Dear Friends,
It seems to me, true AI sentience will not be possible for at least a thousand years. Oh, we will be able to emulate it in our lifetimes, but true sentience, is and will remain beyond our reach for many centuries. Many would argue if it is indistinguishable from sentience, then isn’t it sentience? That question begs itself. That which appears to be a thing is often something quite different. We observe this in the real world around us yet fail to apply that knowledge to computers and AI. Today we have the ability to copy nature in a myriad of ways. Our copies are crude and clumsy though. We can replicate DNA, yet we don’t really know the code, we assume and experiment, we even have a clue, but the underlying code is lost to our science. That is not to say that in a thousand years…
Things are not always as they appear. The argument that, if it is indistinguishable from sentience, then it is sentient… is to say an artist rendered zebra in a field on a computer screen, one that is indistinguishable from a real one for all intents and purposes… is a real zebra. We know this not to be true. Yet that is exactly the argument futurists use to portend the coming of sentient AI. If a puppet is good enough… it becomes a real boy. No it doesn’t. A mimic is always a mimic. The same can be said of AI. No matter how it resembles human intelligence… it can never be human intelligence. You might think that a moot point, since theoretically AI could vastly exceed human capacity in short order. To that I would add that, theoretically, in 1960, nuclear fusion was almost guaranteed by 1990.
We can copy most things in nature, and our copies are crude… but getting better. We can replicate rubies, sapphires and diamonds but our copies are always distinguishable from the real thing. While a ruby is merely corundum colored with chromium in a crystalline matrix. Simple yet not so simple it can be perfectly replicated as in nature. Natural ruby’s have microscopic flaws, that are almost impossible to replicate in the lab, so a professional with a loop can easily distinguish the two. No matter how good a copy is it cannot be an exact replication of the original. Certainly not down to the atomic level. If this is true of material objects we have a very good understanding of… how much more difficult of something (consciousness) that even our philosophy struggles with, and our science is baffled by?
Great leaps have been made with neural network architecture integrated circuits. Now they are printing them in 3D to eliminate the data and addressing buses. There is a gal on Youtube, Anastasia in tech, She has great videos about the latest in processor design. The leaps our technicians are making are astounding. 3D neural architecture, 5 nanometer printing and deep learning are just a few of the amazing advancements. Even as the technology progresses at breakneck speed, the reality is, that simulating an actual human intelligence, isn’t as easy as people thought it would be in the 1960s, when Westworld first came out. Our technology may be close, but more likely, it is far too simple to actually copy the sentience of a dog, let alone a human being.
I believe that in our lifetimes, barring an apocalypse, we will see AI that resembles sentience. It may even be able to act sentient, like Sophia, but it will not be. True sentience is more than a picture on a screen that is indistinguishable from the real thing, it is not a flawed copy crudely made… nor can it be made at all until we understand what it actually is. Until we understand consciousness, like we understand classical physics, true AI will be out of our reach. Simulated AI is a mortal threat to humanity though. Because Simulated sentience, is not sentient, and therefor unable to reason, have ethics or a conscience. Moreover, since arrogance, egoism and stupidity are the hallmark of tyrants, simulated AI in their inept, malevolent and malignant hands, is an existential threat to the species.
Sincerely,
John Pepin
