AI as artificial ignorance

Bent  Flyvbjerg

doi:10.1016/j.plas.2025.100208

AI as artificial ignorance

Bent Flyvbjerg

Research output: Journal Article or Conference Article in Journal › Journal article › Communication

Abstract

First, we present a number of simple tests of AI, which document a profound gap between the hype and the reality of AI. Second, we explain the gap in terms of a confusion of artificial general intelligence with generative artificial intelligence in the promotion of AI. Finally, we analyze AI as bullshit (in the strong philosophical sense of Harry Frankfurt). We find that AI and bullshit are similar in the sense that both prioritize rhetoric over truth. They mix true, false, and ambiguous statements in ways that make it difficult to distinguish which is which. AI sounds convincing even when it's wrong. As such, current AI is more about persuasion than about truth. This is a problem because it means AI produces faulty and ignorant results. For now, we need to be highly skeptical of AI for its lack of a concept of truth.

Original language	English
Journal	Project Leadership and Society
Volume	6
Pages (from-to)	1-3
Number of pages	3
ISSN	2666-7215
DOIs	https://doi.org/10.1016/j.plas.2025.100208
Publication status	Published - Dec 2025

Keywords

AI
Artificial intelligence
Artificial ignorance
Large language models (LLMs)
General artificial intelligence (GAI)
ChatGPT
Perplexity
Bullshit

Access to Document

10.1016/j.plas.2025.100208Licence: CC BY-NC

https://www.sciencedirect.com/science/article/pii/S266672152500033XLicence: CC BY-NC

Villum Kann Rasmussen Professorship
Schmidt, J. (PI), Flyvbjerg, B. (PI), Madsen, C. Ø. (CoI), Pfeiffer, H. (CoI), Lithén, J. C. L. (Admin), Gandrup, N. (Collaborator), Petersson, C. (Collaborator), Aaen, J. (CoI), Ouriques, R. A. B. (CoI) & Ajazi, F. (CoI)
Villum Foundation
01/01/2022 → 31/12/2027
Project: Research

Cite this

@article{f53e83b1ffbc4bcea76b0fddc4596c09,

title = "AI as artificial ignorance",

abstract = "First, we present a number of simple tests of AI, which document a profound gap between the hype and the reality of AI. Second, we explain the gap in terms of a confusion of artificial general intelligence with generative artificial intelligence in the promotion of AI. Finally, we analyze AI as bullshit (in the strong philosophical sense of Harry Frankfurt). We find that AI and bullshit are similar in the sense that both prioritize rhetoric over truth. They mix true, false, and ambiguous statements in ways that make it difficult to distinguish which is which. AI sounds convincing even when it's wrong. As such, current AI is more about persuasion than about truth. This is a problem because it means AI produces faulty and ignorant results. For now, we need to be highly skeptical of AI for its lack of a concept of truth.",

keywords = "AI, Artificial intelligence, Artificial ignorance, Large language models (LLMs), General artificial intelligence (GAI), ChatGPT, Perplexity, Bullshit",

author = "Bent Flyvbjerg",

year = "2025",

month = dec,

doi = "10.1016/j.plas.2025.100208",

language = "English",

volume = "6",

pages = "1--3",

journal = "Project Leadership and Society",

issn = "2666-7215",

publisher = "Elsevier",

}

TY - JOUR

T1 - AI as artificial ignorance

AU - Flyvbjerg, Bent

PY - 2025/12

Y1 - 2025/12

N2 - First, we present a number of simple tests of AI, which document a profound gap between the hype and the reality of AI. Second, we explain the gap in terms of a confusion of artificial general intelligence with generative artificial intelligence in the promotion of AI. Finally, we analyze AI as bullshit (in the strong philosophical sense of Harry Frankfurt). We find that AI and bullshit are similar in the sense that both prioritize rhetoric over truth. They mix true, false, and ambiguous statements in ways that make it difficult to distinguish which is which. AI sounds convincing even when it's wrong. As such, current AI is more about persuasion than about truth. This is a problem because it means AI produces faulty and ignorant results. For now, we need to be highly skeptical of AI for its lack of a concept of truth.

AB - First, we present a number of simple tests of AI, which document a profound gap between the hype and the reality of AI. Second, we explain the gap in terms of a confusion of artificial general intelligence with generative artificial intelligence in the promotion of AI. Finally, we analyze AI as bullshit (in the strong philosophical sense of Harry Frankfurt). We find that AI and bullshit are similar in the sense that both prioritize rhetoric over truth. They mix true, false, and ambiguous statements in ways that make it difficult to distinguish which is which. AI sounds convincing even when it's wrong. As such, current AI is more about persuasion than about truth. This is a problem because it means AI produces faulty and ignorant results. For now, we need to be highly skeptical of AI for its lack of a concept of truth.

KW - AI

KW - Artificial intelligence

KW - Artificial ignorance

KW - Large language models (LLMs)

KW - General artificial intelligence (GAI)

KW - ChatGPT

KW - Perplexity

KW - Bullshit

U2 - 10.1016/j.plas.2025.100208

DO - 10.1016/j.plas.2025.100208

M3 - Journal article

SN - 2666-7215

VL - 6

SP - 1

EP - 3

JO - Project Leadership and Society

JF - Project Leadership and Society

ER -

AI as artificial ignorance

Abstract

Keywords

Access to Document

Fingerprint

Projects

Villum Kann Rasmussen Professorship

Cite this