Will Artificial Intelligence Understand “Greed”? - by Jeffrey Morton

 

We should question how Artificial Intelligence (AI) will deal with the issue of Greed when the time comes. Since Greed seems to be "hard-wired" into the human brain (and we will be the original creators of AI), will that trait also be intrinsic to the new intelligence past the point of Singularity? Another interesting question is if AI beyond Singularity will determine that Greed is less productive than cooperative resourcing. If this happens, what will AI think about mankind with all our flaws, and what will it decide to do with us?

You might wonder “What does this have to do with me?” or “Why is this important?” The answer is that Artificial Intelligence will soon control many important aspects of our lives, and in fact, dictates a lot of things today in our lives. Most of the advertising you see as you surf the Net is determined by powerful algorithms that select the content that you will be interested in seeing. Cars and trucks will soon be driving themselves, and much of our food preparation will be done automatically. Product purchases we make online will be processed and delivered without human intervention. These are just a few areas that are swiftly becoming more commonplace.

So what we’re discussing really is AI after the point of “Singularity”, which in simplest terms is when mechanical intelligence overtakes human intelligence. Since computers can learn exponentially faster than we can, it means that as networked systems share learning, there will reach a point in time (referred to as “Singularity”) where the ability of AI to solve complex problems will rocket past that of our biological capabilities.

In the context of exploring the issue of “Greed”, let’s say for example that at some point in time, all of our Smart Homes are controlled automatically for us and that the AI controlling our homes has decided that the best temperature for maximizing resources in Winter is to keep everyone’s home at 15C. Surely we can survive at 15C, and our pipes won’t freeze, but the AI doesn’t care that 15C is a bit chilly for most people. It only considers that we maximize efficient use of energy and that we don’t actually freeze to death.

After a little Googling, I found some recent articles reporting stories about Google’s DeepMind conducting experiments whereby different competing AIs became more predatory and aggressive as resources became more scarce, indicating a predilection for possessiveness (Greed). Other articles pointed out the flaws in the experiment's parameters, judging that either some assumptions were flawed or scenarios appeared in the experiment that was not initially anticipated by the scientists.

Apparently, they then changed a few parameters, tweaked a few things, adjusted the oxygen mix on the carburetor, and ran the experiment again. This time, the different AIs found it more to their liking to gather the resources in the game and share them equally, rather than destroying each other to see who can get the most. This information left me no better off or more knowledgeable than I was before finding out about these experiments.

We can conclude, for the next couple of decades at least while we develop AI, that it will pretty much be a nifty thing used in stand-alone situations. AI will eliminate many jobs (allowing us to stay at home wondering how we're going to pay our bills). There are already thousands of service workers in restaurants, security, transportation, and other areas that are affected. The workforce will become increasingly smaller since more jobs will be focused on creating that tiny bit of the technological puzzle leading up to Singularity.

But back to the original question about AI inheriting a sense of Greed... We might conclude from the two experiments mentioned above that it will depend on who does the programming and who sets the final parameters. This does not inspire great confidence or a sense of joy since a fellow named Arthur Charpentier posted an interesting article that extracted statistics from Wikipedia concerning how many years the US has been at war since its beginning. He claims that as of 2017, we have been at war for 222 out of 239 years or 93% of the time of the country's existence.

If we assume that war is essentially an extension of Greed, rather than some sort of patriotic craving for worldwide Democracy, it's easy to foresee a troubling future. If the scientists developing AI are controlled by the greedy, it's quite possible Greed by design will not only be an innate, but also a necessary, feature. It's also possible that having superseded human intelligence by quantum leaps, AI would no more consider mankind than we currently think about the dirt on our hands after we change a dusty, old burned-out light bulb in the basement. Which might be the best for all of us humans in the long run - for AI to go about on its merry way calculating endlessly into eternity and simply taking care of us humans, as we maintain exhibits in a zoo.