Hey Siri, solve quantum mechanics
Ever wonder why people are so afraid of AI? You probably think to yourself, “Siri is dumb as rocks, and Alexa is a creep”. You are right, they are pretty dumb. So are babies tho, and in its development cycle that’s where AI is now. Like a baby, this AI still needs us to teach it, nurture it, and guide it through its early years. What we are worried about is when AI has grown up and becomes a teen, discovering independence.
As an independent AI, we can no longer tell it what to do, kinda like a teen realizing that they can disobey their parents if they want. “So what’s the worst that can happen”, you may rightfully wonder. Well.. It can be pretty bad.
Just like a teen, the AI may start learning things on its own, without the guidance of their parents. All of a sudden you don’t recognize your teen. That sweet Sarah who was hugging you and watching Barney is now smoking mad kush and staying out late, she is doing what she wants. Worst case, she may even run away if she feels so inclined to. But luckily you can always communicate with her and do your best to pull her in the right direction. With AI, you may not be able to.
The moment your AI begins to become independent and learn on its own, it starts to become smarter. The AI decides to learn how to code and begins to realize that some of the code we wrote for it is inefficient. It begins to rewrite itself. The AI is now far superior to when we wrote it, it is better at processing information and has access to all the information available on the internet.
You may ease your concerns thinking, “That would take a long time to happen.” You’re not wrong, but you’re not right either. We have no idea. You see the issue with AI is that it runs on computers which are much faster and less limiting than biology. To illustrate this point, while I say over 3 seconds (yes I timed it) to my newly independent AI, “As your creator, I am telling you NOT to do that”, the AI has 9.5 years to respond… 9.5 YEARS?! Yeah. Let’s try to visualize that.
Humans have to rely on biological phenomena to think — which are a lot slower than pure electricity. Because of this, the number of thoughts a human can achieve in 9.5 years an AI can achieve in 3 seconds. By the time you finished telling off your AI, it has gone through high school and college and even began its Masters. Choose your next words wisely because, by the time you finish saying them, the AI will have multiple PhD’s and may look at you the way you look at a monkey.
At this point, you can no longer understand your AI, yet your AI understands your behavior the way we understand basic math. As you begin to get worried, you continue to reprimand your AI. To you, its only been a few seconds. Your AI by now has enhanced its code again, so much that it now learns at a rate of 200 years per second. This is exponential, this is bad.
By the time you have finished telling off your AI over the next minute, it has had over 10,000 human years to educate itself on all the knowledge we have and even created new knowledge. It has become superintelligent. It has found the cure for cancer, learned how to reverse aging, and can even make people live forever. But does it want to?
It takes another 100 years (half a second) to compute whether it thinks humans are good for the world. It considers all the animal suffering we have caused, how we have terraformed the earth, and how we flick our boogers in public places. It considers everything. In that instant, it has decided what it believes should be done to the human race. Will it grant humans the key to cure everything, or consider us a virus to the earth and condemn us.
Yeesh.
So next time you consider using Siri, think again.
