3 ms·
I agree with the article that regulation will work in the near term while it is still expensive to train and operate such systems but for the long run we will n
by throwaway4aday 4y ago
I agree with the article that regulation will work in the near term while it is still expensive to train and operate such systems but for the long run we will need the AI equivalent of ad blockers that can be deployed by individuals to defend themselves against this type of exploitation. Arming everyone with their own personal advocate and security auditor AI is the only way to defend against such exploits. Unfortunately this opens another can of worms similar to how some ad blockers allow certain "good" ads through.
- roywiggins 4y agoDoesn't that basically devolve into my AI agent and your AI agent conferring and trying to mutually convince the other that they're a human before we actually communicate? Not sure that's a great future to live in.
- throwaway4aday 4y agoI don't see an alternative, at least if processing power and developments in ML continue to improve. Eventually you'll get AI clickfarms where people are just running hundreds or thousands of AIs out of a warehouse trying to scam people day and night. Arming everyone with their own AI bodyguard would prevent a lot of that from succeeding. The other options aren't too appealing e.g. only trusting face to face communication which basically means throwing away the internet and telephones and even hand written letters. Trying to regulate your way out will either lead to a war on AI where you're knocking down doors and arresting people who are trying to do legitimate independent ML research or you are just bombarded by offshore agents in countries with no regulation.