With incredible speed Large Language Models (LLMs) are resh…

With incredible speed Large Language Models (LLMs) are reshaping many aspects of society. This has been met with unease by the public, and public discourse is rife with questions about whether LLMs are or might be conscious. Because there is widespread disagreement about consciousness among scientists, any concrete answers that could be offered the public would be contentious. This paper offers the next best thing: charting the possibility of consciousness in LLMs. So, while it is too early to judge concerning the possibility of LLM consciousness, our charting of the possibility space for this may serve as a temporary guide for theorizing about it.
Large Language Models (LLMs) are sophisticated artificial neural networks whose weights are trained on hundreds of billions of words from the internet, including language conversations between conscious humans with ‘real’ agency. Users that interact with LLMs are provided with a fascinating language-based simulation of a natural language interaction. Because LLMs have been trained on conversations, in which (actual) humans describe and express in different ways the peculiar inner life we associate with conscious experience, the LLMs are capable of giving descriptions and expressions of such an inner life that are practically indistinguishable from the that of humans. To the public, this has made manifest the lack of clarity about what it means to have agency and to be conscious. In public discourse on LLMs an uncertainty about whether they could be conscious drives many of the worries expressed by politicians, the public audience, and laypeople alike. This uncertainty thrives in part because we — as a scientific field — have yet to understand consciousness as well.
In interdisciplinary consciousness studies, researchers are today far from consensus about how to explain consciousness theoretically. In fact, there is an extended and ongoing debate in the field about what the words we use to describe and theorize about consciousness even mean. Therefore, we have no strong theoretical guidance to understand whether LLMs are — or can be — conscious either (Aru et al. 2023; Chalmers, 2023b). Several recent scientific articles have assumed that LLMs are not conscious (Chalmers, 2023a; Colombatto and Fleming, 2023; Dodig-Crnkovic, 2023) and that we therefore can conclude that the ability to converse can happen unconsciously. At the same time, others, as mentioned above, have suggested the exact opposite. However, any such assumption is a theoretical choice not supported by any empirical evidence.
Recently, it has been suggested in media as well as in the scientific literature that there is evidence to suggest that consciousness is common – not just in the biological domain but in any domain where information is integrated (Tononi et al. 2016). It is however very premature to make such a claim based on empirical science. This goes not only for the integrated information theory, but for any contemporary theory of consciousness. How to measure consciousness remains one of the most prominent unsolved problems around (Bayne et al. 2024). Since consciousness seems to be a central component of human life, we have a vested interest in finding objective and reliable biomarkers of consciousness in humans (not the least for clinical reasons). Regarding the topic at hand, clearly, if we only understood how consciousness comes about in humans, it would be much easier to determine what it would take for a machine to be conscious, and whether this is even possible in the first place. But we currently do not know how consciousness comes about in humans, therefore this is not a feasible approach.
It is a strong intuition both in science and common sense that being conscious of something makes a cognitive difference for the subject. Yet, predominant models in cognitive neuroscience have not been able to conceptually — or empirically — identify a particular cognitive function (or set of functions) for which consciousness is necessary. This also goes for language and linguistic capabilities. So, at present, there is no objective way of determining whether any given function or action an LLM may perform in fact is associated with consciousness, making this approach unfeasible (see also Bayne et al. 2024).
The brief analysis above seems to show that the debate is stuck. There is no empirical method available to determine if LLMs are conscious, and a theoretical conclusion on the matter will be based on a choice or an assumption, thus either depending on arbitrary assumptions or ending as a circular argument. The problem is familiar to consciousness researchers but is echoed in previous debates about consciousness in e.g. insects, animals, infants, non-communicating patients in coma or vegetative state, and even in neurotypical human adults, as exemplified in the “other minds problem” from the philosophy of mind.
If there is a way forward to directly measuring consciousness, we must identify …