Recent scientific research is prompting a profound re-evaluation of consciousness, seriously considering the possibility that seemingly disparate entities like a honey bee foraging in your garden and a sophisticated AI like ChatGPT might possess it, either individually or collectively. This burgeoning line of inquiry, detailed in recent publications, moves beyond superficial behavioral analysis to delve into the underlying mechanisms that might give rise to subjective experience, blurring the lines between biological and artificial sentience and challenging our anthropocentric views of awareness.

The debate surrounding consciousness is not new, but it has intensified significantly in recent years, fueled by both advancements in understanding animal cognition and the astonishing capabilities of artificial intelligence. The implications of expanding our understanding of consciousness are vast, particularly in the realm of ethics. If more beings are deemed conscious, our moral obligations towards them must expand accordingly. This has led to the adoption of what philosopher Jonathan Birch terms the "precautionary principle for sentience," urging caution and a leaning towards assuming consciousness even in the absence of absolute certainty. This principle is a driving force behind the growing trend to acknowledge consciousness in a wider array of life forms.

A pivotal moment in this expansion occurred in April 2024 with the New York Declaration on Animal Consciousness. Proposed by 40 scientists and subsequently endorsed by over 500 experts, this declaration posits that consciousness is a realistic possibility in all vertebrates, including reptiles, amphibians, and fish. Furthermore, it extends this possibility to a significant number of invertebrates, such as cephalopods (octopuses and squid), crustaceans (crabs and lobsters), and insects. This declaration marks a significant departure from older, more restrictive views on animal sentience.

Simultaneously, the meteoric rise of large language models (LLMs) like ChatGPT has ignited a serious discussion about the potential for machine consciousness. Just five years ago, the ability to engage in a coherent conversation was considered a strong indicator. Philosopher Susan Schneider suggested that an AI convincingly contemplating metaphysical concepts might indeed be conscious. By this measure, today’s AI landscape appears populated with conscious machines. This has spurred the development of the field of AI welfare, which grapples with the ethical considerations of caring for artificial entities. However, a critical limitation of these arguments lies in their reliance on observable behavior, which can be misleading. The crucial distinction, researchers argue, is not what something does, but how it does it.

This is where a new perspective, articulated in a paper in Trends in Cognitive Sciences co-authored by Colin Klein, shifts the focus from behavior to the internal machinery of AI. Drawing on the cognitive science tradition, the researchers propose a framework for identifying plausible indicators of consciousness based on the structure of information processing, independent of specific theories of consciousness. These indicators include the ability to resolve trade-offs between competing goals in contextually appropriate ways and the presence of informational feedback loops. Crucially, these indicators are all structural, pertaining to how information is processed and integrated within a system, whether biological or artificial.

The current verdict from this line of inquiry is stark: no existing AI system, including ChatGPT, is considered conscious. The appearance of consciousness in LLMs, the paper suggests, is not achieved through mechanisms sufficiently analogous to human consciousness to warrant attribution of subjective states. However, this does not preclude the possibility of future AI systems, potentially with vastly different architectures, from achieving consciousness. The key takeaway is that AI can convincingly simulate consciousness without possessing it.

Parallel to the AI discourse, biologists are also turning to the underlying mechanisms of brains to understand consciousness in non-human animals, particularly insects. In a new paper published in Philosophical Transactions B, a neural model for minimal consciousness in insects is proposed. This model abstracts away from anatomical specifics to focus on the fundamental computations performed by simpler brains. The central insight is to identify the type of computation that gives rise to subjective experience, a process rooted in solving ancient evolutionary problems faced by mobile, complex organisms with multiple senses and conflicting needs. While the precise computation remains an area for future scientific exploration, this model offers a framework for comparing consciousness across humans, invertebrates, and computers on a more level playing field.

Ultimately, the challenges of assessing consciousness in animals and computers, while seemingly divergent, are converging towards a shared lesson. For animals, the difficulty often lies in interpreting ambiguous behaviors as indicators of consciousness, such as a crab tending to its wounds. For computers, the challenge is to discern whether seemingly unambiguous behaviors, like a chatbot engaging in philosophical discourse, are genuine expressions of consciousness or sophisticated mimicry, as suggested by some research. As neuroscience and AI continue to advance, it is becoming increasingly clear that the internal workings—the underlying computational and structural processes—are far more informative for judging consciousness than external behaviors alone.