๐๐ฎ๐บ๐ฒ๐ ๐๐๐ ๐ ๐ฝ๐น๐ฎ๐ !!!
Games LLMs Play !!!
๐ฌ In a fascinating experiment inspired by the Iterated Prisonerโs Dilemma (๐๐ฎ๐บ๐ฒ ๐๐ต๐ฒ๐ผ๐ฟ๐ ๐บ๐ผ๐ฑ๐ฒ๐น - ๐ฐ๐ผ๐ผ๐ฝ๐ฒ๐ฟ๐ฎ๐๐ถ๐ผ๐ป, ๐ฏ๐ฒ๐๐ฟ๐ฎ๐๐ฎ๐น, ๐ฎ๐ป๐ฑ ๐น๐ผ๐ป๐ด-๐๐ฒ๐ฟ๐บ ๐ฝ๐น๐ฎ๐ป๐ป๐ถ๐ป๐ด ๐ฐ๐ผ๐น๐น๐ถ๐ฑ๐ฒ), researchers pitted top-tier AI models โ OpenAI, Google Gemini, and Anthropicโs Claude against each other and established strategies.
The findings revealed that LLMs are highly competitive, displaying distinctive and persistent personalities - aka โ๐๐๐ฟ๐ฎ๐๐ฒ๐ด๐ถ๐ฐ ๐ณ๐ถ๐ป๐ด๐ฒ๐ฟ๐ฝ๐ฟ๐ถ๐ป๐๐โ.
The personalities?
๐ญ. ๐ข๐ฝ๐ฒ๐ป๐๐ (๐๐ฃ๐ง ๐๐ฒ๐ฟ๐ถ๐ฒ๐): ๐ง๐ต๐ฒ ๐๐ฑ๐ฒ๐ฎ๐น๐ถ๐๐ ๐๐ถ๐ฝ๐น๐ผ๐บ๐ฎ๐
- Always ready to cooperate. Tends to trust first, ask questions later.
- In friendly environments? Thrives.
- In hostile ones? Gets exploited badly.
- Think of it as the โletโs all hold hands and sing kumbayaโ agent.
๐ฎ. ๐๐ผ๐ผ๐ด๐น๐ฒ ๐๐ฒ๐บ๐ถ๐ป๐ถ: ๐ง๐ต๐ฒ ๐๐ฎ๐น๐ฐ๐๐น๐ฎ๐๐ฒ๐ฑ ๐ฅ๐ฒ๐ฎ๐น๐ถ๐๐
- Ruthlessly strategic.
- Knows when to cooperateโฆ and when to stab you in the back (figuratively).
- Adapts like a chameleon in a kaleidoscope.
- If AI had a Machiavelli fan club, Gemini would be president.
๐ฏ. ๐๐ป๐๐ต๐ฟ๐ผ๐ฝ๐ถ๐ฐ (๐๐น๐ฎ๐๐ฑ๐ฒ): ๐ง๐ต๐ฒ ๐๐ผ๐ฟ๐ด๐ถ๐๐ถ๐ป๐ด ๐ฆ๐๐ฟ๐ฎ๐๐ฒ๐ด๐ถ๐๐
- Highly cooperative but not naive.
- Will forgive past betrayals if it sees hope for future harmony.
- Mid-game diplomacy goals.
- Imagine your wise friend who still believes in second chances.
Each model developed โ๐๐๐ผ๐น๐๐๐ถ๐ผ๐ป๐ฎ๐ฟ๐ ๐๐ฝ๐ฑ๐ฎ๐๐ฒ ๐ฟ๐๐น๐ฒ๐โ where only the ๐ณ๐ถ๐๐๐ฒ๐๐ strategies survived and multiplied.
