HIDDEN COSTS OF DEEP LEARNING
To understand intelligence, we have to take a step back and see how things manifest in this world. In order for something to manifest, there needs to be some kind of space where the manifestations appear. Then in that space, needs to appear some kind of movement. Take for example atomic particles, which appear in a space where they move closer to each other, or further apart from each other. Sometimes scientists refer to this process as “clumping” or more commonly “self-organization”. Is this intelligence?
The fundamental question of intelligence, is “what is intelligence?”. Where do we find intelligence, and how can we precisely describe the way it manifests. In the modern days, we are also very much interested with the question of how to manufacture it, how to create intelligent systems. This leads us right back to the basic facts of space and movement; in order for intelligence to manifest, there needs to be space and in that space there is movement. At this point it’s important to consider a subtlety; space does not imply a confined space, or a thermodynamic isolated system, i.e space does not constitute a physical constraint. More so a potency of some sort, a quality rather than a spatiality.
In fact, if we try to find intelligence that does not move, in such analysis it quickly becomes obvious that if something is static and void of any movement or change of any kind, it does not meet the requirements of functioning. Because there is no function, then clearly there is no intelligence. A function is in some way the output of intelligence, the gross characteristic of it. This way we can find that consequently movement is the basis for what we refer to as intelligence. In this fundamental dichtomy, space is the quality upon which movement is possible. This is obvious, as without such a quality as space, no movement can take place.
When you contemplate how things manifest, from the atomic spin to the nebula, and everything in between, every organism, every so-called inanimate thing, every particle, you find that the fundamental relativity is the one of space and movement. All appears in a space due to causes and conditions, and all that manifest and that we can observe, is a result of things coming closer together, or going further apart within that quality of space. While this process is not so clear for us with say bedrock for an example, we can find it to behave in the exact same way. It too continuously change, even though at a scale (in terms of time and magnitude) that is not obvious to us.
The Problem of Absolute Zero
When it comes to movement, the scientists have found that it can not be stopped. This is referred to as the absolute zero i.e. the temperature where atoms stop moving. In a causal system this limit makes sense, and the opposite i.e. attaining such a state does not make sense, as once something stops changing, it breaks the causal chain and nothing ever happens after that. Causality is dependent on constant change, constant functioning. When functioning ends, it ends and there are no causes and conditions for anything to manifest after that. Everything that happens, happens because there is some kind of cause and condition that make it happen. Both randomness and determinism are 100% dependent on change (movement). In the case of absolute zero, we are using the relativity of temperature to measure movement. In fact, movement and temperature are interchangeable in terms of meaning. Temperature is simply a relative measurement of the degree of movement.
Where does this leave us in terms of intelligence? We might for example ask if a wall is intelligent, and if there is intelligence manifesting in a wall. Without the frame of reference we have just briefly gone through, this would be a very confusing questions. Indeed, most “experts” working in the field of machine intelligence and even modern philosophy, are oblivious to this type of idea about intelligence. Most will not hesitate in answering “no, wall is not a manifestation of intelligence”. Actually, the kind of fundamental concept of intelligence we are discussing now, clearly shows that wall is intelligent, because it is a manifestation of movement in space. If we want to impose more relativities, beyond just movement, we can find that in fact the wall kind of intelligence is superior to human intelligence in many ways. To understand this, we have to remember that the idea of intelligence has to do with functioning. If walls were made of humans, then houses would have a big problem staying together because of the way human intelligence manifest. In respect to the kind of intelligence wall represents, it is very predictable, and its physical attributes are very rigid
This type of contemplation on the topic of intelligence gives us far more room in terms of exploration when it comes to asking and answering the fundamental questions underpinning machine intelligence. For example, we come to understand that the comparison between the intelligence of a man and the intelligence of a machine to a far extent play out as the comparison between man and the wall. This type of approach grants us quite a bit of flexibility, in contrast to say models of intelligence that focus on sentience or cognitive functions as the basis for discussion on intelligence. That type of approach is very much coming from the human perspective, and creates unnecessary limits for an era characterized by co-operation of man and intelligent machines.
Greater than Human Intelligence
Some fear, that the machine will become more intelligent than us humans, showing a profound misunderstanding of both intelligence generally, and machine kind of intelligence more specifically. Such a view is driven by confusion and fear. Actually the machine is already able to perform many kinds of functions (intelligence) far superior to us human. That seems to be the whole point; having our own abilities augmented by the abilities of the machine. Not augmented by the means of physical augmentation, the hallmark of the Cyberpunk genre, but by the kind of invisible connection we already to some extent have with machines. This concept is partially explained in the seminal 1960 paper by JCR Licklider, Man-Machine Symbiosis [read here]. Whereas machine intelligence promises to allow us humans embrace our human intelligence more, namely creativity and empathy, there are no groundings for suggesting that machine alone could have greater than human intelligence. This is simply proven using a line of reasoning; human intelligence is associated with the causes and conditions that result in humans and the functions thereof, whereas machine intelligence is not. Like it is impossible for a banana to grow from an apple tree, it is impossible for human intelligence to appear in a machine. This type of philosophical precision is very important, as otherwise we can just say anything (and in have people believe it). As the great philosopher Ludwig Wittgenstein said in Tractatus Logico-Philosophicus:
What can be said at all can be said clearly, and what we cannot talk about we must pass over in silence.
Indeed there is another part of this matter, a practical one where some actual risks do exist and a serious conversation needs to be had “before its too late”. Because of the causal relationship between man and machine, namely the facts that machine intelligence is and always will be underpinned by its origin (being an invention of humans), and that human decisions are heavily influenced by inputs from machines, the risks are quite clear. The most significant threats mostly associate with the decisions we make regarding investment in to rerflatively simple systems that lack any autonomy, but provide inputs for humans to make decisions. Those, and the way machine intelligence based decision making support systems are programmed, and the pressures “to deliver” at the level of individual programmers. The most important point to understand in this regard, is that the greatest risk with machine intelligence era is not associated with the machine becoming too intelligence, but in us humans remaining ignorant.