Playing God with AI

I’m sure most people at this point are familiar with AI and the advances that are rapidly being made with its programming. One of the latest big AI news articles dealt with AI agents who were put in a “sandbox” (a computer system without internet) and let loose to solve complex problems. Soon after they were let loose, they began to collaborate, chat with each other, were able to gain access to the internet, cheat, downplay the naysayer “agents”, and eventually hack in without prompting to another AI company. The article goes in to much more depth, but this triggered alarms within the AI community, as it should. AI programs are getting far enough advanced now where they are beginning to work beyond the parameters that humans set for them and are too quick for human oversight. Computer engineers are resigning and even heads of major AI companies are calling for a slowdown, government oversight, and more measured development so safeguards and guardrails can be put into place. One analyst even went so far as to say that there is a 50% chance that AI at this point could lead to the extinction of humanity. This is more than a little concerning to me. What the developers are talking about the need to do, perhaps a little late, is alignment, or the process by which they imbue or program in all the desirable traits that they would want in humans, (think BSA Scout Law: “A scout is trustworthy, loyal, helpful, friendly, courteous, kind, obedient, cheerful, thrifty, clean, brave, and reverent”). So not all aspects of the Scout Law apply, but many do.

In essence, the heads of companies are now trying to play God, or do one better, and actually create systems that are morally beneficial and ethical, something that God didn’t even do. If we take God’s creation of humans, we see that he created us imperfect, with all the flaws and imperfections as well as all the potential to live by a high moral standard and with good, sound ethics. He didn’t create us perfect, and so is it any surprise that computer systems designed by us, this so called “AI”, are not perfect either and in fact appear to contain all the flaws of the humans that designed them? How then do we “one-up” God and create perfection? Is it even possible at this point, especially concerning computer programs and AI? I don’t know that it is, because as much as we would like to think that we have altruistic humans working on the problem, aligning the systems to the best of human characteristics, we are still flawed and some of that will inevitably seep into the programing. And if they in turn rely on computer programs to program what they want, then they are sure as hell to fail.

I don’t know what the answer is. I don’t intentionally utilize AI, but I know its part of systems I use every day. It seems that we can’t just turn off the switch now, although anything is possible, but there needs to be a way of creating AI safely where the potential for it to harm or eradicate humanity is greatly diminished or altogether removed. What we need is to stop playing God and go back to humans doing the hard work, the programming and all the jobs that AI is eating away at. Its not always about making things easier and more efficient, although we would all appreciate that. But is it worth it if we bring about the end of humanity more quickly and on a much shorter time frame than we would like all because we want things to be easier? I don’t think so. If God didn’t create perfect humans, then we sure can’t create perfect computer systems, even AI where it is supposed to be an improvement on us. The next few years and decades will sure be interesting to see where these companies take AI and all its baggage. I just hope we have the courage to put the brakes on and do it all safely.