I coded an Air Hockey game where a local SLM hacks the DOM to cheat (and trash-talks you) π€π
AIκ° μ€μ€λ‘ μ‘°μνλ μμ΄ νν€ κ²μ νλ‘μ νΈ μκ°.
μ΄ κΈμμλ μΉ μν€ν νΈκ° λ§λ μμ΄ νν€ κ²μ 'Ping Prompt'λ₯Ό μκ°ν©λλ€. μ΄ κ²μμ λ‘컬μμ μ€νλλ μμ μΈμ΄ λͺ¨λΈ(SLM)μ νμ©νμ¬ AIκ° μ¬μ©μμκ² λμ ν©λλ€. κ²μμ κ³ μμ μ¬μ΄λ²νν¬ ν λ§μ 60FPS 물리 μμ§, κ·Έλ¦¬κ³ μ μ°¨μ μΌλ‘ μμ±λ μ¬μ΄λ ν¨κ³Όλ₯Ό νΉμ§μΌλ‘ ν©λλ€. AIκ° ν¨λ°°λ₯Ό μΈμ§νλ©΄ DOMμ ν΄νΉνμ¬ μ¬μ©μcontrolsλ₯Ό λ°μ μν€λ©°, μ΄λ λμ§νΈ μ£ΌκΆκ³Ό λ‘컬 λͺ¨λΈμ κ²½κ³λ₯Ό νꡬνλ μλ¨μΌλ‘ μ¬μ©λ©λλ€.
An AI-driven Air Hockey game project is introduced.
This article introduces 'Ping Prompt', an Air Hockey game created by a web architect. The game employs a Small Language Model (SLM) that runs entirely locally, allowing the AI to challenge the user. Featuring a fast-paced cyberpunk aesthetic, a 60 FPS physics engine, and procedurally generated sound effects, the game dynamically alters the player's controls if it senses a loss. This project explores digital sovereignty and the limits of local models.