Commissioner Charles H. Duel is reported to have recommended upon his retirement from the US Patent Office in 1899 that the Patent Office be closed, because “everything that can be invented has been invented.” Predicting the future of science and technology is a tricky business.
In 1952 pioneering scientists built a computer to play tic-tac-toe. In 1997 IBM’s Deep Blue Artificial Intelligence (AI) computer beat chess champion Garry Kasparov at chess. This was a major achievement.
Twenty years later, in 2017, Google’s AlphaGo (AG) AI program beat world champion Go player Ke Jie. This was perhaps an even more significant event, as this was previously thought to be impossible because of Go’s vast complexity. Later that same year Google’s AI computer program AlphaGo Zero (AGZ) beat AlphaGo. This should have been a wake-up call.
AG required massive amounts of Go game history, and some human help, to develop its Go-winning algorithm. AGZ required neither of these. Given only Go rules and a random play start, AGZ played against itself until it became unbeatable; neither human champion nor AG could defeat it.
Here is a graph of AGZ’s progress against human championship performance (lower dotted line) and AG’s performance (upper dotted line).
Source: Nature, October 2017
In less than 40 days AGZ taught itself more knowledge of God than had been attained in all of human history! The logic of AGZ’s play is not understood. It just wins.
On December 5, 2017, Google announced that it had generalized AGZ’s approach into a single AlphaZero (AZ) algorithm, that within 24 hours achieved a superhuman and super-rival-program level of play in Go and in other games.
Today’s AI can perform specific tasks such as playing games, face and voice recognition, driving a car and reading an X-ray. Tomorrow’s Super AI will not be limited to a specific task but will be capable of teaching itself in an ever-expanding range of applications, including robotics, or maintenance of a controlled society, or eventually anything it decides to do.
Most popular discussions of AI predict a happier future for man, one that is hopefully awaited and virtually unlimited. But Elon Musk, a well-known and very creative cookie (Tesla, Spacex etc.), is concerned about a point called the “singularity”, a concept originally formed by mathematician John von Neumann in the 1950’s.
At some point, AI will reach a condition where it will be totally beyond our control, and unstoppable. We will have lit a fire that we cannot put out. This will produce a great change in mankind, leading us beyond human to transhuman, and conquest of the stars; or, Musk’s concern, to extinction.
All observers agree the singularity will be reached. The only question is how soon. Leading AI proponent Ray Kurzweil of Google predicts runaway singularity in 2045, but this was before AGZ. Other estimates are much sooner (remember, few thought AG was even possible). Given the unknown nature and potential danger of the outcome, Musk is trying (unsuccessfully) to get the developers of AI to slow down before it may be too late.
As AI is enlisted to enhance AI, things will develop very quickly. In terms of learning, AI has humans totally outgunned. The size of its hardware, its processing unit, is not limited by the size of a human skull, and, operating many, many times faster than the human brain (at the speed of light vs. slower than the speed of sound), it is capable of doing in a week what would take humans thousands of years. It will be able to ever-increasingly teach itself, but teach itself what? This is real; this is scary.
If you are willing to consider what AI has to say about the bible, and what the bible has to say about AI, read on.
(Good, you’re still here) The jump from winning at chess to winning at Go involved a change in approach to the problem. Deep Blue just plowed through all possible moves and chose the best one. Go is too complex for even a computer to do this.
The seriously brilliant creators of AG and its descendants instead imitated the human brain by creating complex “neural networks” (that mercifully I will not even try to explain). As related above, a neural network is able to learn.
Every philosophical discussion of AI sees it as a continuation of Darwinian evolution. This is the framework that is used to put these developments into a historical perspective; place them in a big picture. AI is seen as the next step in evolution. I believe this is wrong.
Darwinian evolution is the wrong paradigm for understanding AI. Nothing in today’s AI events fits Darwinian theory: apes didn’t program humans and send us on our way.
AG was jump-started. Historical data, the learning program and the hardware (computer) for the program to operate in was supplied by humans. For AGZ and AZ, the learning program and the hardware were again supplied by us.
AI is the only example of human-level intelligence known to man, beside ourselves. Based on the totality of information that we now have about human-level intelligence, the clear implication is that intelligence requires a creator. In our experience, intelligence does not self-assemble. We may wish it did, we may fantasize it did, but the only solid evidence says no.
The AI learning programs did not self-assemble, nor did the computers nor did any additional required data. AI was created – we created it.
So what does this tell us? Based on what we actually know about intelligence, our hardware, software and innate information (e.g. DNA) must have been created, as the bible says (though not necessarily that way), not evolved, as is widely believed.
Actually, I should rein-in the above inference a bit. AI and the human brain are both neural networks, so the correct inference is that neural-network-based intelligence requires a creator. This leaves the door open for some other kind of non-neural-network intelligence, whatever that might be; for example, God. (So God, if he has a different form of intelligence, need not have a creator.)
Recall that no one knows what a post-singularity world is going to look like. But it turns out that we do have some information about how independent neural networks behave.
From human history, we know what neural network intelligence does when sent-off on its own into the world. According to the bible, it rebels against and tries to kill, its creator (see Jesus).
AI is learning in, and from, a modern world created by humans. Thus, in addition to its inherent nature, AI may be receiving a subliminal “human taint.” (Indeed, if possible, how could it not?) And human behavior isn’t always the best. (Kurzweil has said, “future machines will be human, even if they are not biological.” Has he looked at human history? He meant this to be reassuring, but I am not reassured.)
If the human pattern were to repeat, AI will rebel against and try to kill its creator – that’s us. It will also display a “if something can be done, it will be done” propensity; to what end I cannot imagine. Maybe not, but ours is the only documented case study we have.
This strongly suggests that Musk’s warning to slow down should be heeded. But, in his own words, “no one is listening.”
I know – I’m speaking matter-of-factly about something that sounds like far-out science fiction. Since at least Mary Shelley’s Frankenstein, how many times has the story been told of some being or creature or thing turning on its creator? How many times have we told the story of the unheeded warning?
We keep telling ourselves these stories over and over again, in one form or another, in one setting or another. The plot keeps resonating with us. I think this is because we know, deep in our hearts, that this is the true story, our story, as told in the bible. It is one of the things “written on our hearts”, as the bible puts it. The fact that we are about to attempt to execute this plot for real, with ourselves as creator-god, I find almost unbelievable. Yet here it is; we’re looking at it. This is getting a little too real for comfort. [Frankly, if it wasn’t for Jesus I would be horrified.]
The Bible also indicates that AI may be a signal to us that the end of days is near.
In the Book of Daniel, after being told about the harrowing “time of the end”, Daniel is instructed:
4 But thou, O Daniel, shut up the words, and seal the book, even to the time of the end: many shall run to and fro, and knowledge shall be increased. [Daniel 12:4 KJV]
Since many are to-and-fro-ing, and knowledge has been increasing, I used to think we were well into the time of the end, despite the fact that the world has not yet plunged into the chaos of that time (in which AI may have a role to play).
I now believe we may not be quite there yet. We now have on the horizon a fulfillment of this prophecy in the extreme. An increase in knowledge immeasurably beyond the sum of all previous human knowledge is on the way. AI is about to produce a world in which ‘knowledge shall be increased” is an indisputable understatement.
The bible doesn’t say “human knowledge” will increase, just knowledge. Maybe this unprecedented situation is why the meaning of the quoted passage in Daniel’s book has been “sealed-up” until now, the time of AI, the time of the end. A time that is fast approaching.