註冊並分享邀請連結,可獲得影片播放與邀請獎勵。

Jason Lowery
@JasonPLowery
Space Force Guardian & Engineer | Personal Account | Views expressed on this platform are mine alone, not the views of the U.S. Department of War
加入 August 2021
3 正在關注    160.2K 粉絲
The single biggest difference between humans and AI agents is the amount of actions they execute on a computer per unit of time. AI agents can look and act just like humans, but execute 10,000X more actions per unit time. that's the tell. so the key to defending cyberspace from AI slop or AI attacks is to target ACTION, not IDENTITY. trying to detect AI is a fool's errand; targeting its BEHAVIOR is the key. So if you want to deter AI from accessing your website/system, simply make it expensive in computational tokens to ACT on your website. Make every command or button cost a token, and suddenly it's 10,000X more costly for AI to act than a human. You can't stop AI, but you can make it extraordinarily expensive to use it. We already know this intuitively, because everybody who has ever run openclaw or their own personal AI agents knows that the single biggest LIMFAC is compute tokens. your AI is only as good as your compute token wallet. but AI compute is constantly getting cheaper and better, and some countries have far better AI compute than others, creating a major imbalance of power. so the key to long term success is to create some kind of universal, non-sovereign, geopolitically neutral compute token network. that is, we need some kind of network that accounts for a bunch of transferable tokens that are computationally expensive to produce. that no amount of AI compute can produce more of. we need a proof-of-compute token network. and we need it to be fully decentralized so that no single person or polity can gain permanent or centralized control over it. we need this network to be globally dispersed. and we need to figure out a way to distribute those compute tokens in a way that's fair, that gives no single person or polity an advantage over the others. perhaps in some kind of first-come first-served basis, where the users who provide the computational power are given the tokens and free to spread them out however they want. if we had this compute token network, the final step to saving ourselves against AI is to simply adopt the habit of making every online action cost some fraction of a compute token, so that it's orders of magnitude more expensive for AI agents to operate online than humans. the first nation to discover this compute token network and use it to its defensive advantage against the threat of agentic AI will win the 21st century. BTW that compute token network is Bitcoin.
顯示更多
0
166
2K
445
轉發到社區