural edge doesn't happen many times it happens only
once, effectively or conceivable very few times.
But to learn stuff you need to see things many times, whether you're
learning a language, or whether you're learning to play a musical instrument, or
anything that you learn you need to see and practice many, many times.
The characteristics that you're trying to embed in your memory stores.
Again, memory, we're not talking about individual memory.
We're talking about human species memory.
So it's no good to use images and
imagine that we are somehow learning from the image in its entirety.
We're learning about lightness and darkness and its relationship to
our perceptions in terms of little patches of light and dark.
And these are patches that are on the order of size of the receptive fields.
In the centre of vision, in the primary visual cortex,
remember we talked about those?
Those are fractions of a degree or maybe a degree or two, and
you take a small piece of images like that and sample them many millions of times,
of course you can get a pretty good idea of what the relationships
have been over the millenia of evolutionary time.
What the relationships have been between luminous relationships in the real
world and little patches of the real world that we see time and time again.
And how those might be related, could be related to our
experience based on that empirical evidence, so
what we're going to do is to try and explain this phenomenon.
This simplest of all the luminous discrepancies that we've talked about
in terms of experience derived from natural images.
And we're going to talk about the experience, in terms of sampling
natural images with little patches like this, many, many millions of times.
And extracting the gist of our human experience, from that evidence.