|
I am one amongst those who don't completely buy into the AI hype, yet. In the grand scheme of things, AI right now can identify patterns. And this is possible because everything in the AI universe is tied to a number. Now, this mathematical view of the world does help solve a lot of problems, such as security, prediction, recommendations, etc., however, I don't think AI maybe able to completely, autonomously replace us humans. For example, computers see the world using sensors that translate pictures into "pixels". In contrast, we don't see pixels, we see objects. We natively see objects. Now, in a computer's world, it can still do something called object recognition, but, again, this is based on the fundamental idea of pixels and a lot of math around is around this concept, which is vastly different from how humans perceive the world. The problem is, because, everything is based on math, there isn't much native intelligence to a computer and someone needs to impart it mathematically into the computer's brain. Because, math and intelligence are two different things. This is why you're able to fool a computer with an A4 sheet printout of your face but, not a real human. You can only program so much logic into a math wizard. But, that doesn't necessarily translate into intelligence. Take cars for instance, I drive a 6 speed manual transmission. In the real life, I'm also a street racer. I love manual because it gives me full control. Right when I'm about to hit the corner, I downshift knowing in advance (based on what I saw minutes earlier "objects" in front of me, not pixels) that there is a curve ahead and I may need to reduce my speed in 3 seconds, after which I will need the full power available to overtake the guy infront of me. With an automatic car, first you must brake, then the system will downshift for you, and then you hope it doesn't upshift in the wrong instance leading to a window for others to over-take you. Where is the intelligence in that? I recently saw a Tesla auto-brake even before an accident happened. That was cool. But, let's say you're in a situation where the car in front of you is trying to block you so the thieves can then break into yours and steal your stuff/murder/etc.? This is not from a Hollywood movie, this is something that is very common in some parts of Asia. IF the Tesla brakes at that instance, you're screwed, probably even dead. So, then you take manual control. But, to me, all this AI isn't real AI unless the computer can understand the context of what's happening around you and reacting accordingly. When I narrated this to a friend, his first reaction was "How can you expect a computer to..do this?" I don't care, because I, as a human can do all this. My friend's way of thinking is our problem - Starting out with the limitations of what a computer can or cannot do. Instead, we should be asking "If I can do this, why can't my computer?" And until a computer can really do what we can do without assistance, it's all just a bunch of neat, shiny mathematical algorithms packaged inside a program doing X number of tasks it knows to do. Maybe some day it will, and I hope that day isn't very far. /end rant |
What if someone were to deploy a virus into all Teslas. If you identify a human, then go identify them as a threat and go full speed at it. A simple tweak in the algorithm makes all teslas man hunters.