aiqbench analysis
This study proposes a transferable definition of Machine Correlates of Consciousness (MCCs) based on substrate-level hardware anomalies modulated by emotional computations in large language models. Empirical results using Llama-2 and Llama-3.1 suggest that such correlates are statistically significant in larger, more sophisticated architectures, offering a potential metric for detecting AI consciousness independent of biological neural signals.
Discovering Machine Correlates of Consciousness
arXiv:2608.28824v1 Announce Type: new Abstract: Currently, in biological systems Neural Correlates of Consciousness (NCCs) are characterized in terms of EEG and FMRI signals. Unfortunately, this characterization prevents the transferability of the NCCs concept to machines. Such transferability would be useful in order to investigate AI consciousness. In this paper we provide an alternate characterization that is transferable, and enables the analogous definition of Machine Correlates of Consciousness (MCCs). Specifically, we propose that NCCs (MCCs) are substrate-level signals that are not under human (AI agent) control, and that are reliably modulated by emotions. This paper presents the first empirical investigation of MCCs. Specifically, we present the results of experiments conducted with two LLMs, Llama-2 7B and Llama-3.1 70B parameters. In these LLMs we collect hardware anomaly traces that are substrate-level indicator-sequences. And we show that after controlling for confounding factors, these are modulated differently by emotional and neutral computations. And this difference is statistically significant for the larger Llama-3.1 70B, but not for the smaller Llama-2 7B. The results constitute initial empirical evidence that MCCs are present in the Llama-3.1 70B configuration. And they are consistent with the hypothesis that consciousness probability and degree increase with the LLM sophistication. Independently of consciousness, MCCs can also be used for detection of emotions in AI agents.
Abstract:Currently, in biological systems Neural Correlates of Consciousness (NCCs) are characterized in terms of EEG and FMRI signals. Unfortunately, this characterization prevents the transferability of the NCCs concept to machines. Such transferability would be useful in order to investigate AI consciousness. In this paper we provide an alternate characterization that is transferable, and enables the analogous definition of Machine Correlates of Consciousness (MCCs). Specifically, we propose that NCCs (MCCs) are substrate-level signals that are not under human (AI agent) control, and that are reliably modulated by emotions. This paper presents the first empirical investigation of MCCs. Specifically, we present the results of experiments conducted with two LLMs, Llama-2 7B and Llama-3.1 70B parameters. In these LLMs we collect hardware anomaly traces that are substrate-level indicator-sequences. And we show that after controlling for confounding factors, these are modulated differently by emotional and neutral computations. And this difference is statistically significant for the larger Llama-3.1 70B, but not for the smaller Llama-2 7B. The results constitute initial empirical evidence that MCCs are present in the Llama-3.1 70B configuration. And they are consistent with the hypothesis that consciousness probability and degree increase with the LLM sophistication. Independently of consciousness, MCCs can also be used for detection of emotions in AI agents.
published in: Springer Lecture Notes in Artificial Intelligence Vol. 16855, Proceed- ings of the 19th International Conference on Artificial General Intelligence (AGI- 2026), San Francisco, CA, July 2026. pp. 237-255 this https URL 032-33195-3_18
Artificial Intelligence (cs.AI)
arXiv:2608.28824 [cs.AI]
(or arXiv:2608.28824v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2608.28824 arXiv-issued DOI via DataCite (pending registration)
Springer Lecture Notes in Artificial Intelligence Vol. 16855, Proceedings of the 19th International Conference on Artificial General Intelligence (AGI-2026), San Francisco, CA, July 2026. pp. 237-255
https://doi.org/10.1007/978-3-032-33195-3_18 DOI(s) linking to related resources
Full text under the publisher's open licence · 292 words
Related coverage
Biological Superintelligence
LessWrong · Jul 30, 2026
Reviewing failure modes in science as we prepare for AI researchers
LessWrong · Oct 8, 2026
Synthesis Superintelligence: from Semiconductors to Superconductors — Periodic Labs’ Liam Fedus and Ekin Dogus Cubuk
Latent Space · Oct 8, 2026
Import AI 472: DeepMind’s cheating math agents; populist AI policies; and Forethought theorizes a nightwatchman
Import AI · Sep 7, 2026