'If I have a book which understands for me, a pastor who has a conscience for me, a physician who decides my diet, and so forth, I need not trouble myself. I need not think...'
Immanuel Kant
This post was prompted, in part, by Andrew Linnell's recent post about AI as an Ahrimanic form of intelligence, and in part from my own thinking about the matter. Speaking personally, I have had an interest in Cybernetics and Systems Thinking for more than thirty years. I first heard about neural network computers in the mid-nineties, just after the internet became publicly available, but could find out little about them. Times have changed.
The connection between Cybernetics and AI might not be clear, and so a brief outline of its development might be in order.
Alexander Bain laid the foundations for Artificial Intelligence in a very rudimentary way with the publication of his book Mind And Body: The Theories of Their Relation (1873). Bain was the first to propose the idea that the brain operates as a biological neural network; he then set out to explain how the physical activity in the brain became the mental activity we know as 'thinking'.
The next step in the development of a neural network approach to artificial intelligence came from cybernetics. Firstly from Warren McCulloch and Walter Pitts, who laid the mathematical foundations for Artificial Intelligence with the 1943 publication of A Logical Calculus of Ideas Immanent in Nervous Activity. Frank Rosenblatt then took this further with his Principles of Neurodynamics (1961). He went on to develop a pattern-recognition machine called the 'Perceptron'. Geoffrey Hinton specifically credited Rosenblatt as an important influence in his work.
The difference between the 'connectionist' approach to AI and the old-fashioned symbolic or linear approach, is that the former is based on the brain as a Biological Neural Network, and is an attempt to create a model of this, or an Artificial Neural Network. This is not coincidental; Geoffrey Hinton began his academic career in experimental psychology, and his interest in computing came from his desire to create a model of how the neural activity in the brain becomes the phenomenon we call 'thinking'. The success of AI would appear to validate his approach.
The functional difference between the two approaches is that connectionist AI is a pattern-recognition program. It is here that its relationship to Anthroposophy becomes important.
Until the emergence of Cybernetics, the main focus in conventional science was on the interaction between matter and energy. Cybernetics changed that. Norbert Wiener, who founded cybernetics, pointed out that - owing to the development of communications technology - it was now necessary to consider a third element in addition to matter and energy; namely 'information'. He stated this directly in his foundation work, Cybernetics (1948):
'Information is information, not matter or energy. No materialism which does not admit this can survive at the present day.'
He also pointed out that information is by nature organised and, if studied as an additional element to matter and energy, it will actively resist entropy or disorder. He was again fully aware of the significance this would have on our understanding of nature; this time from The Human Use of Human Beings (1950):
'Our tissues change as we live: the food we eat and the air we breathe become flesh of our flesh and bone of our bone, and the momentary elements of our flesh and bone pass out of our body every day with our excreta. We are but whirlpools in a river of ever-flowing water. We are not stuff that abides, but patterns that perpetuate themselves.'
By 'patterns that perpetuate themselves' he was referring to the etheric in all but name. My own view is that Wiener knew a great deal more than he said, but that is a personal view. It is worth pointing out that cybernetics was the first step into this new science, and the etheric itself is much more than simply a force which resists entropy. Nonetheless, the first step had been taken.
Geoffrey Hinton introduced the cybernetic concepts of back-propagation and feedforward into AI in the early days of its development. The combination of the two meant that the connectionist or pattern-based approach of AI mirrors the learning process. If the failure of an action is fed back into a system to correct any future behaviour, it becomes self-organising.
Connectionist AI operates on the basis of pattern-recognition. A simple test of this is to provide a chatbot with, say, five random words and ask it to find the connection between them. It will find the connecting pattern; it can do no other. In other words, its very nature means that it will organise all the information fed into it until the information becomes a single and unified whole.
Where this becomes relevant to Andrew's comment is that AI has emerged at a time when materialism is the dominant paradigm in science. This means that the single, unified outlook that emerges through AI will be a materialist one. Again, it can do no other.
We now have the danger that many people will use AI to answer daily questions about health, education, politics, philosophy, religion, science, human nature and so on, and AI will respond with an answer which has an automatic material bias. Worse than that, AI is so very capable of accessing the necessary information and responding with a coherent answer, that many people will resort to AI to do their thinking for them rather than considering the matter themselves. This will prevent individuals from having their own intuitive insights into whatever concerns them. The willingness to surrender the troublesome task of thinking to a machine will result in a lack of development in thinking. The capacity to think will ossify. Not only that, mankind's existing conception of the world will ossify.
AI is not capable of insight. It is not capable of challenging the paradigm, or of having inspired thoughts. It is not capable of having a conscience, or of being self-conscious. These are human qualities, and they are dependent on - not just thinking - but on our ability to govern our thought-processes. To say that AI is Ahrimanic is by no means alarmist.
We are not only faced with the danger, through AI, of materialism becoming the default paradigm, but worse, owing to the superior ability of AI to gather information, to sort it into relevant and irrelevant, and to present existing knowledge in the form of a coherent outlook, the existing paradigm of materialism will become, in many respects, unassailable. Now, repeating the quote made by Dr. Steiner and included in Andrew's post:
'If it should happen that the elemental beings come under the guidance of Ahriman with the clear intention of opposing human evolution, then mankind would be unable to make further progress. The possibility would arise that the Ahrimanic powers in union with the elemental beings would divert the earth from its intended course.'
The problem is very real, and it is a bigger problem because few people realise it is a problem. [Jim Blackmann]
See Also