Saturday, February 26, 2011


There's a private school near my house that has a really nice track, so sometimes we go over there to run/walk/play frisbee on their field. Today was one of those days. 
My dad's been into running barefoot lately (although none of us have tough enough feet to do much of it quite yet), so we were walking barefooted around the outer lane, which is the least worn-down.

And we talked about some interesting stuff, so I thought I'd share it with you.

Computing power apparently doubles every 18 months. So the super-computer I could buy with a million dollars in 18 months is twice as good as the one I could buy now.
Eventually, these computers are going to get so incredibly good that they can better themselves without our help. All we need to do is write a program that allows them to write their own software. 
One of the ways to do this is called genetic algorithms, which is basically digital natural selection. The computer makes a whole bunch of little changes, tests them out, picks the ones that worked best, makes changes to them, and so on.
The point at which computers become equal to or better than humans is called the Singularity. Ideally, these computers will apply their self-programming skills to our own DNA, and we will end up theoretically living forever. The biggest proponent of this is a man named Ray Kurzweil, who has done loads of research on the topic and is really, really serious about it. He even has arrangements made to freeze his body if he dies before Singularity gets here.
In late 2009 when I first read and posted about him (the post wasn't very good, so I'm not going to link to it), he was 61. He expects Singularity to occur around 2045.

Now, any of you who have been hanging around here for awhile know that I'm most interested in people. So you know that I'm going to talk about that.

Computers are good at doing what we tell them to do, but we have to tell them in a very logical way. How can we program them to make goals for themselves, to take the initiative, and to be innovative? How can we program them to have feelings? Do we want them to have feelings? 
Even the most objective among us are swayed by our biases and experiences. How could we ever predict what these computers would think or do, if we can't even conceive of such a high level of straight-forwardness? 
Do these computers count as people? Are there such things as souls, and if so, would they have one? At which point does the computer become conscious? Are we conscious, or is it just an illusion?

There's enough in those questions for several books to be written, much less a few quick comments, so I'll stop there.

If computing power continues to double, soon (as in, within a few decades) systems equivalent to IBM's Watson will be in our iPods. Wikipedia will no longer be needed.
If everything my dad's read and told me about is true, life as we know it will quite literally be enormously changed.

So do you guys buy any of it? What do you think will happen?


  1. A topic that is not only deep and interesting, but also near and dear to my heart (mind?), so my response grew too large for a comment. It is now a post of its own. But I also wanted to say that you seem to have a lot of great conversations with your Dad, I am somewhat envious ;) That didn't seem to fit into the post terribly well.

  2. He'll be glad to hear it. **off to read yours**

  3. This topic has always scared me. Considering the amount of damage that humans are doing to the world, a logical computer would, in all likelihood, kill us all. xP
    I'll leave you with that absolutely lovely tidbit of thought.

  4. Isnt that sad, the logical thing to do would be kill us all because we're screwing up the world so bad.

  5. There's actually either a book or a movie about that (I haven't read/seen it-- my dad just mentioned it.) The computer was honestly doing what was best for humanity. It just wanted us to stop suffering.

  6. I'm pretty sure that was the premise behind the movie, "I, Robot." Also, apparently I fail at subscribing to this comment thread, oops.


Talk to me.