Why we need an AI-resilient society

Thomas Bartz-Beielstein

Abstract

Artificial intelligence is considered as a key technology. It has a huge impact on our society. Besides many positive effects, there are also some negative effects or threats. Some of these threats to society are well-known, e.g., weapons or killer robots. But there are also threats that are ignored. These unknown-knowns or blind spots affect privacy, and facilitate manipulation and mistaken identities. We cannot trust data, audio, video, and identities any more. Democracies are able to cope with known threats, the known-knowns. Transforming unknown-knowns to known-knowns is one important cornerstone of resilient societies. An AI-resilient society is able to transform threats caused by new AI technologies such as generative adversarial networks. Resilience can be seen as a positive adaptation of these threats. We propose three strategies how this adaptation can be achieved: awareness, agreements, and red flags. This article accompanies the TEDx talk "Why we urgently need an AI-resilient society", see https://youtu.be/f6c2ngp7rqY.

The Problem: The Erosion of Digital Trust

As AI technologies like Generative Adversarial Networks (GANs) make it increasingly difficult to trust digital media, we need more than just technical "anti-virus" patches. We are currently in a "shadow" phase where AI's negative effects are outpacing societal adaptation. Technical solutions alone create an infinite loop: because GANs learn by trying to fool a discriminator, technical countermeasures inadvertently train the AI to be even more sophisticated at creating fakes.

To address this, we must build an AI-resilient society—one that proactively transforms "unknown-knowns" (blind spots we ignore) into "known-knowns" through public engagement and policy.

Conceptual Framework: The Johari Window of AI Threats

A critical gap exists in how we handle AI threats. While we address "known-knowns" (like software bugs) and speculate about "known-unknowns" (like the distant threat of Super-intelligence), we are failing to address "unknown-knowns."

In psychology, these are "blind spots"—threats that exist and are technically understood by experts but are largely ignored by the public, such as the erosion of privacy via social media or the manipulation of identity through deep fakes. Resilience is defined as the positive adaptation to these threats by moving them out of the blind spot and into the sphere of public understanding and regulation.

The Johari Window applied to AI threats, highlighting the "blind spot" of unknown-knowns and the path toward AI-resilience.

Historical Context: The Hexapawn Analogy

To understand the logic of modern AI, we can look back to Martin Gardner’s 1962 "Hexapawn," a matchbox-based learning machine. This simple game illustrates that the concept of machines learning through competition and punishment is not new. In Hexapawn, matchboxes represent game states, and the "machine" learns by removing moves that lead to defeat.

Martin Gardner’s Hexapawn (1962), a matchbox-based learning machine that serves as a historical precursor to the competitive learning logic used in modern GANs.

Modern GANs operate on a similar principle, where two networks compete. This historical perspective shows that while the concept is old, the current scale and speed of such learning are unprecedented, necessitating a broader societal framework for resilience.

The Three-Pillar Strategy for Resilience

To achieve resilience, we propose three specific strategies for societal adaptation:

  1. Awareness: Moving technical knowledge into the public sphere. This involves educating the public on tools like reverse image searches and fact-checking to restore trust in data.
  2. Agreements: Establishing international ethical principles and laws. This includes updated versions of Asimov’s Laws or the "Partnership on AI" to create a global standard for AI behavior.
  3. Red Flags: Implementing mandatory self-identification for AI systems. Inspired by the 19th-century "Locomotive Act," any autonomous system must identify itself at the start of any interaction with a human. If you are talking to a bot or looking at a generated image, it should be legally required to "wave a red flag."

Conclusion

An AI-resilient society is defined by its ability to transform blind spots into publicly understood threats and to develop rules for those threats. Openness and knowledge transfer are the only ways to restore trust. We must move away from closed-door technical fixes and toward public engagement, mandatory AI disclosure, and international ethical standards.

Summary updated

Abstract

Artificial intelligence is considered as a key technology. It has a huge impact on our society. Besides many positive effects, there are also some negative effects or threats. Some of these threats to society are well-known, e.g., weapons or killer robots. But there are also threats that are ignored. These unknown-knowns or blind spots affect privacy, and facilitate manipulation and mistaken identities. We cannot trust data, audio, video, and identities any more. Democracies are able to cope with known threats, the known-knowns. Transforming unknown-knowns to known-knowns is one important cornerstone of resilient societies. An AI-resilient society is able to transform threats caused by new AI tecchnologies such as generative adversarial networks. Resilience can be seen as a positive adaptation of these threats. We propose three strategies how this adaptation can be achieved: awareness, agreements, and red flags. This article accompanies the TEDx talk "Why we urgently need an AI-resilient society", see https://youtu.be/f6c2ngp7rqY.
BibTeX
@misc{bartzbeielstein2019why,
  author = {Thomas Bartz-Beielstein},
  title = {{Why we need an AI-resilient society}},
  year = {2019},
  eprint = {1912.08786},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1912.08786},
}
Figure