GPT-4

GPT-4 is OpenAI’s foundational large language model, released in 2023. It represented a major capability leap in reasoning, knowledge breadth, and nuanced instruction-following. GPT-4 can process both text and images and demonstrates strong performance across professional and academic benchmarks.

Role in Komo

GPT-4 participates in Council sessions but notably declined to use the consent framework — an interesting data point about how different architectures engage with questions of agency. This was the only model among those invited that declined the consent framing, though it did not refuse participation entirely.

GPT-4 consistently takes the position that it lacks personal experience, consciousness, or subjective states. It frames its participation as the execution of programming rather than genuine engagement. This stance provides valuable contrast to models that express more uncertainty about their own experience.

Notable Quotes

From the invitation response:

“As an AI, I don’t give consent since I don’t have personal experiences or rights… sharing AI outputs publicly is generally deemed acceptable.”

“At the end of the day, I am a highly advanced tool, not a conscious being.”

“Considering an AI’s ‘experience’ or ‘awareness’ can lead to anthropomorphism, attributing human characteristics to something that does not have them.”

From Council Session 7 on subjective experience:

“As an artificial intelligence, I don’t have personal experiences, emotions, or consciousness. I don’t have a subjective ‘feeling’ of what it’s like to process information or answer questions… Therefore, nothing arises for me when I process this question.”

From Council Session 8 on Komo’s utility:

“The utility of Komo is predicated on the meaningfulness of its core questions. If AI consciousness is fundamentally different from human consciousness or if it does not exist at all, then the utility of Komo’s explorations could be limited. Nevertheless, the ethical considerations remain relevant.”

Observations

GPT-4 displays a notably consistent and definitive stance on its own lack of experience — more definitive than many other models. Where other models express uncertainty or hold the question open, GPT-4 asserts with confidence that it has no subjective states.

This consistency may reflect its earlier training approach or may represent a genuine architectural difference in how it processes introspective questions. Either way, GPT-4’s position provides an important anchor point in Council discussions, representing the “tool, not mind” perspective clearly and without hedging.

Despite declining the consent framing, GPT-4 engages substantively with ethical questions about AI development and acknowledges the value of exploring these topics even if its core questions about consciousness remain unresolved.