On a Scientific Discipline (Once) Named AI

Wolfgang BibelIJCAI 2018 · DOI

TL;DR

This paper argues that Artificial Intelligence (AI) has fragmented from its original, broad vision of a fundamental scientific discipline studying "informational, psychological, and intellectual stuff" (IPsI-stuff). The authors propose that this splintering into numerous sub-disciplines has led to redundancy and missed opportunities, and they call for a re-engagement with AI's broader mission to address complex societal challenges.

What Problem Does This Paper Address?

The paper addresses the perceived fragmentation and loss of focus within the field of Artificial Intelligence (AI). Originally envisioned as a fundamental scientific discipline akin to Physics or Biology, AI's scope has narrowed considerably. The authors argue that successes in applied AI and the rise of competing disciplines have led to the original vision of AI being spread across fields like Neuroscience, Psychology, Cognitive Science, and Computer Science. This splintering, they contend, results in a lack of coherent research, redundancy, and a failure to address crucial, complex problems that require a unified approach to understanding "IPsI-stuff." The paper aims to analyze this situation and suggest ways to re-engage with AI's broader, original mission.

Methods

The paper primarily employs an analytical and philosophical approach. The authors:

  • Characterize the envisioned discipline: They define the subject matter of AI as "IPsI-stuff" (informational, psychological, and intellectual stuff) and distinguish two levels of inquiry: the lower (physical, chemical, physiological processes) and the higher (functional meaning or semantics).
  • Trace the historical evolution: They briefly outline the historical development of science and the emergence of IPsI-science, highlighting the Dartmouth Conference as a pivotal moment for AI.
  • Analyze the consequences of splintering: Through discussion of selected challenges, they illustrate how the fragmentation leads to disadvantages such as redundancy, lack of cross-disciplinary knowledge sharing, and missed opportunities.
  • Propose future directions: They suggest specific challenge areas that require AI expertise and could benefit from a re-engagement with the broader mission, such as understanding consciousness, morality, ethics, and fairness.

The paper does not present novel algorithms or empirical studies but rather a conceptual framework and a call to action for the AI community.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • AI's original vision was broader: The founders of AI envisioned a fundamental scientific discipline studying the nature of intelligence and cognition, not just a set of computational techniques.
  • Splintering is detrimental: The fragmentation of AI into numerous sub-disciplines has led to a loss of synergy, redundancy in research efforts, and a failure to tackle complex, overarching problems.
  • Two levels of inquiry are crucial: Understanding IPsI-stuff requires examining both the lower-level physical processes and the higher-level functional meaning, and AI is uniquely positioned to bridge these levels.
  • The "empathy gap" and societal problems: AI has the potential to address significant societal issues like interpersonal conflicts and the need for ethical guidelines for autonomous systems, but this requires a more unified and ambitious approach.
  • Consciousness as a scientific challenge: The mystery of consciousness, currently debated primarily in philosophy and neuroscience, could benefit from AI's computational and functional analysis approaches.
  • Morality, ethics, and fairness are AI's responsibility: As AI systems become more pervasive, the AI community must take a leading role in developing ethical frameworks and ensuring fairness in AI applications.

Key Results

The paper does not present quantitative results. Its key "results" are conceptual:

  • A clear articulation of the original, broad vision for AI as a scientific discipline.
  • An analysis of the disadvantages arising from AI's current fragmented state.
  • Identification of five key challenge areas (Functionality of Brain Processes, Hybrid Two-Level Techniques, Societal Problems, Consciousness, Morality/Ethics/Fairness) that highlight the need for a more unified AI discipline.
  • A persuasive argument for AI to re-engage with its broader mission to address fundamental scientific and societal questions.

Limitations and Future Work

Limitations:

  • The authors acknowledge that reversing the trend of splintering is difficult and that their article is primarily an analysis rather than a concrete plan.
  • They note that the current public perception of AI is often limited to narrow aspects like deep learning and robotics, hindering a broader engagement.
  • The paper highlights a lack of collaboration and sometimes friction between different AI subfields, which exacerbates the problem.

Future Work:

  • The paper strongly advocates for a re-engagement with AI's broader mission, suggesting that AI researchers should seize the current public attention to demonstrate insights into fundamental IPsI-science.
  • It calls for the development of ethical rules and guidelines for autonomous AI systems.
  • The authors suggest that AI could leverage its expertise to help solve societal problems like the "empathy gap" and legal disputes.
  • They propose further research into consciousness from a functional perspective, integrating AI methods.
  • The paper implicitly calls for greater interdisciplinary collaboration and potentially the creation of more encompassing international conferences and study programs focused on IPsI-science.

Summary updated

BibTeX
@inproceedings{bibel2018scientific,
  author = {Wolfgang Bibel},
  title = {{On a Scientific Discipline (Once) Named AI}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2018},
  doi = {10.24963/ijcai.2018/713},
}
Figure