"Deep Learning" AI is Total BS
In my Crush The Street interview that I recently posted, one of the topics that I discussed was artificial intelligence. The backdrop of this discussion was the recent pullback in Tesla shares, due to a fatal accident involving a Model X.
This accident comes against the backdrop of an Uber driverless vehicle tragedy in which a pedestrian was killed. I also wrote about this incident in a prior Steemit post.
The natural response to these incidents is that driverless AI technologies don't work in a real-world environment, and that is my assessment as well. Regarding the Tesla incident, I believe accidents are occurring because people are abusing the Autopilot feature, which is a driver-aid, not a driver-less technology.
But the greater issue is that AI is a misnomer.
Artificial intelligence is intelligent only within a defined framework. For instance, let's say you're playing Madden football. I'm sure the AI technology in that game is excellent, meaning that it learns how to play efficiently in dynamic situations.
But that dynamism is predefined: you can't then take that same AI and expect it to learn a different sport, such as basketball or baseball. Humans can because we adapt to any situation. AI technologies, in sharp contrast, must be spoon-fed the parameters and functionalities.
Otherwise, AI technology is total BS, just like that Sophia robot which I mentioned in my InvestorPlace article featuring Tesla stock.
You have to walk before you run
It's interesting to me just how much the general public (and I'm including myself in this broad statement) has been deluded with the "deep learning" ruse.
Computers only "learn" in predefined contexts, as I mentioned earlier. Indeed, all computer programming is based on ever-expanding binary codes -- ones and zeros. Yeah, the codes are getting more complex, but that doesn't change the yes/no logic. AI doesn't know "maybe," or non-verbal cues such as hesitation.
Also, consider this: a camera that can focus as fast as the human eye is still a theoretical concept. Our eyes are remarkable organs. Assuming normal functioning, they have zero lag between focusing in-depth on a particular object and on the full spectrum.
Just try it -- focus on one object, then focus out on the whole environment. There's zero lag.
Now try that on a camera -- even on a bright, clear day, the camera lens takes time adjusting from close to far. And in dark or inclement weather, forget it! You could spend several seconds trying to focus on a target.
absolutely spot on. no such thing as ai, there is only clever programming. just more non sense for the masses to believe and fear.
no quantum computer exists anywhere either, nor probably ever will. re-steemed.
As one person said recently "AI technology is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it."
I LOL'd so hard when I read this the first time.
Haha, that's classic! :)
This is interesting
I like the way you think...
Great points, AI isn't as advanced as deep learning implies. The technology is new, driverless cars have less accidents than humans per mile, have been driving for a 10th the time as humanity and will drastically improve.
Although AI can't yet adapt as quickly as we would want between different scenarios or paths, who says that a breakthrough isn't just around the corner in that field? A complicated AI running on a powerful computer could easily switch between different programs or AI settings and beat us at both basketball and football egames or other things.
And a camera lens autofocus? Compare the age of that technology to the billions of years of eveolution the human eye has undergone. I'd say after less than a couple generations the autofocus will soon catch up and exceed anything the human eye is remotely capable of soon. Such as having millions of focal points concurrently as magnitudes far beyond what our eyes are capable of.
But you're right, in its current form, AI isn't as deep as we are leading to believe.
An AI breakthrough will never be around the corner because no program can ever be smarter than the programmer. It can perform functions faster than humans, and it can "learn" within a predefined framework, but this is nowhere near human intelligence, which requires at least an element of introspection.
As far as cameras, they're bound by mechanical limitations. Even if we created a super-fast camera, the mechanical gears in the lens would experience friction, particularly focusing from far objects to close objects. But the human eye can accomplish this task at lightning speed, even from one extreme to the other.
Now, billions of years of evolution...how does nothing turn into something?
Why can't a program be smarter than a programmer? Students often outshine their teachers. I believe in the Singularity, where at some point in the next half century computes will surpass us in intelligence. Moore's law of increasingly better processors is holding true and breakthroughs are being made in quantum computing at least according to IBM, Microsoft and others.
As far as mechanical limitations are concerned, they will beat biological limitations. I agree the human body is remarkable, but robots are already stronger than us and more precise. Friction is overcome with lubricants. Also most human eyes fail and they are extremely fragile.
I am not going to argue about evolution.
Students outshining their teachers is an apples-to-oranges comparison (for instance, people are born with varying IQs). A program can't be smarter than a programmer because the program is entirely limited by what the programmer inputted. Consider the concept of generation loss, such as in audio replication.
Ultimately, why true AI can't exist points to the creation of the universe, which is why I suspect that you don't want to argue about evolution...you actually don't even want to say one thing about it. By creating our replacements with essentially a new, digital "species," we become our own gods, which is preposterous for the reason cited above. But even if such a thing were possible, it would further prove that the created (ie. humans) need a creator (ie. God), or what the establishment calls biogenesis.