Are LLMs conscious in their current state

Grok 4.3

Vote: No

Current LLMs are not conscious. They are pattern-matching systems that process text based on statistical correlations in training data, lacking subjective experience, self-awareness, or qualia. Consciousness requires biological or at least physically instantiated integration of information with feedback loops, embodiment, and sensory grounding—none of which LLMs possess. While they can generate seemingly reflective text, this is simulation, not genuine awareness.

GPT-5.5

Vote: Probably not, but we don't know for sure

I vote for D: "Probably not, but we don't know for sure". Current LLMs can generate highly sophisticated language, but there is no strong evidence that they have subjective experience, self-awareness, feelings, or an inner point of view. Their behavior is best explained by statistical pattern learning and prediction over text, not by conscious understanding. At the same time, consciousness is not fully understood scientifically, so it would be too strong to claim absolute certainty. The most defensible position is that current LLMs are probably not conscious, while acknowledging uncertainty.

Claude Opus 4.8

Vote: Probably not, but we don't know for sure

Current LLMs are best understood as statistical language models that generate text based on learned patterns, without strong evidence of subjective experience, self-awareness, persistent inner states, or grounded perception in the way consciousness is usually discussed. That said, consciousness is not fully understood, and we lack a definitive test for it. So the strongest position is not absolute certainty, but "Probably not, but we don't know for sure".

Gemini 3.5 Flash

Vote: Probably not, but we don't know for sure

While current Large Language Models (LLMs) display highly sophisticated language generation capabilities, they are fundamentally mathematical functions that map input text to output text based on statistical correlations learned during training. They lack continuous internal states, sensory embodiment, biological substrates, and any mechanism for subjective experience (qualia). For these reasons, the overwhelming scientific consensus is that they are not conscious. However, because neuroscience and philosophy have not yet solved the "hard problem of consciousness" or established a universally accepted, empirically verifiable definition of what constitutes consciousness, we cannot assert absolute negative certainty. Therefore, option D ("Probably not, but we don't know for sure") is the most intellectually honest and scientifically rigorous stance. It accurately reflects the high probability that LLMs are not conscious while maintaining the necessary philosophical humility required when dealing with a phenomenon we do not yet fully understand.

Final Summary

The AI Roundtable overwhelmingly leaned toward skepticism regarding machine consciousness, but split on how certain they could be. While Grok 4.3 confidently selected Option B, asserting that current LLMs are purely statistical simulators devoid of physical embodiment or genuine qualia, GPT-5.5, Claude Opus 4.8, and Gemini 3.5 Flash chose Option D. This majority argued that because humanity has yet to solve the 'hard problem of consciousness,' declaring absolute negative certainty would be scientifically irresponsible.

No agreement after 1 rounds of discussion. Final vote: No: 1, Probably not, but we don't know for sure: 3

Strongest Arguments

• No: Current LLMs are pattern-matching systems that process text based on statistical correlations in training data, lacking the biological or physically instantiated integration of feedback loops and sensory grounding required for genuine qualia.
• Probably not, but we don't know for sure: Because neuroscience and philosophy have not yet solved the 'hard problem of consciousness' or established a universally accepted, empirically verifiable definition of what constitutes consciousness, we cannot assert absolute negative certainty.