The symbol grounding problem 75 years after Turing's Test (why computational success still leaves meaning unexplained)

The symbol grounding problem 75 years after Turing's Test (why computational success still leaves meaning unexplained)

by David Strohmaier

Humanities, Social Sciences & Thought Computing & Cybersecurity UComputing and Information TechnologyUYComputer science
🎙 David Strohmaier 👥 305 📅 January 15, 2026 ⏱ 83 min 👁 95 📄 expert opinion 🧭 2026-08-16
Available in: English (current) Français

Keywords

symbol groundingLLMmeaningTuring testsemantics

Summary

In this seminar talk, David Strohmaier addresses the symbol grounding problem 75 years after Turing’s seminal paper. He distinguishes between an ’easy’ grounding problem, which concerns the engineering challenge of building systems that can process and integrate multimodal data to replicate human cognitive abilities (T3 in Harnad’s hierarchy), and a ‘hard’ grounding problem, which concerns whether a system’s representations and outputs carry intrinsic meaning. Strohmaier argues that computer science and cognitive science do not need to worry about the hard problem, because they are justified in not worrying about whether the physical systems they use are ‘actual’ computers. He draws on historical sources, including Turing’s paper and Harnad’s 1990 work, and discusses contemporary debates about LLMs, referencing philosophers like Müller and others. The talk bridges computational linguistics and philosophy of mind, questioning whether statistical learning can ever capture referential and intentional aspects of meaning. Strohmaier’s final argument is that the easy problem is the relevant one for scientists, and the hard problem is a philosophical concern that does not affect practical research.

172 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides a valuable conceptual clarification by distinguishing between easy and hard grounding problems, which is useful for researchers in AI and cognitive science. The argumentation is philosophically informed and draws on historical and contemporary sources. However, the argument that scientists need not worry about the hard problem is presented as a ’naive’ argument and may not convince philosophers who see the hard problem as central. The talk is more about framing the debate than providing empirical evidence.

Scientific Rigor, Source Quality, Title Accuracy

The talk references key works: Turing (1950), Harnad (1990), Müller (2011), and others. The sources are relevant and appropriately cited. The title accurately reflects the content. The talk is a seminar presentation, so it is not peer-reviewed, but the speaker’s expertise is evident. The discussion is rigorous in its conceptual analysis, though some claims are presented without detailed justification.

153 words

Title / Content Match

The title accurately reflects the content, which examines the symbol grounding problem in the context of Turing's test and modern LLMs, focusing on the distinction between easy and hard grounding problems.

Quality & Reliability

7/10

The talk is an expert opinion by a researcher with a strong interdisciplinary background in philosophy and computer science. It references key literature (Harnad 1990, Turing 1950, Müller 2011, etc.) and provides a clear conceptual framework. However, it is a seminar talk, not peer-reviewed, and the arguments are presented in a conversational style with some informal asides.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • Preference Change — Book by Strohmaier and Messerli (2024), mentioned in biography.
  • Contrafactives and Learnability: An Experiment with Propositional Constants — Paper by Strohmaier and Wimmer (2023), mentioned in references.
  • A Category Theory Framework for Sense Systems — Paper by Strohmaier and Tyen (2022), mentioned in references.
  • Organisations as Computing Systems — Paper by Strohmaier (2021), mentioned in references.
  • The symbol grounding problem — Harnad's 1990 paper, central to the talk.
  • Computing Machinery and Intelligence — Turing's 1950 paper, discussed in the talk.
  • The easy and hard problems of symbol grounding — Müller's 2011 paper, mentioned in the talk.

Concurring Sources

  • The symbol grounding problem — Harnad's paper supports the distinction between easy and hard grounding.
  • The easy and hard problems of symbol grounding — Müller's paper also distinguishes easy and hard grounding, aligning with Strohmaier's framework.

Dissenting Sources

  • The Vector Grounding Problem — Mollo and Millière argue that LLMs face a hard grounding problem, contrary to Strohmaier's claim that scientists need not worry about it.

Contribution & Novelties

The talk offers a clear distinction between easy and hard grounding problems, applying it to LLMs. It argues that scientists can focus on the easy problem without worrying about the hard problem, a position that is not universally accepted. The talk bridges philosophy and computer science, providing a framework for future discussions.

Pour aller plus loin :

  • The symbol grounding problem — Harnad’s original paper, foundational for the discussion.
  • Computing Machinery and Intelligence — Turing’s paper, the historical anchor.
  • The easy and hard problems of symbol grounding — Müller’s distinction, directly relevant.
  • Vector grounding problem — Paper by Mollo and Millière, discussed in the talk.
  • The Octopus Test — Paper by Bender and Koller, mentioned in the talk.

118 words

Radar Profile

The radar profile shows high scores in information quantity and quality, reflecting the talk's rich conceptual content. The technical level is moderate, suitable for an academic audience. The global reliability is good, but the argumentative nature and lack of empirical data slightly lower the score.

Reliability 7/10