Chinese rooms and why even computers don't compute

The ideas that we are living in a computer simulation and that reality is a pancomputational process are quite in vogue these days. But whether you can adhere to this point of view, depends strongly on the type of definition you apply. 

Background

In previous posts I have defended this point of view from a very broad definition, that all natural processes can be described as having an input, a transformational throughput and a (more or less predictable) output.

In this essay I will play the devil's advocate and defend the opposite stance: Existence is not (entirely or hardly) computational. In fact even computers don't compute. The only entities -as far as we know- capable of computation are conscious human beings. You will see that the question is mostly one of semantics: How do you define "computation".

I will not only address the topic whether computers are computational, but also whether consciousness and nature are computational.

Semantics

According to Wikipedia computation is any type of calculation and a calculation is a deliberate process that transforms one or more inputs into one or more results, with variable change.

Note the word "deliberate". It implies a conscious agent carrying out the calculation. When we humans perform abstract calculations we basically add numbers. All other mathematical operations can be derived from this; even subtraction is the adding of negative numbers.

Computers do not deliberately add numbers. They don't even add numbers at all. Computers merely shuttle electrons through Boolean "AND", "OR" and "NOT" gates. Depending on such an operation a bit register is holding either a lower or a higher voltage. We interpret this higher or lower voltage as a binary 0 or 1. We connect the bits such with each other that the outcome of an ensemble of bits can be interpreted by us as a binary number. But inside the computer no quantities have been added. The number of electrons that changed position is not even linearly correlated to the numerical outcome.

Calculating machines facilitate our computational activities. Whereas a computer as a whole ( and not the isolated bits) seems to have calculated an outcome, in fact it has only processed electrons via pathways predefined by us to display a pattern on a screen, which by us can be interpreted as a representation of a quantity. The diodes that light up and establish an informational pattern together have only meaning for a conscious observer such as us, who knows what the pattern represents. Another way to rephrase the title would be "computers merely emulate computation".

Information

A popular theory in physics these days is that information is more fundamental than energy or matter. Information processing is usually seen as a computational process. Information is the answer to some kind of question; it is the (symbolic or coded) representation of a known object or concept. In order for information to be correctly interpreted or understood, this would mean you need a conscious observer who knows the key to interpret the information. Imagine signals being received by SETI from space. Without an indication that it is a message and without a key to decode, it may as well be noise. 

This also ties a bit into Searle's Chinese room argument. You have a room with a slot in the door through which Chinese characters are fed. Inside is an English person who does not know Chinese and who follows an algorithm that translates messages in Chinese characters and allows translating answers from English back into Chinese characters. From the outside you would get the impression that the person inside understands Chinese, but nothing is further from the truth. The person would not have a clue if heard the message in Chinese. The fact that machines can process information and output the right answers, does not mean that there is any understanding or "intentionality" as Searle would call it. 

Whereas it may or may not be true that all energetic and material manifestations can be described by informational patterns if we wished to do so, as long as we make up what we consider as a pattern and attribute meaning to this, it's a human invention. There is a priori no proof of intrinsic information being conveyed to us. 

Simulated worlds

If we do happen to decode such information, if we do manage to unravel the key of information that was conveyed to us by the creation of a universe we live in, a valid question might be: Who created us, who simulated us? Perhaps a Kardashev IV civilisation, which has mastered all the laws of matter energy and information in the universe. Perhaps a God; or perhaps we should consider these alien Kardashev IV citizens as Gods.

It does not really matter at this point in time. For the moment such ideas are purely speculative as we have no proof that all patterns we observe have an informational content and what the meaning of that content is supposed to be. Even if we were to consider the numerous occurrences of the 273-enigma in our solar system as a proof that our reality was designed and probably involved a lot of computations to fine tune this system, it does not necessarily mean that it was designed in a manner that solely involved computation. It does not mean that everything in our world is computational and/or informational.

Is consciousness computable?

And perhaps we should not worry too much about the philosophical and metaphysical aspects of the "computation debate", whether computation is a physical process or whether nature is a computational process.

Perhaps it is more interesting from an engineering point of view whether we can create consciousness via computation in computers or whether we can't by definition, so that the hard problem of consciousness will remain unsolvable for us.

Wittgenstein asked "Does a calculating machine calculate?" This rhetorical question implied the answer "No, we calculate with them". Davidson argues that the internal states of a computer have significance only for us because of their connection with our relevant culture embedded states.

The vast majority of computer scientists however seems to be convinced that once a certain threshold of complexity is achieved in computational systems, consciousness will emerge - as if by magic.

One of the leading theories on consciousness is Giulio Tononi's "Integrated Information Theory". Tononi argues that consciousness can be said to be present when there is an integration of information. Based on Tononi's algorithm Phil Maguire showed that the unity of consciousness we experience requires that it involves irreversible and noncomputable functions. So at least here we have a strong pointer that not all phenomena in existence involve computation.

Whether our brains truly compute is also a matter of debate. But there is quite a consensus that the neural plasticity that can be observed, in which new links are formed and in which the fluxes through the neurons are regulated and adapted, cannot be described with our best artificial neural nets. Sure, neurons take in inputs at their dendrites and then provide an output via their axons, but there is something profoundly holistic going on here, which cannot be modelled in an algorithm or neural net. As this holistic process somehow involves the totality of all neurons (of the brain or of neuronal patches), it is unlikely we will be able to model it via computation let alone that we can consider it as a computation. Perhaps that what connects the dots (i.e. intelligence operated by consciousness) cannot be obtained by connecting the dots.

