Each signatory country must incorporate these paragraphs in its laws
Same as done for ABC weapons, war crimes, civil rights, and copyright laws
The Three Laws of Robotics by Asimov are clearly outdated but there’s still a lot of unwarranted hype about AI. Note it just very quickly copies general knowledge. It’s knowledgeable but not smart, wise, or moral. Guilty are those who plugged it in, fed it data, and didn’t throw out crappy results.
Present AI is completely on the wrong track. It’s good at stringing words. Not that it understands them. It looks at what word sequences are popular in certain contexts and then copies. But Generation Z isn’t taken by words. It has a visual mindset, learns TikTok speak. Also, AI can be used for evil.
Those liking fake, forgery, counterfeit, and fraud have a good time coming.
1. Just as ‘orders are orders’ is no excuse to do evil, nor is ‘AI told me so.’
2. Humans are responsible for what information they feed AI and to censor what results it produces. Immoral and false results must be corrected or discarded. The products must say which quality control and fact-checking procedures took place and which people or legal entities are accountable.
3. All AI produce must bear a clear audio or visual hallmark that can’t be removed and must be evident on every frame and detail of the output.
4. Actors who use AI for immoral gains, manipulation, deceit, theft, and destructive goals, or don’t disclose all information as stipulated in these regulations, must be isolated by the signatories of the treaty and prevented from getting AI service, hardware, software, data, feedback, and upgrades.
5. AI results produced by immoral actors must be clearly marked as such whereby the mark cannot be removed and must be evident on every frame and detail of the output. They cannot profit from their immorality.
6. Like packaged food must state its calories, any AI result should specify the human work hours AI here saved and the number of jobs it cost.
7. Also must be disclosed what percentage of the inserted data comes from men, Whites, English speakers, white-color workers, the West, Straights, people between 25 and 60, healthy and able-bodied people, etc. Just like nutrition should hold sufficient fiber, AI data must have enough diversity. Supremacy and inequality must be slowed and not be given more privilege.
8. The products must say if certain computer programs were used that have been employed on the battlefield or could be used in the war theater.
9. Just like cigarettes, AI output must carry a WARNING: ‘Don’t blindly trust or get intimidated by computer performance. It just amassed an impressive amount of information. Therefore, stay critical as AI answers will often lack perfection, intelligence, wisdom, and an ethical compass.’
10. In the case of a run-away bot, people at risk must be alerted instantly.
11. In school, children must be taught that people are not like computers but much more intelligent, and AI is not human, smart, wise, or moral. Computers might be treated with respect, but they never think or feel.