Intelligence Made Simple: Part 2
Please see also Part 1
Are there other intelligences in the cosmos?
How would we know?
By what universal standard can we know intelligence when we see it?
A Universal Definition of Intelligence.
This is what I call 'square one' analysis. Putting aside all unanswerable questions about metaphysical entities, let us look at the basic building blocks of, not life as we know it, but the chemistry of the cosmos.
We know how elements combine to form molecules. We know that the electrostatic fields around molecules cause them to take on particular shapes in three dimensions. We also know that there are particular zones on some molecules that fit like jigsaw pieces to the mirror zones on other molecules. There are yet other molecules which, by linking temporarily to the two joined molecules can separate them. From this, you can build an intelligent device.
Intelligence is the manipulation of information in some form. If we treat the attachment of molecules one to another as a means of storing information, a 'memory', then we have the basis of a Turing machine. The Turing machine can be shown to be capable, given enough time, of computing anything that is computable. The digital computer is a large-scale implementation of the basic Turing machine model.
A Turing machine, in order to function, must be capable of discriminating between two states. These may be voltages, chemical concentrations or holes in a strip of paper. Let us consider voltages as the easiest thing to intuit. We could use zero volts as one reference, and one volt as another. In practice, due to errors of measurement, it is more convenient to have a lower cutoff slightly above zero, say 0.1V. Let us call the 1.0V level A and the 0.1V level B.
In order to determine that a voltage is greater than A, or less than B, there must be reference values to compare with. A reference value is the smallest possible model of memory. In biological entities there are detectors which discriminate e.g. food from poison, light from dark, up from down. All of these mechanisms can be viewed as comparators. In every comparator, the reference item can be viewed as a simple memory device.
The Turing machine, in order to operate, must be able to determine that a value is either >=A OR <=B
Values between A and B are indeterminate, and are ignored. Alternatively, we can take the value between A and B and call it C, in which case we ignore all values above A or lower than B.
The fundamental and universal mechanism behind intelligence is the ability to:
determine that A is not B, by comparing input with memory, ignoring the range of possible values A - B
determine that C is not A, by comparing input with memory, ignoring all values less than A
determine that C is not B, by comparing input with memory, ignoring all values greater than B
determine that C is neither A nor B, by making two tests.
Armed with the basics, we can move from intelligence in the abstract to the notion of 'an' intelligence.
This is what I call the ignorance model of intelligence. It is entirely divorced from anthropic use of language. It is, I suggest, of universal application.
Please see also Part 1
Are there other intelligences in the cosmos?
How would we know?
By what universal standard can we know intelligence when we see it?
A Universal Definition of Intelligence.
This is what I call 'square one' analysis. Putting aside all unanswerable questions about metaphysical entities, let us look at the basic building blocks of, not life as we know it, but the chemistry of the cosmos.
We know how elements combine to form molecules. We know that the electrostatic fields around molecules cause them to take on particular shapes in three dimensions. We also know that there are particular zones on some molecules that fit like jigsaw pieces to the mirror zones on other molecules. There are yet other molecules which, by linking temporarily to the two joined molecules can separate them. From this, you can build an intelligent device.
Intelligence is the manipulation of information in some form. If we treat the attachment of molecules one to another as a means of storing information, a 'memory', then we have the basis of a Turing machine. The Turing machine can be shown to be capable, given enough time, of computing anything that is computable. The digital computer is a large-scale implementation of the basic Turing machine model.
A Turing machine, in order to function, must be capable of discriminating between two states. These may be voltages, chemical concentrations or holes in a strip of paper. Let us consider voltages as the easiest thing to intuit. We could use zero volts as one reference, and one volt as another. In practice, due to errors of measurement, it is more convenient to have a lower cutoff slightly above zero, say 0.1V. Let us call the 1.0V level A and the 0.1V level B.
In order to determine that a voltage is greater than A, or less than B, there must be reference values to compare with. A reference value is the smallest possible model of memory. In biological entities there are detectors which discriminate e.g. food from poison, light from dark, up from down. All of these mechanisms can be viewed as comparators. In every comparator, the reference item can be viewed as a simple memory device.
The Turing machine, in order to operate, must be able to determine that a value is either >=A OR <=B
Values between A and B are indeterminate, and are ignored. Alternatively, we can take the value between A and B and call it C, in which case we ignore all values above A or lower than B.
The fundamental and universal mechanism behind intelligence is the ability to:
determine that A is not B, by comparing input with memory, ignoring the range of possible values A - B
determine that C is not A, by comparing input with memory, ignoring all values less than A
determine that C is not B, by comparing input with memory, ignoring all values greater than B
determine that C is neither A nor B, by making two tests.
Armed with the basics, we can move from intelligence in the abstract to the notion of 'an' intelligence.
An intelligent entity is any entity which, by ignoring irrelevant aspects of its environment, can abstract information from that environment, and use the information so abstracted to promote and prolong its own structure and status in that environment.
This is what I call the ignorance model of intelligence. It is entirely divorced from anthropic use of language. It is, I suggest, of universal application.
It seems like distinguishing btwn A, B, & C is more rooted in logic than intelligence. But by extension we know that pure logic produces a poor & unrealistic model of the world. Further it seems like the vast differences btwn human intelligence & a Turing machine attest to the latter's inadequacy of representing intelligence. Unless perhaps we just haven't seen a Turing machine of adequate complexity.
In this matter, I think we can learn a valuable lesson from the creationists. Theology, after all, is built on a very carefully thought-out structure. It combines rigid logic with assumptions about the origin of the universe that were widely accepted for centuries. But those latter assumptions are only partly responsible for why it went wrong. In some extreme cases, theologians was built such ornate structures that their complexity & rigidity would rival our modern statistical models. But in hindsight, it was similar to a social scientist fishing in the data for statistical significance: the structures were merely means to an end of supporting one's own opinion.
I guess I would advise caution in starting from such seemingly incontestable logic. But admittedly it's easier to criticize than to build, & off the top of my head I've nothing to offer in it's place.