Wednesday, September 29, 2004

Moore's Law

I'm about a quarter of the way through an article by Steve T. Jurvetson on Ray Kurzweil's site and thought I'd jot down a couple of things before I forget them.

This article discusses Moore's Law and how abstractions of it can be shown to have been going on for 100 years (there's a nifty little graphic that I'd copy here, but my Flickr account is full for the moment due to a test I'm running).

Jurvetson talks about how NASA AMES shut down their wind tunnels this year, opting for computer modeling instead, this contributing to the ability to rapidly model things without ever having to create physical models. This, along with the general discussion of Moore's Law that Jurvetson is undertaking got me to thinking about abstracting Moore's Law even further beyond computational power into the types of problems that computing has been able to tackle as the power increases. Here's a quick, off-the-top-of-my-head list.

- Simple Calculations (+,-,*,/)
- Accounting-Type Math
- Basic Scientific Math
- Macro-Level Modeling
- Complex Modeling (e.g. Weather)

This got me to thinking about what the next order of problems that computers will be able to undertake. Again, this is after 30 seconds of thinking only. Discuss.

- Complex Multi-System Interactions
- Multi-System Optimization (based on human defined parameters)
- Multi-System Optimizaiton (computer derived parameters)

I'd also like to say that, at this point, I'm not convinced that exponential progress will ultimately be sustainable. This "singularity" that has gained popularity just doesn't sit right with me. I think that the technology behind Moore's Law may keep the problem domains that can be solved via computation moving along nicely, but I feel that as things start really feeling out of hand that we'll start reigning things back. I think it's as naive to think that it will reach a point we can't stop it as it is to think that things won't continue progressing (sounds like a contradiction, I know).

Anyway, more later.