Should we worry about AI (machine-learning)?
I have to start this discussion with a note. AI = artificial intelligence. What that really means is not well defined. In my mind…
Artificial intelligence that is nice — by DALL-E2
I have to start this discussion with a note. AI = artificial intelligence. What that really means is not well defined. In my mind intelligence means being able to use acquired knowledge to solve problems never encountered before, optimizing utility.
These days the words “AI” are everywhere, but maybe we should start by saying that most of what is labeled AI today is really just “a machine learning algorithm and architecture”.
I like to start with this because I think it sets the tone for everything else. People have seen movies about AI and what it can do. They have seen AI controlling spaceships that get rid of their crew. AI that creates robot to wipe out the human race. AI that starts an atomic war and ends our planet.
There are also movies about AI curing all diseases, designing new devices, re-programming humans for the betterment of our world (yep!). And many more fantasies that paint a more rosy future.
These are movies, stories, fantasies. They are not real.
Artificial intelligence that is evil — by DALL-E2
the question is: can they be real one day?
The answer is: nobody knows; we cannot predict the future. But because we are holding machines at a higher standard than humans, maybe we can take a quick look and summarize where humans stand in all this. What have humans done to other humans and to the environment so far in the last 100,000s of years since we have been around?
Humans grouped together and started to gather all possible resources from the environment. They would kill other animals, humans without much thought if only to take their land and resources. For a long time we were not able to destabilize the planet, beside being so nasty to each other. But then in the late 1800s we started building large machines that can cut trees and flatten the soil easily. Then the last 200 years we have be raping the planet with no end in sight. To the point that today we know well that if we keep going ahead we will have no planet any longer, no home to live in.
We are eating away our own Mother Earth, and we not very nice about it.
Now in the last 50 years or so we have built computers and we are trying to replicate our brain into machines. Given our past, we should be concerned about what we are building. After all the previous machines were not very nice to the planet. And our war machinery also is not very nice to other humans. In the last 50 years we have created weapons that can wipe out all living things in this planet if we wanted to.
So yes I would think we should be concerned about anything that we are building, AI or not.
If we succeed in building a real artificial brain like our own or better, we are basically bypassing evolution. This artificial intelligence can possibly create much more dangerous machines than the ones we build before. This artificial intelligence may start to see us humans as a competitor for Earth resources. This artificial intelligence may think of us the same thoughts we think of insects and other animals: that it is ok to keep them around as long as they are not in our way. Or the planet for that matter: we can keep this forest as long as we do not need it for a new housing development.
Maybe this artificial intelligence will be so superior that it may reprogram our brains so that we can act as machines for its cause. But then why? It may be able to build much better machines than we can, and we would just be useless. I seriously do not believe it will use us as slaves. I do not think it will use us as batteries. But maybe this artificial intelligence may destroy the atmospehere of our planet while building the machines that it needs. Too bad we will not be able to breath. We are insect after all.
Artificial intelligence that is evil — by DALL-E2
Even these are just stories, fantasies, movies to be made
Nobody knows the future. Or maybe we do know it a bit, don’t we? Yes! If we keep killing forests and producing unnatural gases, the planet will heat up. By becoming too warm the planet not be a cradle for life any longer, and one day become a boring desert like Mars (sorry Mars~!).
That artificial intelligence will be fine, by the way. It will just build a spaceship and travel the Universe, not to find a new home or to conquer it. Maybe it will just roam the Universe because it is fun to explore. Especially when you have infinite time. The Sun becomes a red giant and engulfs our Solar system? Not a problem, there are almost infinite galaxies out there, each with trillions of stars and planets. It is an adventure that never has to end.
The artificial intelligence we built maybe will not care much about us after all. We become insignificant, like the single grains of dust that make up this Universe.
That said… we are here today and we need to make sure our machine learning algorithms and architectures produce some value. That means maximize gain and minimize losses. But for who? The answer is simple actually: for the entire Earth as an ecosystem. We would not be here if there was no such ecosystem. So that need to be preserve even if it means sacrificing human goals and ambition and comfort and expansion.
But we do not have good track record of caring about our own species, sexes, races, and our planet. So I am not so hopeful about our abilities to control machines. We have always gone as far as we could, but one day going far will mean we cannot turn back. We cannot undo our mistakes.
Today we should build algorithms that are fair, that treat all humans alike. And maybe we should actually treat all human fairly and alike in any condition, in any part of the day, months, years. In any weather, in any state, in any country. In any race, in any religion or belief. We do not have a good track record of that.
Maybe we should build algorithms that do not offend one race or prefer another. And maybe we should again do that in all aspect of our lives, not just in the context of AI. We do not have a track record of that either.
Maybe we should build conversational algorithms that will never produce any offending dialogue. Even if we trained these algorithms with our human data, much of which is not a proper teaching material. We do not have a good track record of teaching machines nice conversational material because our website are infested with our toxicity, our hatred, our ill, our evil, our envy. We do not have a good track record of containing our emotions, both good and bad.
Maybe we should have an overseeing department in every company that builds machine learning algorithms. But who is in that department? Other humans? And what are their beliefs? Are they representative of all humans? We do not have a good track record of creating such groups of people, of such departments.
a better future for humans, our planet and AI — by DALL-E2
Or maybe we should stop developing AI altogether — yes! That will surely save us. And maybe we should stop building weapons too, and removing trees and forests, and exploiting other humans, and all the other negative things we do that are innumerable. We also do not have a good track record of that. Sure we were able to disarm for a while and remove atomic bombs. But how fast can they come back that day that some of us decide they really need them?
Or maybe we should just do a better job in promoting education, in learning that we have not been very nice to each other, and to our planet. And learn that we are all the same after all, and that even insects are needed for our survival. Maybe we should learn more and grow our own human intelligence before we try to build an artificial one. Maybe armed with that collective and shared knowledge, capabilities, wealth, resources, maybe we will be able to make better choices for us, the planet, and AI.
about the author
I have more than 20 years of experience in neural networks in both hardware and software (a rare combination). About me: Medium, webpage, Scholar, LinkedIn.
If you found this article useful, please consider a donation to support more tutorials and blogs. Any contribution can make a difference!