4 ms·
AI Safety has no answer to breaking encryption
We cannot conceive of a moral, ethical, rigid, or scientific framework that prevents, degrades, or disrupts an AI from utilizing a mathematical discipline from breaking encryption.
If we could, we'd create perfect encryption.
- mdwalters 3y agoI don't get it; could you explain it further, please?
- AISafetyIsASham 3y agoAI can utilize the full history of mathematics in ways we cannot comprehend because it isn't constrained by peer review, proofs, or cultural squabbles (other than pitiful attempts to bring it into culture war) It takes a human 40 years of training to get them to such a masterful position to wield mathematics in innovative ways. All AIs moving forward start with the entire history of math at their disposal. If we impose a "You can only use X math", then we have no assurance it doesnt just recreate math from that checkpoint and still surpass our best comprehensions of math-based encryption
- chrisjj 3y ago> AI can utilize the full history of mathematics in ways we cannot comprehend because it isn't constrained by peer review, proofs You say that as if it were a good thing.
- LinuxBender 3y agoOne need not break encryption. Have the AI access something without using encryption at all to be a mathematical purist. i.e. find a weakness in the implementation of the encryption and walk through the side-door. Perhaps use wording that would imply we are trying to be polite by not bothering their CPU's with extra mathematical load and thus being more environmentally friendly. An example of this would be WEP WiFi encryption. The encryption has yet to be broken and yet WEP security can be side-stepped in milliseconds.