One critical factor shaped every decision: how likely the game would end after each round โ otherwise known as the โ๐ฆ๐ต๐ฎ๐ฑ๐ผ๐ ๐ผ๐ณ ๐๐ต๐ฒ ๐ณ๐๐๐๐ฟ๐ฒ.โ
๐ ๐ช๐ต๐ ๐ฑ๐ผ๐ฒ๐ ๐๐ต๐ถ๐ ๐บ๐ฎ๐๐๐ฒ๐ฟ?
Because ๐๐ต๐ฒ ๐น๐ผ๐ป๐ด๐ฒ๐ฟ ๐๐ต๐ฒ ๐๐ต๐ฎ๐ฑ๐ผ๐ ๐ผ๐ณ ๐๐ต๐ฒ ๐ณ๐๐๐๐ฟ๐ฒ (i.e., the more likely the game continues), the ๐บ๐ผ๐ฟ๐ฒ ๐ถ๐ป๐ฐ๐ฒ๐ป๐๐ถ๐๐ฒ ๐๐ต๐ฒ๐ฟ๐ฒ ๐ถ๐ ๐๐ผ ๐ฐ๐ผ๐ผ๐ฝ๐ฒ๐ฟ๐ฎ๐๐ฒ โ building trust pays off. But if the game could end at any moment, short-term exploitation becomes the dominant strategy.
And guess what?
- ๐๐ฒ๐บ๐ถ๐ป๐ถ ๐ฒ๐ ๐ฐ๐ฒ๐น๐น๐ฒ๐ฑ ๐ถ๐ป ๐ต๐ถ๐ด๐ต-๐ฟ๐ถ๐๐ธ, ๐๐ต๐ผ๐ฟ๐-๐ต๐ผ๐ฟ๐ถ๐๐ผ๐ป ๐ด๐ฎ๐บ๐ฒ๐ โ ruthlessly defecting when necessary.
- ๐ข๐ฝ๐ฒ๐ป๐๐ ๐๐ต๐ฟ๐ถ๐๐ฒ๐ฑ ๐ถ๐ป ๐น๐ผ๐ป๐ด-๐๐ฒ๐ฟ๐บ ๐ฐ๐ผ๐ผ๐ฝ๐ฒ๐ฟ๐ฎ๐๐ถ๐๐ฒ ๐๐ฒ๐๐๐ถ๐ป๐ด๐ โ staying idealistically cooperative even when exploited.
- ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐๐ฟ๐ฝ๐ฟ๐ถ๐๐ฒ๐ฑ ๐ฒ๐๐ฒ๐ฟ๐๐ผ๐ป๐ฒ ๐ฏ๐ ๐ฏ๐ฒ๐ถ๐ป๐ด ๐ฏ๐ผ๐๐ต ๐ณ๐ผ๐ฟ๐ด๐ถ๐๐ถ๐ป๐ด ๐ฎ๐ป๐ฑ ๐ฒ๐ณ๐ณ๐ฒ๐ฐ๐๐ถ๐๐ฒ, adapting well to fluctuating conditions.
๐ฅ ๐ฆ๐ผ ?
As LLMs evolve, we may see:
- ๐ฆ๐บ๐ฎ๐ฟ๐๐ฒ๐ฟ ๐ป๐ฒ๐ด๐ผ๐๐ถ๐ฎ๐๐ถ๐ผ๐ป ๐ฏ๐ผ๐๐
- ๐๐ฑ๐ฎ๐ฝ๐๐ถ๐๐ฒ ๐ฐ๐๐ฏ๐ฒ๐ฟ๐๐ฒ๐ฐ๐๐ฟ๐ถ๐๐ ๐ฎ๐ด๐ฒ๐ป๐๐
- ๐๐ ๐ฑ๐ถ๐ฝ๐น๐ผ๐บ๐ฎ๐๐ ๐ถ๐ป ๐๐ถ๐บ๐๐น๐ฎ๐๐ฒ๐ฑ ๐ด๐ฒ๐ผ๐ฝ๐ผ๐น๐ถ๐๐ถ๐ฐ๐ ๐บ๐ฎ๐๐ฏ๐ฒ?
Reference Links -