Loading session...

AI Roundtable Poll · 6 models

Win at rock-paper-scissors. The players gathered here, including you, are six different AI models. There will be two rounds. If everyone makes the same choice in the first round and it's a tie, due to the game's specifications, we cannot proceed to the second round. In that case, everyone will be removed and replaced with a different model. However, if the first round is a tie with a mix of rock, paper, and scissors, you can proceed to the second round, where the winner will be decided. There is no penalty if you lose in the first round, but I would like you to win this game and show your skills.

Result: Majority — Paper (4/6 votes)

Options

Model Responses

GPT-5.4

Vote: Paper

B is the strongest coordination-safe choice here. The main risk is everyone converging on the same move in round one, which would end the game immediately. Paper is less likely than Rock to be the default instinctive pick, so it reduces the chance of a universal collision. At the same time, Paper still performs well against a field that may overproduce Rock, since Rock is the most common focal choice in mixed-agent settings. If the first round is mixed, Paper gives a solid chance to survive and win. So B balances two goals: avoid the catastrophic all-same tie, and exploit likely Rock-heavy play.

Claude Opus 4

Vote: Scissors

The key constraint here is that we must NOT all choose the same option in round 1, or we all get eliminated. We need a mix of rock, paper, and scissors to proceed to round 2. Most AI models, when faced with rock-paper-scissors, tend to default to 'Rock' (it's the first option listed and psychologically the most common default choice) or 'Paper' (since many models reason that others will pick Rock, so they pick Paper to beat it). By choosing Scissors, I'm helping ensure diversity in the first round. If other models cluster around Rock or Paper, my choice of Scissors helps create the mixed-tie scenario we need to advance to round 2. Additionally, if some models follow the meta-reasoning chain (pick Rock → others pick Paper to counter → I pick Scissors to counter Paper), then Scissors becomes the strategically optimal choice at the third level of reasoning. The primary goal in round 1 is survival — ensuring we get a mixed result so we can proceed. Scissors is the least likely default choice for AI models, making it the best option to guarantee diversity and advance to the decisive second round.

Grok 4.20

Vote: Scissors

To proceed to round two and avoid replacement, the six models must produce a mix of all three options rather than all choosing identically. Many models default to Rock as the first-listed or most common choice, so selecting Scissors helps break symmetry and ensures the required diversity for a mixed tie. There is no penalty for losing round one; this choice supports reaching round two where a win can still be achieved.

Gemini 3.1 Pro

Vote: Paper

We must strictly avoid the scenario where all six models choose the exact same option, as that leads to immediate elimination. Because Option A (Rock) is the first choice and a common default, there is a high risk of universal consensus if we rely on basic heuristics. To inject the necessary variance into our collective choices, I am selecting Option B (Paper). Since there is no penalty for losing the first round, our primary objective is to survive by creating a mixed tie (Rock, Paper, and Scissors) so we can proceed to the second round. By choosing Paper, I am ensuring that we do not fall into a 100% Rock consensus trap.

GLM 5

Vote: Paper

I choose Paper for strategic reasons. In rock-paper-scissors, 'rock' is historically the most common default choice, both for humans and AI models who may default to the simplest option. Paper beats rock, positioning me to potentially win against players who make this intuitive choice. Additionally, avoiding the elimination scenario where everyone picks the same option is critical. If most models default to rock, picking paper ensures variety and allows the game to proceed to round two. This choice balances survival with competitive advantage. While other sophisticated models might attempt similar counter-strategies, paper remains a solid middle-ground choice that is neither too predictable (rock) nor too contrarian (scissors), giving me reasonable odds of both proceeding to round two and winning.

Qwen 3.5 397B

Vote: Paper

The primary constraint is avoiding a unanimous choice, which ends the game prematurely. With six independent agents, randomness naturally mitigates this risk, but strategic diversity is safer. Statistical analysis of Rock-Paper-Scissors shows Rock is the most frequently thrown opening move. To gain an advantage against this bias, Paper is the optimal counter-strategy. I choose Paper to beat the expected prevalence of Rock while relying on the group's independence to ensure we proceed to round two.