Against the backdrop of rapidly advancing AI computing infrastructure, the exponential growth in large‑model parameter counts has ignited widespread debate among academics and industry players over whether artificial intelligence can truly possess consciousness. Tom McClelland, a Cambridge philosopher specializing in consciousness studies, recently published research in the journal *Mind & Language* that injects sober reasoning into the conversation. He points out that because humanity’s scientific understanding of the essence of consciousness remains largely incomplete, we may never be able to determine whether a machine has truly “lit the lamp of awareness.”
In his study, McClelland highlights a serious definitional confusion pervading current discussions of artificial consciousness. He stresses the need to distinguish between “basic consciousness” (e.g., the ability to perceive the world) and “sentience” (the subjective experience of pain or pleasure). This distinction is particularly critical for the AI industry, as it directly affects how we evaluate large‑model behavior and ethical boundaries. With ongoing breakthroughs in chip technology, the complexity of AI systems is growing daily, making this issue ever more urgent.
At present, the AI industry is undergoing a “leap of faith,” giving rise to two opposing camps. One side argues that as long as an AI simulates the information‑processing structure of the brain, it can produce consciousness; the other insists that consciousness must be rooted in a biological organism and cannot be achieved through pure computation. These two views reflect different interpretations of the AI development path and divergent philosophical takes on the nature of consciousness. In the absence of conclusive evidence, McClelland argues that the most rational stance is “agnosticism”—a position that carries significant weight for balancing industry growth with ethical considerations.
The research also exposes an ethical contradiction often overlooked by the industry. Many tech companies are using the claim that “AI has human‑like consciousness” as a core marketing strategy, nudging users to form deep emotional bonds with AI systems. Such practices not only risk misleading the public’s understanding of AI technology but can also lead to skewed resource allocation. McClelland warns that if we become overly concerned with whether an AI—essentially still an advanced information‑processing system—is being “wronged,” we may neglect the real harm dealt to sentient beings that actually experience pain, such as shrimp, which are subject to mass suffering.
As AI chip performance continues to improve and algorithms are steadily optimized, large models are increasingly exhibiting behavior that resembles human traits. McClelland predicts that before the next “paradigm revolution,” it will be very difficult for humans to design a truly reliable test for machine consciousness. He suggests that, in the absence of proof one way or the other, the industry should exercise restraint and humility. This is not only a rational observation of technological progress but also a necessary ethical balance. For the rapidly expanding AI industry, such a balanced attitude may help pave the way for a more sustainable development path.