Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generative AI-based closed-loop fMRI system

Mikihiro Kasahara, Taiki Oka, Vincent Taschereau-Dumouchel, Mitsuo Kawato, Hiroki Takakura, Aurelio Cortese

TL;DR

The paper addresses the risk that malicious generative AI could steer cognition or decision-making and the lack of neural or computational countermeasures. It introduces DecNefGAN, a closed-loop architecture that couples decoded neurofeedback with GAN-like generative AI, using MVPA decoding and CLIP-based personalization to map brain states to AI-generated stimuli. The work outlines the architecture, decoding/generation pipeline, potential issues, and ethical considerations, and discusses applications in security and psychiatry. By enabling the study of human resilience to AI manipulation, DecNefGAN aims to illuminate neural mechanisms of counteracting AI influence and to inform safeguards for cognitive safety and robust human-AI interaction.

Abstract

While generative AI is now widespread and useful in society, there are potential risks of misuse, e.g., unconsciously influencing cognitive processes or decision-making. Although this causes a security problem in the cognitive domain, there has been no research about neural and computational mechanisms counteracting the impact of malicious generative AI in humans. We propose DecNefGAN, a novel framework that combines a generative adversarial system and a neural reinforcement model. More specifically, DecNefGAN bridges human and generative AI in a closed-loop system, with the AI creating stimuli that induce specific mental states, thus exerting external control over neural activity. The objective of the human is the opposite, to compete and reach an orthogonal mental state. This framework can contribute to elucidating how the human brain responds to and counteracts the potential influence of generative AI.

Generative AI-based closed-loop fMRI system

TL;DR

The paper addresses the risk that malicious generative AI could steer cognition or decision-making and the lack of neural or computational countermeasures. It introduces DecNefGAN, a closed-loop architecture that couples decoded neurofeedback with GAN-like generative AI, using MVPA decoding and CLIP-based personalization to map brain states to AI-generated stimuli. The work outlines the architecture, decoding/generation pipeline, potential issues, and ethical considerations, and discusses applications in security and psychiatry. By enabling the study of human resilience to AI manipulation, DecNefGAN aims to illuminate neural mechanisms of counteracting AI influence and to inform safeguards for cognitive safety and robust human-AI interaction.

Abstract

While generative AI is now widespread and useful in society, there are potential risks of misuse, e.g., unconsciously influencing cognitive processes or decision-making. Although this causes a security problem in the cognitive domain, there has been no research about neural and computational mechanisms counteracting the impact of malicious generative AI in humans. We propose DecNefGAN, a novel framework that combines a generative adversarial system and a neural reinforcement model. More specifically, DecNefGAN bridges human and generative AI in a closed-loop system, with the AI creating stimuli that induce specific mental states, thus exerting external control over neural activity. The objective of the human is the opposite, to compete and reach an orthogonal mental state. This framework can contribute to elucidating how the human brain responds to and counteracts the potential influence of generative AI.
Paper Structure (8 sections, 2 figures)

This paper contains 8 sections, 2 figures.

Table of Contents

  1. Introduction
  2. New closed-loop fMRI framework of human and generative AI
  3. Overall Architecture
  4. Generative AI architecture for specific mental status and human counteraction
  5. Potential issues and solutions
  6. Applications
  7. Ethical challenges
  8. Concluding remarks

Figures (2)

  • Figure 1: Architecture of DecNefGAN (Decoded Neurofeedback based Generative Adversarial Networks of human and generative AI).
  • Figure 2: Generative process based on decoding of latent brain states. Decoding the strength of a mental state from the brain, then associating the strength of the decoded mental state with pre-collected ratings from the participants. The mapped strength of the identified mental state is then correlated with image features in the latent representations of CLIP. From these latent spaces, generating images corresponding to levels of a specific mental state becomes possible.