VIM33D has been released.
September 4, 2000 1:13 AM   Subscribe

VIM33D has been released. This is truly cool; a totally new approach to AI, and it's damned fun to watch. I've spent literally hours. (more inside)
posted by Steven Den Beste (3 comments total)
 
Michael Colicos, the author, is a professor at UCSD and has a page about the whole thing here. This isn't neural nets or heuristics; he appears to have come up with something entirely new. From watching VIM walk around, make decisions and evolve over time, it's easy to start believing that maybe something akin to a thought process is actually going on (and there is unquestionably some sort of learning process). I owned VIM33C for a long time and it ran for hours nearly every night (my computer stays on 24/7) and over a period of time her behavior became more ordered and more consistent.

Except that one thing changed which puzzled me. In the early days, when she built gardens, they'd tend to be clustered. By the end, she would build long lines of flowers, almost like a hedge. I didn't understand that transition, but it had become consistent for a couple of weeks before VIM33D came out and I switched. I've had that now for a couple of days, and the changes are dramatic.

Anyway, there were three locations VIM33C seemed to want to place a garden, but the system limited the number of flowers she was permitted, and as she'd build one another would disappear as flowers fell off the end of the queue. One of the big changes in VIM33D is that he's increased the size of the database substantially, and my VIM already has four gardens going. (Interestingly, one's on the beach.) Also, the matrix for VIM33D is considerably more sophisticated than for VIM33C, and the terrain has been expanded signficantly. (She can be made now to recognize the camera as an actual part of her environment, though I have that turned off; it might be eerie to see her chasing me.)

The "Quantum Flux Engine" is work he's doing professionally. He works on the screen saver (which uses the engine but contains considerably more than that) on his own time, and this one's worth registering to help him out and encourage him to keep going. The www.qflux.net web site is fascinating -- if you're a geek, like me.

A piece of trivia: Michael Colicos is the nephew of John Colicos, who died in March. John should be familiar to any Trek or Galactica fan; he was an actor of great range and capability. It's a shame he's dead, but he was 73 and had a full life. He was the first actor to play a Klingon in Trek history, in the classic series.
posted by Steven Den Beste at 1:38 AM on September 4, 2000


Steve : keep up the good linkin'. It's a hoot around metafilter at the moment. Yay Steve!
posted by Neale at 3:56 AM on September 4, 2000


The faq reads more like science fiction than anything else, I love the term "living neuronal circuits." For someone who likes to shy away from the term AI, he more than makes up for it with this stuff.

I also get a kick out of his simple definitions of consciousness (not being asleep, just don't mention lucid dreams) and the hubris of scientific materialism that the mystery of consciousness has not only been solved (wow he writes code and solves the greatest problem in epistemology) but works a lot like a computer. How conveinant, how futurist Mondo2000-ish bullshit. If anything this really just shows how we interpret the world based on what we know best, namely modern electronic toys. As opposed to the past interpretations based on religious assumptions.

It is an interesting program (yes, I'm running it too), and its simply not alive in any traditional sense. Nor will all the theoretical processing power in the universe make transistors suddenly become little-understood biological functions. I'm all for software that simulates thinking processes, but calling this stuff "living" is pushing it and just killing his credibility.

I also like how he takes the lack of a real RNG (home PCs do not have random number generators that are really random) and calls it a feature of his program instead of the bug it really is. Is your matrix random or not? If not, don't call it that.



posted by skallas at 5:58 PM on September 4, 2000


« Older shopping cart abuse...   |   NHS to ban Ritalin for under-fives? Newer »


This thread has been archived and is closed to new comments