Mid Sweden University

Some studies have presented early traces of Artificial Intelligence (AI) in ancient Greece such as in the tales of Homer. However, if AI should involve creating and not only imagination, the consensus seems to be that the history of AI starts with the ideas presented by the Catalan philosopher, theologian and poet Ramon Llull. In his Ars generalis ultima (The Ultimate General Art) the main was to create new content and knowledge based on pre-defined concepts, a system that he completed in 1308. Another reason for starting the history of AI with Ramon Llull is that his Ars generalis ultima inspired Gottfried Wilhelm Leibniz to the system he presented 1666 in his dissertation On the Combinatorial Art. At the age of 20, Leibniz had created a theory for how to generate knowledge based on a rule-based combination of symbols. Leibniz’s system was built around the main idea that all human thoughts, no matter how complex, are combinations of basic and fundamental concepts, resembling how sentences combine words, and that the words are combinations of letters.

Leibniz who was a multi-talented mathematician, scientist philosopher received a lot of respect from his contemporary colleagues. However, there were also some criticisms with parallels to the critique of GenAI chatbots today. The current chatbots have been called stochastic parrots (Bender et al., 2021), indicating that the grammatically correct language they generate lacks a deeper understanding of the processed language. The strongest critique of Leibniz’s rule-based combinatorial system was the contemporary author, Jonathan Swift. In his well-known book Gulliver’s Travels, the Leibniz system is called 'The Knowledge Engine' (Rogers, 2017), and something that Gulliver experiences when the visits the Grand Academy of Lagado. An academy where the teacher claimed that his student could write books in any subject without the least assistance from genius or study. Swift's critique could be summarised as that a machine with a set of rules is not enough to handle an ambiguous entity as a context sensitive language. Today when different AI systems outplay humans in chess and excel in pattern recognition, language might be the challenge that remains. 

The idea for further research is to investigate the AI concepts that can be found in the time span after Leibniz, and before Alan Turing's computerised modern AI. Two interesting persons in the history of AI to investigate would be Baruch Spinoza and Ada Lovelace, but there are of course more persons and concepts to explore. Interesting also to look at the early attempts of computer-based chatbots such as the Eliza system (Berry, 2023), and how to pass the Turing test without a deeper understanding of the used language. Finally, in the contemporary development of generative AI it seems relevant to involve the contributions of the Nobel Prize laureates Geoffrey Hinton and Demis Hassabis.