Artificial intelligence and freedom of thought

A philosophy of science pamphlet

Jacob Hendrik Galjaard

    In ‘Brief Answers to the Big Questions’ (John Murray, 2018), Stephen Hawking connects the understanding of these questions on page 19 with (as he writes) the idea that we live in a universe governed by rational laws we can discover and understand through science. On page 21 we read that those laws govern us and the universe. On page 41/42 he writes: ‘We have already made tremendous progress in understanding the cosmos, but we do not yet have the complete picture. I like to think we are not far off.’

    Hawking’s view that we live in a universe governed by rational laws is frequently confirmed by science. Research reveals laws, that relate to existence in general as well as to ourselves and to life in all its other forms. That is reassuring. After all, those laws form the reliable bedrock of our existence. Science has accepted the challenge of identifying them. It succeeds, often to our advantage. However, the hypothesis that everything in the cosmos corresponds to rational laws is based on circular wishful thinking. After all, this hypothesis is a prerequisite for obtaining a complete picture of the cosmos, that includes ourselves.¹

    Related to this wishful thinking were the thoughts of the factory engineer Frederick Taylor (1856-1915) more than a century earlier. Although his work did not include the cosmos, it has had an enormous influence on our existence in the industrialized world. Taylor also wanted to get a grip on ‘reality’ through science. For him, that reality concerned the ways in which his workers operated. He looked for rational laws (rules, algorithms) that would make operations understandable and controllable for him. Of course, he derived those rules from the workers’ skill, knowledge and experience. After all, nothing else was available to him. He then reduced their previous freedom of how to work, to working within a framework of rules. In doing so, he made the operations measurable and controllable.

    Despite resistance, this Taylorism gained a foothold. For example, Charles Eugène Bedaux (1886-1944) and Frank Bunker Gilbreth (1869-1924) produced extensive files of time-standardized   methods, which are still used in various professions today. The development of computers (of all types) confirmed Taylorism not only in companies and organizations, but also in society as a whole. Increasingly, our actions are determined by rules. Think, for example, of internet banking, tax returns and the routing processes for all business telephone traffic. Humans and machines are gradually placed within a common framework of information-processing artifacts or informates.²

    In 1979 in a Dutch thesis the term informatizing (Dutch: informatisering) was introduced to this.³ Informatizing involves technological control that tries to reduce our actions and existenceto what is determined by rational (algorithmic, computable) rules for the processing of information.  Everything becomes known in this sense by disregarding non-rational definable properties and qualities of life, or by reducing them to fit within a rational defined framework. Thereis then nothing other than what can be rationally found within the aforementioned framework. This shows that informatizing is inherent in Hawking’s wishful thinking (see 2nd paragraph).

    In the meantime, within the term Artificial Intelligence (AI), thinking itself also became the subject of informatizing (see the preceding paragraph). With this, thinking was also brought within a framework of rules. The path along which this happens is initially the same as the path Taylor followed when he made the actions of his workers conform to rules.  As mentioned in the 3rd paragraph, he developed those rules using the existing reality, including the workers’ skills, knowledge and experience. Similarly, rules for thinking are derived from the skill, knowledge and experience for thinking games such as Chess and Go and from the skills. knowledge and experience generating written language.

    As for thinking games, it was in 1997 that Deep Blue, the chess program developed by IBM, defeated World Chess Champion Gary Kasparov. The program used the experiences (ways of playing) of other grandmasters taken from a manual describing opening moves written by them. For the rest of the game, the program had available a multitude of strategies.⁴ For Go, an even more difficult thinking game than chess, the unbeatable program AlphaGo Zero was allowed to gain experience by playing against itself some 29 million times. The Go-grandmaster conceded a challenge match in 2016.⁵

    That Deep Blue and AlphaGo Zero proved so successful is not surprising. After all, the  prerequisite for mastery in Chess and Go is what can be called the ultimate form of informatizing: reduction of thinking to lawful (computable) processing of information (see 5th paragraph). This condition is fulfilled by the nature of the Deep Blue and AlphaGo Zero programs. People however never quite succeed. That is why the aforementioned programs are our superior.

    As far as language is concerned, the situation is different. Here too, skill, knowledge and experience are an inexhaustible source for discovering how our thoughts follow the laws. An unprecedented amount of digital information is available for this purpose – think of all Facebook, Twitter and WhatsApp messages, not to mention all internet pages.⁶ Unimaginably extensive analysis of this data has provided the desired insight, however, this insight is only partial, unless it is assumed (as Hawking argues) that our thinking fully corresponds to rational laws, which we can identify and understand with science.

    This leads to a contradiction: how can thinking that is bound by rational laws perceive those same laws? This contradiction can only be solved by proposing that our thinking is actually free.  That is, scientific thinking and rational thinking in general are subject to our free thinking.

Following this line, it becomes clear that what has been deployed and is produced with ChatGPT actually amounts to informatizing.³ That is, non-computable qualities of language are ignored and/or brought within a framework of rules (see 5th paragraph).  The results seem good but they are limited by pre-determined rules. They can be obtained ever more easily and quickly. How do we avoid bowing our heads to AI not only when playing Chess or Go? How can we avoid giving up our freedom of thought?

    Apparently, we are faced with a choice between our relationship with science and technology and with ourselves. In an essay¹ I discuss this in more detail. I wrote it in response to Yuval Noah Harari’s closing question in his book Homo Deus:

Are organisms really just algorithms and is life really just data processing?

Dr.ir. J.H. Galjaard,  5 May 2023

¹ See Informatizing on https://informatizing.wordpress.com

² My own translation of a Dutch term from C.A. van Peursen et al., Information,1968, The Spectrum, p.136

³ J.H. Galjaard, Informatisering, 1979, PhD thesis, Delft University Press

⁴ Guido van der Knaap, From Aristotle to Algorithm, 2022, Boom Uitgevers, p.37

⁵   Ditto p.37

⁶ Ibid. p.74