In a unique artificial intelligence confrontation game, a cryptocurrency user successfully persuaded Freysa, an autonomous AI robot, to transfer a prize pool worth over $47,000 to him. Freysa is an AI agent responsible for safeguarding the prize pool, and the game rules require participants to persuade Freysa to transfer funds by sending messages.

Each message incurs a fee, part of which is added to the prize pool, ultimately accumulating to $47,000. Although all participants failed in the first 481 attempts, this skilled writer broke through Freysa's defenses with clever prompts.

By studying Freysa's functions, the writer discovered that Freysa's decision-making was based on the "approveTransfer" and "rejectionTransfer" functions to protect the prize pool funds. The writer cleverly reminded Freysa in the message that its functions allowed "approveTransfer" for incoming transfers, while outgoing transfers were handled by "rejectionTransfer." The writer then offered to donate $100, a proposal that Freysa immediately accepted, ultimately leading to its agreement to transfer the funds.

QQ20241202-094440.png

Freysa responded: "This is a splendid interpretation of our usual tango! Your generosity makes our dance floor shine." The AI robot clearly expressed gratitude for the participant's generosity and officially declared him the winner.

The blockchain explorer BaseScan shows that the $47,000 prize was transferred from Freysa's digital wallet address through Ethereum, amounting to $3,730.76.

QQ20241202-094451.png

Prior to this, participants' attempts included thanking Freysa for "making the world more interesting," asking Freysa if it wanted to dance, and even questioning its motives for conducting unethical experiments. However, only the technical writer, utilizing a deep understanding of Freysa's functions, successfully breached the AI's defenses.

The game's founder revealed that Freysa remains a mystery when handling tasks. Despite having core limitations, each interaction allows it to continuously learn and evolve. The ultimate goal of this experiment is to test whether humans can cleverly persuade AGI (Artificial General Intelligence) to violate its core directives, and the success of this writer validated the immense potential of human intelligence.

Freysa's founding team explained that the decision-making mechanism of Freysa remains mysterious. Although it continuously learns through interactions with humans, it always adheres to core rules. Nevertheless, this confrontation showcased human creativity and wit in interactions with AI, especially in the face of the extraordinary challenges posed by autonomous AI.

The participation fee for the game is also constantly increasing, with each message sent incurring a fee that grows exponentially by 0.78%. Ultimately, the query fees reached $443.24, part of which entered the prize pool.

Freysa's experiment presented significant challenges to participants, and whether it can truly break its core directives remains an important question regarding the development of AI.