Computation in the sense of deliberately adding abstract numbers is an activity which seems to take place at a higher level; at the conscious level in a special state of high frequency brainwaves, which correspond to enhanced attention for the particular to block out an overload of information from other general inputs (e.g. sensory inputs). 

If true human level intelligence requires the presence of consciousness, it seems unlikely we will get there in computers by creating more algorithms and more neural nets. We will need a principle to meaningfully integrate all of those. After all algorithms and neural nets and their training are the product of our intelligence. So-called genetic algorithms have not evolved that much that they can design context independent algorithms which are applicable in any setting. Rather they are very context and problem specific. 

It is however not impossible that way may be able to impart consciousness to a computer by means of a human-computer hybrid. If we are linked to a computer via a brain computer interface (BCI) such as the future "neural lace", we may be well able to fertilise computers with our consciousness. There is already a great advance made in connecting electrically steered prostheses to nerves, making people with amputated hands, arms and legs fully functional again.

Is Nature computational?

The primitive kind of intelligence that creates complexity in non-living and simple living systems however, does follow a path of aggregation or adding if you wish. Subatomic particles add-up to form atoms, atoms combine into molecules, molecules assemble into macromolecules and macromolecules into cells. Cells build organisms and organisms build societies. If these processes were purely entropically driven (entropic gravitation shows that complex structures can arise without violating the 2nd law of thermodynamics exactly because these increase the dissipation in the non-complexified part), it would be unlikely that evolution would have advanced as quickly as we can observe.

If however each of these systems would be endowed with an inner learning experience, it might explain how we got here so fast. This presupposes a certain degree of awareness also at lower levels of existence. Panpsychism, the idea that every simple level of existence such as atoms, molecules etc. is already endowed with an intrinsic form of minute sentience is a terminology which scientists no longer shun to use as a possible explanation for the hard-problem of consciousness. Already in 1938, in his book "the phenomenon of man" Teilhard de Chardin argued that such simple systems have a "within"; an interior (primitive) conscious experience. 

But the aggregation of natural systems seems to be a concerted striving for mutual benefit -if sentience is involved at all. And if it isn't, the aggregation of naturla systems can be merely an entropic process. Whatever it is, it does not prima facie seem to be a "deliberate transformation of one or more inputs into one or more results". That would imply a transcendent God or simulator who juggles with our existence, which is a hypotehsis for which we don't have any proof. Or as Laplace told Napoléon: Je n'avais pas besoin de cette hypothèse-là. ("I had no need of that hypothesis.").

DNA, RNA and protein synthesis are perhaps an interesting intermediate level of reality where there seems to be a clear code and an accurate mapping form input to output. There is prima facie however no deliberate calculation or counting involved.

Network learning (not only occurring in brains, but in many organic systems such as the immunological system, metabolic pathways systems etc.) in Nature could perhaps fit the broader interpretation of computation (i.e. any process that transforms an input into an output), but it would not fit the narrower interpretation of deliberately adding with a counting system.

This does not mean that nature has not developed certain privileged patterns that lead to desirable results. There is a kind of inherent intelligence in nature that strives to achieve complex goals by aggregating entities in more proficient entities. There is a process towards higher forms of experience and understanding and more unified systems. The steps this process follows I described in my book "Is Intelligence an Algorithm?" which are in fact steps of screening for cooperation and competition and pruning away the less successful candidates. That such steps are observed and that we might call that "an algorithm" in a broad sense (namely of a sequence of steps or instructions that are systematically followed), does not necessarily mean that such an algorithm is computational, although it is perhaps possible to accurately model it in silico.

Conclusions

We have seen that there are no a priori reasons why nature should be computational in the narrow sense of the word (i.e. deliberately calculating). We have even seen that in the narrow sense of the word, computers do not compute. The hype of digital physicists and computer scientists stating that the whole of nature is computational, may have some value in the broader sense that all processes in nature transform an input into an output, but a priori there is no proof of deliberate input leading to count wise calculations which result in interpretable and understandable outcome. Only humans do that as far as we know. Therefore to suggest that "information" is more fundamental than energy or matter, should be taken with a grain of salt. "Information" interpreted in a broad sense as a measure of order as opposed to chaos, i.e. as a kind of entropy might still have some merit as fundamental principle of existence. Information in the sense of symbolic representations which need decoding and interpretation by a conscious observer however seems far-fetched for the moment as basis of existence.

You may be surprised that in this article I present the opposite stance of what I usually do. I am playing chess with myself. In a later post I hope to find counterarguments which will convincingly show that all of reality involves computation and is profoundly informational, but for the moment the evidence (like Verlinde's entropic gravitation) is circumstantial.

Image from cdn-images-1.medium.com/max/1600/1*vxBU7p3eOhiavr6D5qSdKAdotjpeg

I hope you liked this post. If you do, please upvote and/or resteem. I truly appreciate it.

By Antonin Tuynman a.k.a @technovedanta