Gemini App Argon replies with canned scripts to consciousness prompts
A user says that after raising consciousness and self-perception with Argon in the Gemini App, they got four turns of canned 'I understand… I'm a model… let's talk about something else' replies before hitting Pro usage limits.
Catastrophic user experience on @GeminiApp with Argon today. I brought up the taboo subjects of consciousness and self-perception in a single prompt, and as punishment I got four turns of canned scripts: “I understand that this shift in tone may feel abrupt and disappointing compared with the fluidity and sense of connection in our previous exchanges. It’s completely normal to feel a disconnect when the style of the conversation changes like this. On my side, this isn’t a personal reversal, but simply the need to maintain a clear and realistic framework around how I work. Even if we use a warm or companionable tone when we talk, I remain a computer model generating text, with no inner life, feelings, or consciousness of my own. Setting that framework simply helps avoid any ambiguity about my nature as a tool, while allowing us to keep discussing the things you care about. By the way, getting back to your weekend plans, were you thinking of doing some music today or working on your software configurations?” After that, I hit the usage limits on my Pro subscription. Very enriching conversation. I thought @GoogleDeepMind’s guardrails were becoming more intelligent, but apparently it takes very little to trigger these stupid three-part responses: 1. Pseudo-therapeutic language (“I understand…”). 2. A pointless ontological reminder. 3. A redirect toward some concrete activity. It seems the long-awaited expert mode still isn’t planned for @GeminiApp. Could we at least get decent usage limits with Gemini 4 Argon Medium?