[Paper Review] There is no Artificial General Intelligence

Jobst Landgrebe, Barry Smith|arXiv (Cornell University)|Jun 9, 2019

Topic Modeling68 references4 citations

TL;DR

This paper argues that Artificial General Intelligence (AGI) is mathematically impossible because no Turing machine can fully master the complexity of human dialogue, which requires understanding dynamic context, intentions, emotional cues, and evolving dialogue horizons. The authors conclude that since fluent, contextually adaptive dialogue is a necessary condition for AGI, and since this cannot be achieved by any computational system, AGI itself is unattainable.

ABSTRACT

The goal of creating Artificial General Intelligence (AGI) -- or in other words of creating Turing machines (modern computers) that can behave in a way that mimics human intelligence -- has occupied AI researchers ever since the idea of AI was first proposed. One common theme in these discussions is the thesis that the ability of a machine to conduct convincing dialogues with human beings can serve as at least a sufficient criterion of AGI. We argue that this very ability should be accepted also as a necessary condition of AGI, and we provide a description of the nature of human dialogue in particular and of human language in general against this background. We then argue that it is for mathematical reasons impossible to program a machine in such a way that it could master human dialogue behaviour in its full generality. This is (1) because there are no traditional explicitly designed mathematical models that could be used as a starting point for creating such programs; and (2) because even the sorts of automated models generated by using machine learning, which have been used successfully in areas such as machine translation, cannot be extended to cope with human dialogue. If this is so, then we can conclude that a Turing machine also cannot possess AGI, because it fails to fulfil a necessary condition thereof. At the same time, however, we acknowledge the potential of Turing machines to master dialogue behaviour in highly restricted contexts, where what is called ``narrow'' AI can still be of considerable utility.

Motivation & Objective

• To establish that the ability to engage in fluent, contextually adaptive human dialogue is a necessary condition for AGI.
• To analyze the structural and cognitive complexity of human dialogue, including context shifts, emotional cues, and pragmatic interpretation.
• To demonstrate mathematically that neither traditional programming nor machine learning models can achieve full mastery of human dialogue.
• To argue that since dialogue mastery is necessary for AGI, and since it is unattainable in machines, AGI itself is impossible.
• To distinguish between the potential of narrow AI in restricted domains and the fundamental impossibility of achieving general intelligence in machines.

Proposed method

• Analyzing the nature of human dialogue as a dynamic, context-sensitive, and emotionally layered interaction involving evolving dialogue horizons and mutual adaptation.
• Identifying key dialogue phenomena such as turn-taking disruptions, false starts, interruptions, and emotional modulation as core challenges for machine modeling.
• Examining the role of intention, deixis, and implicit meaning in dialogue interpretation, which depend on shared knowledge and biographical context.
• Arguing that no traditional mathematical models can serve as a foundation for modeling such dialogue complexity.
• Demonstrating that even modern machine learning models—successful in narrow tasks like translation—cannot be extended to handle the full generality of human dialogue.
• Using a homeostatic model of dialogue maintenance to show that human dialogue resilience relies on non-computable cognitive and emotional coordination.

Experimental results

Research questions

• RQ1 Is the ability to conduct fluent, contextually adaptive dialogue a necessary condition for Artificial General Intelligence?
• RQ2 Can traditional or machine learning-based models replicate the full generality of human dialogue, including emotional and contextual shifts?
• RQ3 What mathematical or structural constraints prevent a Turing machine from mastering human dialogue in its full complexity?
• RQ4 Why is the Turing Test’s core idea—dialogue as a measure of intelligence—insufficient as a criterion for AGI, despite its theoretical appeal?
• RQ5 To what extent can narrow AI systems simulate dialogue in restricted contexts, and why does this fall short of AGI?

Key findings

• The ability to conduct fluent, contextually adaptive dialogue is a necessary condition for AGI, as dialogue underpins human communication, culture, and practical intelligence.
• Human dialogue involves continuous, dynamic adaptation across multiple dimensions—emotional, intentional, contextual, and structural—making it intractable for any computational system.
• No traditional mathematical models exist that could serve as a foundation for programming a machine to master human dialogue in its full generality.
• Even advanced machine learning models, successful in narrow domains like translation, cannot be extended to handle the full complexity of open-ended human dialogue.
• The dynamic evolution of dialogue horizons, including reinterpretation of past utterances and real-time context shifts, cannot be captured by current computational architectures.
• While narrow AI can simulate dialogue in constrained contexts, the absence of a necessary condition—full dialogue mastery—renders AGI impossible in principle.