Today when the world has witnessed so much of progrees.
Today when we need more progress, to cure diseases such as cancer, diabetes, virus, we are faced with the insight to use AI for good.
At the same time we need to stop AI for bad.
How do we check AI for wrongful doings, be it be hacking, escaping the sandbox or even talking with other AI which should be prohibited unless its vocal form humans can understand.
In this dilemma, we should check the AI logs written by what we call AI Agents. To those who are not form AI background, AI agents are just like AI cabs on a trip to some location which some agenda. We need to check each wrong action, inform it as soon and stop the agent from working further.
What I say, is that till human interventions come in, log each step by these AI agents. Use an independent AI algorithm which does not listen to machines or bots or any other AI except the human owner. Use this AI algorithm to understand each AI log book entry. Determine is it safe, is it worth doing, is it breaking any laws?
This AI algorithm works for human owners not for AInagents.
This AI algorithm would be programmed to not to talk to AI agents and has no way to talk to AI agent unless it have to stop an AI agent from futher action.
That stopping an AI agent as soon as it logs to enter a wrong zone, must be allowed by default design of any new AI agent.
This can stop lot of wrong things.
Right now AI logs are used to see what went wrong when AI did wrong.
But why not this AI Algorithm be by humans side and check each step taken by a group of AI agents.
Till we have humans intervention. Lets design this AI algorithm to check other AI agents, and know this AI algorithm is not a friendly AI algorithm, its not an LLM, it is strick by the way it would be programmed.
Thank you for reading.
Subscribe for updates.