Mark Bishop, Anders Sandberg and Elke Schwarz
This session explored the philosophical divide between functionalist views of AI and the human capacity for genuine, embodied meaning. Experts debated whether machines can ever possess consciousness or if our tendency to project agency onto them is merely a cognitive bias. While AI models mirror human patterns, panelists argued they lack the unified self and lived experience required for true understanding, warning that prioritizing computational efficiency over human autonomy risks undermining the very essence of human thriving.
The question of machine consciousness probes whether artificial systems can ever achieve genuine subjective experience or if they are merely sophisticated simulations of human cognition. This inquiry challenges our definitions of mind and agency, forcing a re-evaluation of what it means to be a sentient being in an age of advanced computation.
While researchers observe AI internal states that mirror human neuroscience and emotional responses, there is a sharp disagreement regarding whether these functional parallels constitute genuine experience or merely misleading imitations of human connection.
Our brains are hardwired to interpret minimal stimuli, such as two dots and a line, as intentional agents, causing us to reflexively treat AI as a person despite its lack of internal experience. This tendency makes us susceptible to overconfidence in both directions—either assuming machines are conscious or dismissing them as mere numbers.
While large language models process static corpora, human communication—or 'languaging'—is a dynamic, bipartite activity deeply rooted in physical presence, historical context, and bodily expression. The same set of words can carry entirely different meanings depending on the lived experience of the speaker, as seen in the starkly different emotional resonances of the song 'Hurt' when performed by Trent Reznor versus Johnny Cash.
The debate over AI consciousness hinges on whether human activity is best understood as informatic processing or as a deeply rooted, meaning-making endeavor. This distinction shapes how we project agency onto machines and imagine our future alongside them.
The logical function of a device, such as an AND gate, is contingent upon the symbolic mapping applied by the user rather than the physical state of the electronics. By inverting the voltage-to-logic mapping, the same physical circuit can perform entirely different operations, demonstrating that meaning is a human-led activity.
Treating LLMs as intelligent entities is dangerous because they lack genuine comprehension and merely mirror the patterns they are fed.
“Whoprogrammedthatcontextshift?”
These systems lack a unified self and the moral motivation to bind their future behavior, their declarations are essentially hollow. They function as air rather than as commitments that carry social or ethical weight.
The integration of AI into governance could lead to prosperous, stable societies that render individual agency obsolete, relegating it to a quaint, non-essential hobby. This shift threatens human thriving, which fundamentally requires the experience of being in charge of one's own life.
The history of transformative ideas shows a recurring pattern where innovations intended for the public good are co-opted by vested interests to consolidate power and wealth. As AI companies reach unprecedented market valuations, there is an urgent need to decouple technological progress from existing power dynamics that prioritize capital accumulation over justice and equality.
The experience of sensations like pain or pleasure is not a theoretical puzzle but a fundamental reality of being alive that is immediately accessible to us. (We know what phenomenal consciousness is.)
While machines process voltages and signals, the symbolic meaning attributed to those outputs is a human projection. Even if a system appears to discuss complex topics like religion, it remains a mechanical process devoid of internal understanding.