It looks like you're using an Ad Blocker.

Please white-list or disable in your ad-blocking tool.

Thank you.


Some features of ATS will be disabled while you continue to use an ad-blocker.


The Inventing Machine

page: 1

log in


posted on Apr, 21 2006 @ 09:00 AM
This is a machine that invents devices.

Like a friendly, non-biological form of the Borg Collective of science fiction fame, 80 personal computers, using artificial intelligence (AI), have combined their silicon brains to quickly design a tiny, advanced space antenna...

To design the ST5 space antenna, the computers started with random antenna designs, and through the evolutionary process, refined them. The computer system took about 10 hours to complete the initial antenna design process.

So, a machine that invents other machines to fit a specified need. This is pretty freaky, and possibly dangerous. Any thoughts?

posted on Apr, 21 2006 @ 10:59 AM
sweet find! really interesting that 'they' have released this information. things will be a lot more interesting though when the a.i. decides that it needs to be mobile, or able to defend itself/children.... (dun dun dummmm)

posted on Apr, 21 2006 @ 01:00 PM
Or when the AI decides that it would function more efficiently running with a human brain, and invents the interfaces for it.


posted on Apr, 21 2006 @ 01:10 PM
I use a similiar piece of software, though all it does is fix up pictures using genetic algorithims.

This stuff doesn't work on its own, it needs the input of a human controller. We are far from the day when these things are self-reliant.

posted on Apr, 21 2006 @ 01:31 PM
Yes, human input is still needed. Personally, even with all the oodles of breathtaking inventions, I'm holding out for the food replicator. I don't want a microwave oven or even something like a crockpot or bread machine in which one throws in the ingredients and the machine does the work. I want a food replicator, then I'll be impressed. '

posted on Apr, 21 2006 @ 04:18 PM
Yes, this kind of program has actually been around for awhile.

The computer program essentially sets up a mock test-tube evolution of programs. It assigns many random patterns of program reconfigurations, or test parameters, and then tests each one in a hypothetical setting. Whichever outcome produces the best numeric results is selected as the starting point of the next cycle of tests.

This does take a long time though, and so I'm surprised this only took 10 hours, but it is not quite as Artificial Intelligence as you may believe, although it's definitely got the groundwork there.

I mean seriously, how do you invent and think up things? Your brain decompiles some data it knows and tries to reassemble it again in a different way. Sometimes it doesn't make a lot of sense, and so you toss the idea (like the idea of punching someone in the face... you test the idea, it looks good, but further testing of "down the line" consequences - ie, jail - stop you from doing so). Other ideas are good, and so you keep them, refine them, and re-recompile them to find even better ideas.

The difference, our parameters are extremely large and varied. A computer may look at a sword and design the best sword you could possibly have. However, there's a certain "intimidation factor" that the computer doesn't understand, and doesn't have set parameters or functions, and so while it's sword may be "better" - the human sword may win due to elements of the design the computer couldn't handle.

I just fear the day when computer-games are given this advanced of an AI. I'll never win a game again! Call it "super-impossible mode".

posted on Apr, 21 2006 @ 04:33 PM
I've seen something like this for sodaplay, it lets you create sodaracers by choosing an evolution type and then waiting until you satisfied with the results.

posted on Apr, 21 2006 @ 04:51 PM
That's the key passage:
"The AI software examined millions of potential antenna designs before settling on a final one," said Lohn. The software did this much faster than any human being could do so under the same circumstances, according to Lohn. "Through a process patterned after Darwin's 'survival of the fittest,' the strongest designs survive and the less capable do not."

"We told the computer program what performance the antenna should have, and the computer simulated evolution, keeping the best antenna designs that approached what we asked for. Eventually, it zeroed in on something that met the desired specifications for the mission," Lohn said.

That is really an old known concept. FAR from any AI. There is no intelligence behind this just random selecting of random attempts:
You create a number of random antenas and select s(evolution) ome that best fit your needs alter them a little (eveolution) and test again. Youc an do it a little more clever than that bbut basicly its this and nothing more.

If you can't calculate something directly because of its complexitiy this is an approach you can do. But it is rather a slow one and doesnt always garant you the best outcome.

posted on Apr, 21 2006 @ 04:53 PM
It didn't say this in the article (that I noticed, anyway) but this method sounds a lot like either finite element analysis or numerical methods, where basically you input a reasonable design and then keep mathematically iterating it over and over to slowly improve it. There's quite a few programs that do similar things, but I wouldn't classify them as AI. I wonder what is different about this program over older ones that this is considered AI? I guess it depends how AI is actually defined...

posted on Apr, 21 2006 @ 05:00 PM
I don't think it's necessarily AI, but if a true AI were developed and combined with this technology, it would be something. It makes sense to me (and maybe no one else) that that would be the next logical step in the process. If this device can work this efficiently, then one with a functioning AI would work that much better, and be that much more productive for NASA, or whoever else.

posted on Apr, 21 2006 @ 05:24 PM

Originally posted by Rasobasi420
I don't think it's necessarily AI, but if a true AI were developed and combined with this technology, .

of course a true AI can make use of this methode (evolution) or most probabily will use this or something simulare. But first you have to develope that AI, and then it's a question what it is based on. Neuronal-Net's? Quantum-bits? The methode then for the AI to learn might be very different based on it's technologie.

We are no where with AI till today. All we can do today is programming something the way that it looks like it has inteligence for a simple human where in fact all is preprogrammed in combination with a random generation and some logic to search for the best matching answers. Ok we can do a little more than this but to speak about AI today is still somehow funny for me.
My personel view..might be others see this already a little different.

[edit on 21-4-2006 by g210]

posted on Apr, 21 2006 @ 08:01 PM
That is quite interesting.
Im sure were a long way off of 'sentient' AI. These AI's seem to have alot of pre-programming, making them more of an adaptive program than an AI.

But the possibilities of these programs are very enticing. I'd love to see what they can do for aerospace!

new topics

top topics


log in