Skip to main content

Artificial intelligence and its natural limits


An argument with roots in ancient Greek philosophy claims that only humans are capable of a certain class of thought termed conceptual, as opposed to perceptual thought, which is common to humans, the higher animals, and some machines. We outline the most detailed modern version of this argument due to Mortimer Adler, who in the 1960s argued for the uniqueness of the human power of conceptual thought. He also admitted that if conceptual thought were ever manifested by machines, such an achievement would contradict his conclusion. We revisit Adler’s criterion in the light of the past five decades of artificial-intelligence (AI) research, and refine it in view of the classical definitions of perceptual and conceptual thought. We then examine two well-publicized examples of creative works (prose and art) produced by AI systems and show that evidence for conceptual thought appears to be lacking in them. Although clearer evidence for conceptual thought on the part of AI systems may arise in the near future, especially if the global neuronal workspace theory of consciousness prevails over its rival, integrated information theory, the question of whether AI systems can engage in conceptual thought appears to be still open.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2


  1. As a justification of our attribution of the roots of the immateriality of the intellect to Aristotle, we should note here that the traditional division of Aristotle’s De Anima into three books reflects the fact that Aristotle devoted the arguments of the second book to perceptual functions and those of the third book to intellective functions (concept formation, judgment formation and reasoning), among which he also provides an argument for the immateriality of the intellect on account of its ability to represent all material natures. For a thorough discussion of that argument in Aquinas’ interpretation, see (Klima and Hall 2011), pp. 25-59. To be sure, there have been materialistic interpretations of Aristotle’s doctrine of human thought even in ancient and medieval times (by Alexander, Averroës and the Latin Averroists), but even those strictly distinguished sensitive (perceptual) from intellective (conceptual) functions, attributing the latter to a separate, immaterial substance, and arguing that (apparent) human intellective functions are due to the material human soul’s special “uplink” to that substance (often identified with God). In any case, in developing the argument of the paper we explicitly relied only on Aristotle’s Thomistic interpretation. This is only a matter of Aristotelian scholarship, not affecting the substance of our argument.


Download references


We thank the reviewers Massimo Negrotti and especially Albert Borgmann for their helpful reviews of an earlier draft of this paper.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Karl D. Stephan.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Stephan, K.D., Klima, G. Artificial intelligence and its natural limits. AI & Soc 36, 9–18 (2021).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:


  • Artificial intelligence
  • Consciousness
  • Conceptual thought
  • Mind
  • Perceptual thought
  • Turing test