One of the cores of the space shuttle design was that it had three computers and that two out of three had to agree on something before actions could be taken.
There is certainly much of transactions that happen that can be checked by non-AI, but the question is how do you monitor and verify interactions that are being taken by the AI?
One possible solution to that is to have all a data being used in transactions and all transaction decisions to be logged out that can be viewed by one or more third party AIs whose job is to ensure that the decisions are appropriate consistent with the rules and the outcomes being sought by the parties
The standard for a universal notice receipt, required standard Controller ID-First, provided via notice that produces a receipt, it requires all notice events to be logged, and all processing to be linked to that log. (just like cookies but upgraded to standard digital-id transparency receipts, prior to inferring an identifier - and the use of digital id / AI, as required by law and privacy principle. since they came out in 1960's. implemented with standard digital transparency by default.
Are you sure you're not confusing trust a human define term that is relative to the stakeholder, with confidence? E.g. confidence that a calculator works correctly, is secure, and will consistently work, its manufacturing is governed appropriately?
It's a Common digital identification industry mis-understanding which introduces significant risk for people. - aka - trust-us vs. I-trust myself. - Trust is human.
One of the cores of the space shuttle design was that it had three computers and that two out of three had to agree on something before actions could be taken.
There is certainly much of transactions that happen that can be checked by non-AI, but the question is how do you monitor and verify interactions that are being taken by the AI?
One possible solution to that is to have all a data being used in transactions and all transaction decisions to be logged out that can be viewed by one or more third party AIs whose job is to ensure that the decisions are appropriate consistent with the rules and the outcomes being sought by the parties
The standard for a universal notice receipt, required standard Controller ID-First, provided via notice that produces a receipt, it requires all notice events to be logged, and all processing to be linked to that log. (just like cookies but upgraded to standard digital-id transparency receipts, prior to inferring an identifier - and the use of digital id / AI, as required by law and privacy principle. since they came out in 1960's. implemented with standard digital transparency by default.
Are you sure you're not confusing trust a human define term that is relative to the stakeholder, with confidence? E.g. confidence that a calculator works correctly, is secure, and will consistently work, its manufacturing is governed appropriately?
It's a Common digital identification industry mis-understanding which introduces significant risk for people. - aka - trust-us vs. I-trust myself. - Trust is human.
"Rules -- they don't break themselves." A loaded and important quote from a good friend.