CoinTelegraph
Mysten Labs' Evan Cheng says AI needs deterministic protocols to set boundaries for agents operating in an adversarial world.
The argument puts predictability at the center of agent safety: clear, enforceable limits can help constrain what autonomous systems do when facing hostile inputs or environments. For the AI sector, trust depends not only on what agents can accomplish, but also on whether their actions stay within defined rules.
Frequently asked questions
-
What does Evan Cheng say AI agents need?
Cheng says AI agents need deterministic protocols to set boundaries for their actions in adversarial environments.
-
Why are deterministic protocols relevant to AI-agent safety?
They can provide clear, predictable limits on what agents are allowed to do, including when they face hostile inputs.
-
How does this argument connect to trust in AI agents?
The argument is that trust depends partly on whether an agent's actions remain within defined, predictable boundaries.
-
What challenge do AI agents face in adversarial environments?
They may encounter hostile inputs or conditions, making clear constraints on their actions important.
-
What broader issue does Cheng's view raise for AI systems?
It highlights that governing capable AI systems requires attention to both what agents can do and the rules that constrain their actions.
WatcherGuru
Crypto News
CoinDesk
TheBlock
CryptoSlate
WuBlockchain