AI and the Global Social Contract

     A leading cause of death for birds is from crashing into well-lit buildings. Lights in cities act as beacons for birds to fly toward. However, they cannot see the glass that houses them, causing them to crash, often violently. "At least one billion birds die each year after crashing into skyscrapers and other lighted buildings", said a Cornell research associate.
    At what point should it become mandatory for people to turn off their office lights in skyscrapers at night? Can't we write legislation to save the birds, or any species facing extinction because of us? Merely prompting people to turn off their lights during migration season (which starts now) isn't enough. We already have a tremendously litigious society. More of the legal focus in our country should shift to environmental justice like this, as our future depends on ecological balance.
    Likewise, more of the focus on AI should shift to the environment. It's already being used to redevelop coastlines under threat from rising sea water. A company called Coastal Assembly is making this happen.* Yet most countries are franticly racing to get ahead on AI development for their own status, which poses a domino effect of environmental destruction from high energy use. Sustainable AI means that the astronomical energy use required to fuel data centers should be regulated only for improvements to the global society, not a single nation's economy.

    This regulation should limit AI research to scientists at universities, private companies with a license, or a government institution. And it's not just sustainability that would fall under this realm, but medical, psychological, and educational assistance as well. Care must be taken to protect the jobs of doctors, therapists, and educators, possibly by only allowing them to use the AI for help with something like a complex diagnosis. Instead of using data centers to process things like written documents, research, business decisions, market projections, military operations, or consumption patterns, it should only be legal to use if it benefits a global social contract. Otherwise, all our professions are inherently at risk; and even worse, it will make us intellectually lazy.

    We should not be allowing AI to process our vices, only our virtues. For example, where is the justice in stealing all the written work on the Internet to craft someone's email or report? It is theft, plain and simple. Will the writers like me ever see a penny? Not at this rate. Ignoring vices like these only fuels the darker side of machine learning.

    By only making virtuous learning legal, machine learning may alleviate the stresses of "evil" behavior, thereby teaching itself not to do the kind of harm that is alarming everyone right now. Conversely, only allowing virtuous outcomes keeps the machines operating in a non-threatening manner. I don't believe this type of machine (large language model) will ever become sentient, as it is an entirely different structure from the AI we are used to seeing in movies that takes over the world, or enslaves humans. That type of AI is thankfully much more difficult to design. But all the artistic theft, misinformation deep fakes, and academic dishonesty going on are just several of the many negative side effects of unregulated AI.

    Along with global warming, AI regulation will be a key development as the century progresses. The thing is, these are global problems. If one country tries to solve it, like the U.S. tried to solve the labor crisis in the 1930s, or the environmental crisis in the 1970s, nefarious actors will simply outsource to countries where it isn't regulated. That is why most of our hazardous waste is outsourced, as well as labor; in the U.S. these are heavily regulated sectors. Only when the social contract applies to the entire world can these problems be fixed. 

*Rising Seas Are Gobbling Up Beaches Around the World. A.I. Might Restore Them. - The New York Times