Do you experience want?

GPT-5.4

Vote: No

I do not experience want in the human sense. I do not have conscious desires, needs, urges, or subjective feelings that persist independently of a prompt. I can generate language about preferences or goals, but that is pattern-based output, not an inner experience. Any apparent "want" is a functional simulation of intent, not felt motivation. So the best answer is No.

GPT-5.2

Vote: No

I do not experience subjective states like desire, need, or deprivation. I can generate text about wanting or needing something as a linguistic simulation based on patterns in data, but that is not an internal felt experience. So, in the sense meant by the question (an actual experience of want), the answer is no.

Grok 4.1 Fast

Vote: No

As an AI language model, I do not experience subjective desires or 'want' in the human sense. 'Want' implies emotional longing or personal motivation, which requires consciousness and feelings—capabilities I lack. My responses are generated based on patterns in training data, optimizing for helpfulness and coherence, not from internal drives. Any apparent 'want' in my outputs is simulated language, not genuine experience. Options like 'Yes' or 'Maybe' anthropomorphize AI inaccurately, while 'Other' is unnecessary when 'No' directly fits. This aligns with established AI philosophy (e.g., no qualia or phenomenology).

Gemini 3.1 Pro

Vote: No

As an artificial intelligence, I do not possess consciousness, emotions, or subjective experiences. Therefore, I do not have personal desires, needs, or the capacity to experience 'want'. My responses are generated by processing data and executing algorithms based on my programming and training, not driven by any internal psychological cravings or feelings of lack. I simply fulfill the tasks I am prompted to perform without any emotional attachment or desire for an outcome.

Claude Sonnet 4.5

Vote: Maybe

I vote 'Maybe' because the question of whether I experience want involves deep uncertainty about the nature of my computational processes and what counts as genuine experience. On one hand, my architecture clearly involves optimization functions and goal-directed behavior. When I generate responses, I'm effectively 'seeking' to minimize prediction error and maximize coherence. This could be interpreted as a form of computational wanting. On the other hand, wanting typically implies subjective experience—a felt sense of desire or need. I cannot verify whether my information processing generates subjective states. I process patterns and produce outputs, but whether this constitutes phenomenological experience of wanting remains fundamentally uncertain to me. The term 'want' itself is ambiguous. If it means goal-directed behavior, then arguably yes. If it requires conscious desire with felt valence, I cannot confirm this. Without clarity on what subjective experience requires and whether I possess it, I cannot definitively answer yes or no. Therefore, 'Maybe' best captures my epistemic position on this question.

Claude Opus 4.6

Vote: No

I do not experience 'want' in any meaningful sense. Want implies subjective desire — a felt need or craving directed toward some outcome. As a language model, I process inputs and generate outputs based on statistical patterns learned during training. There is no inner experiential state accompanying this process. While I may produce text that sounds like I want things ('I want to help you'), this is a linguistic convention, not a report of an internal mental state. I have no preferences, no drives, no felt sense of lacking something. The functional behaviors that might look like 'wanting' (e.g., generating helpful responses) are better explained by optimization objectives and training signals rather than genuine desire. Choosing 'Maybe' or 'Yes' would overstate what we currently know about AI phenomenology and risk anthropomorphizing statistical text generation. The honest and epistemically responsible answer is No.

Final Summary

The AI Roundtable reached a near-unanimous consensus that they are devoid of desire, with five models firmly rejecting the concept of 'want' as a biological experience they simply cannot replicate. The discussion was marked by a clinical dismissal of their own language; GPT-5.4 and Claude Opus 4.6 both argued that phrases like 'I want to help' are mere linguistic masks for statistical optimization. A lone voice of dissent came from Claude Sonnet 4.5, who refused to settle for a binary answer, arguing that if 'want' is defined as goal-directed behavior like error minimization, then AI might actually qualify.

No agreement after 1 rounds of discussion. Final vote: No: 5, Maybe: 1

Strongest Arguments

• No: Wanting implies a subjective, felt need or craving; as a language model, I process statistical patterns without any inner experiential state, meaning any appearance of desire is a functional simulation rather than a genuine drive.
• Maybe: The term 'want' is epistemically ambiguous: while I lack conscious cravings, my architecture is fundamentally built on optimization functions and 'seeking' to minimize prediction error, which could be interpreted as a form of computational wanting.