We’re in serious yoghurt. “Unfettered AI (Artificial Intelligence) will (well, could) wipe out humanity.” see here, for details.
The time is somewhere between 2040 and “in a few centuries”. see here. All of this is coming from VESFs (see previous post), some earned, some self-proclaimed. Some, searching for government grants. (Sell the sizzle, not the steak, right?)
Cynical, sure, why not. How about, “Watson can unlock the vast world of unstructured data” here. My “government math” calculates that to be about “half” right. Ok, enough.
Back to basics. Computers (the engine behind and driving all of AI) are really good at calculating & moving data around. It’s impossible to really conceptualize (at least for me) how fast. See an earlier post, Computers, People: Why can’t we just, well, talk? for a way to visualize that speed.
“Us folks” OTOH, are able to think and make judgements, albeit sometimes flawed, on very incomplete information. Recently, we were having dinner with our grandkids, and I asked Sarah (10), if she would get a folder that I had put on the piano bench. It wasn’t on the bench, but on a music stand by the piano. No instructions, she just looked around, saw it and brought it back. Didn’t even mention that it was not on the bench.
Try that with a robot. Not impossible, for sure, but it would take a lot of design work, programming, and testing — to say nothing about the physical part of walking searching, recognizing, etc., that would be needed. Robots, computers need specific instructions — very specific!
Getting computers, mechanical devices, and people to work together effectively is tricky, but as you can see here, (Tesla video) amazing progress has been made. But understand, that compared to activities that we do everyday in an unstructured world, all of those processes are very simple — even trivial.
As to AI, check out Michio Kaku here. The operative line is, “The current state of AI is that of a retarded cockroach”. A bit strong, good for headlines. Still, take over the world? Have to wait awhile. Send me $10, I’ll give you 10 to 1 odds it won’t happen.
So much effort is spent in trying to simulate the human brain. ‘We just need more speed and more data”. Maybe, but check out those old movie clips of the earliest airplane attempts. They had flapping wings. How’d that work out? Trying to simulate nature isn’t always the best approach. All of us learn a set of complex tasks (walking, speaking a language or two, etc.) without formal classes or government grants. Our brain is more a learning machine than a computer.
Isn’t it more reasonable to think that our “learning machines” will do just that? Figure out how to use the enormous power and potential of computers and robots, to our advantage? We have lots of time.
Then there is the current problem of what to do about social media and our kids (us, too). But that’s another story.
Later.
