Recap on our data structure

At various times and places we have talked of a data structure which supports consciousness. See, for example, reference 6. This post updates and supersedes that at reference 1, of the srb series. We think the introduction of velocity (see below) sufficient reason to start with a new group search key, src.

The diagram included here makes use of what is sometimes called soft box modelling. Each box stands for some particular entity, some particular sort of thing in the world being modelled. For example cars, birds, colour or temperature. Boxes might also correspond to the tables of a relational database or to the worksheets of an Excel workbook. We might have, for example, a row in the car worksheet for each car in the pool. Or a row in the person worksheet for each person on the payroll. While an arrow linking one box to another stands for a one to many relationship, relationships which lie on a spectrum from parent-child to property.

Example of parent-child might be this father has many children, this car has many wheels or this team has many players. Examples of property might be this tyre is of this brand or this soup is that temperature. A relationship of this kind might easily be expressed by a column in one of our Excel worksheets. Or, looked at from the other end of the relationship, there are many soups in the world with this particular temperature.

Diagrams of this sort are a very useful tool for design and communication, but can get terribly complicated. They can become an industry. Here we have kept things simple by cutting corners. A cutting which associates to the fact that consciousness does not need to present a wonderfully complete and accurate picture of the world to the subject, even to the human subject, just a picture which is good enough. A picture which works most of the time, a picture which helps to keep us alive and reproducing. An arrangement which leaves room for consciousness to be tricked by the unusual, by that which has not been allowed for in the course of its evolution. See references 4 and 5 for one example among many.

The diagram describes the consciousness of one individual, the subject. A chunk of the world as seen (or heard) from his (or her) point of view. With the trickiness that a chunk may well include some part of the subject itself. A description which is rather biased towards vision, very important to most vertebrates and hypothesised to have kick-started the whole consciousness business, but which is intended to include other modalities.

So, the principal entities, working from left to right, are the scene through to the cell. With, for example, a frame being made up of a number of layers (parent-child) and a layer belonging to some particular frequency band (property).

We suppose consciousness to be divided into frames, with each frame lasting for the order of a second.

Using a metaphor from the cinema, frames are grouped into takes and takes are grouped into scenes. Very roughly speaking, the subject is in one place, in one situation for the duration of a scene. A take is an incident in, some chunk of that scene. While a frame is some usually short period of time during which the subject keeps some particular object in focus. There is an object of attention. Keeping an object in focus may well involve moving both head and eyes. Maybe even the body.

A scene carries a lot of baggage which may be good for all the frames in the scene. So if a scene involves a polar bear, it may be appropriate for the compiler to load up information from memory about polar bears. Similarly, a take carries a lot of baggage which may be good for all the frames in the take. While the boundaries of scenes and takes may at times seem a bit arbitrary, carrying baggage at both the scene and the take levels ought to make for efficiency savings.
Scenes, takes and scenes can be rather arbitrarily stopped by new input from either outside or from inside. Some new sensory input or some new (unconscious) computation.

Moving from time to space, frames are organised into layers, perhaps up to twenty of them, each corresponding to a frequency band of the oscillations of the activity of the underlying neurons.

Layers are mostly fixed, but some of them have velocity, usually a uniform speed and direction for the duration of the frame, but we also allow rotations and predictions of change more generally.

Using an analogy from the theatre, we might have an actor moving across a painted background – with either the actor or the background being the object of attention. The way this diagram has been drawn excludes there being two such objects. The object of attention will often have no velocity (in the sense that the word is being used here) and, other things being equal, less detail is available for objects on layers which have velocity than on those which do not.

The original idea was that frames were fixed. They were built by a compiler and then were more or less fixed for the duration. However, this did not seem altogether satisfactory and we are now allowing velocity because there is tension between the notion that a frame lasts for around a second and the fact that an object of interest may traverse a good chunk of the visual field in that sort of time. We allow for this fact to the extent that we can predict and display it. Translations easiest, rotations harder and other predictions could be tricky, though possible. With one example of this last being the back and forth movements of arms and legs in the context of what is mainly a translation, a person walking along the street.

The predicted frame getting out of step with the real world might be one reason for starting a new frame.

It is important to note that the brain is doing a lot of processing in real time with limited resources, which means that there is also tension between getting movement into the subjective experience and getting structure. What we actually get will be a compromise.

We will probably allow layers which have been deactivated, which do not contribute to consciousness, to the subjective experience, at the time in question, but that is not included in this diagram.

A layer will contain layer objects and layer objects are organised into contiguous groups of elements, with all the elements in a part taking the same pattern and all the parts in a layer object taking different patterns – it being the pattern which defines the part. Layer objects will be separated one from another by empty space, space where there is at most noise, certainly not elements with patterns. A layer object might correspond to something out in the real world, out in the field of vision, like a polar bear, with the parts of the layer object corresponding to the larger parts of the bear, but it might also correspond to something in some other modality than vision or something rather more abstract altogether.

An element is an instance of a pattern - but it is by no means true that all instances of a pattern are the same. We might, for example, have a pattern for colour which varies in hue across a part. We might think of a pattern as a box of some particular shape, with stuff inside – very much the stuff of reference 2.

We will probably allow colour patterns which are invisible, which hold things together but which do not figure in the visual experience. Probably also layers which are transparent – the alpha channel of people who make movies, possibly invented at reference 3.

We will allow links between parts on adjacent layers, but that is not included in this diagram. As a result of which, patterns function a bit like local identifiers, working with location to bind the various bits of an object in consciousness together. We suppose that we have enough different patterns to do this.

Cells are arranged in a two dimensional array, millions of them altogether, with each cell mapping onto a location on the underlying bit of cortical sheet, a location which we suppose to contain tens, if not hundreds, of neurons. An element is then made up of a number of contiguous cells,

The value of a cell is a non-negative integer, perhaps not more than 10 or so, describing the amplitude of the oscillations of the activity of the neurons underlying that cell at the frequency band and during the interval of time in question.

The frame is brought to life by high frequency oscillations of activation scanning its structure.
Our hypothesis is that consciousness amounts to the combination of the low frequency oscillations defining the data and the high frequency oscillations scanning that data.

Further work

We are working on an alternative way of defining patterns to that described at reference 2. The rectangles with soft centres which we have been talking about so far hold a lot of data, which is good, but do not like being moved very much and do not seem to sit too well with neural nets, which is bad.

We are working on what exactly it means to be scanning a moving structure, that is to say scanning a part of a layer object in a layer with non-zero velocity.

But the next post in this new series has the working title ‘In praise of the homunculus’.

References

Reference 1: http://psmv3.blogspot.co.uk/2017/03/a-new-start.html.

Reference 2: http://psmv3.blogspot.co.uk/2017/03/soft-centred-patterns.html.

Reference 3: Compositing Digital Images - Thomas Porter, Tom Duff – 1984.

Reference 4: https://en.wikipedia.org/wiki/Motion_aftereffect.

Reference 5: https://www.youtube.com/watch?v=oNhcpOIQCNs.

Reference 6: http://psmv3.blogspot.co.uk/2017/04/its-chips-life.html.

Groups search key: src.