The machine that invents

St. Louis Post-Dispatch ^ | 01/25/2004 | By Tina Hesman

[Momaw Nadon]

Technically, Stephen Thaler has written more music than any composer in the world. He also invented the Oral-B CrossAction toothbrush and devices that search the Internet for messages from terrorists. He has discovered substances harder than diamonds, coined 1.5 million new English words, and trained robotic cockroaches. Technically.

[FormerlyAnotherLurker]

Mine figured out the combination to the gun safe and was examining the Barrett M82 when I took an ax to it.
The ammo was in a different safe or else... [shudder]

[Old Professer]

Cross-action toothbrush; functional equivalent: "Wax on, wax off."

[Loc123]

Take a bunch of words on paper, put them in a jar, and shake. Interpret the random results. That is the creativity machine for the most part. The only difference is the scientist adds some algorithms to the “secret layer” to rule out useless combinations based on set rules. These useless combinations are only applicable to certain systems (IE the citation of the car—obviously the car needs certain constants in place).

They key is, an analyst needs to analyze the results of the tempered combinations. That takes creativity.

Also, it is doubtful as of today that this method of combination can produce mechanical models of the universe since often times the forces are imaginary/contrived. The machine only deals with knowns and their combinations.

Interestingly, I had earlier predicted that machines could simulate reality given the knowns, but that they couldn’t interpret it. Even so, I thought perhaps they could “interpret” with us inputting desirable states of the agents/objects in the system. The computer could then use the criteria the scientist inputs and that would execute that combination of agents/items. This is similar to human thought except that we can unilaterally decide what is desirable, whereas a machine follows our orders.

Resonse to the journalists piece: complete biased. She presents two angles on the creativity machine. 1) It has creativity and will be magnificant. 2) It has creativity and will be detrimental. Post-humanists aside, there seems to be a more credible, more accepted viewpoint: it doesnt have creativity and is an upgraded randomness generator, and this has been around for years (New Scientist article I cited before).

Don't get me wrong; this is a great analytical tool. But the truth about the machine has been covered up by an agenda. The author even admits that this is his religion (see his website).

[Old Professer]

Your link sends me to an exe file, do you suppose you could explain?

[Southack]

[MHGinTN]

Silicomazing!

[Libertarianize the GOP]

ping

[RightWingAtheist]

Here's a link to an article I posted a while ago, about a robot which could "do" science: http://www.freerepublic.com/focus/f-news/1058579/posts

[Ernest_at_the_Beach]

Others say it is something far more sinister - the beginning of "Terminator" technology, in which self-aware machines could take over the world.

I'll send a note to Arnoold to see if he feels threatened!

[PatrickHenry]

At last, the explanation for a certain freeper, now banned.

[SC Swamp Fox]

bump for later

[Junior]

Bump.

[E. Pluribus Unum]

Your link sends me to an exe file, do you suppose you could explain?

As I said, it is a program I wrote about 12 years ago. A DOS executable.

It opens with two screens showing the mathematics of the feedforward/backpropagation neural network algorithm.

The first screen is a mathematical model of a neuron. The second screen shows a three-layer neural network and the mathematics involved in feeding back the pattern error into the weights during the training cycles.

The next screen shows a virtual scanner as letters are scanned across it, then it begins to train. You can watch the weights change and the error diminish as the network gets better and better at recognizing characters.

[E. Pluribus Unum]

This was part of a package I wrote and advertised in PC Magazine for $6000. I sold over 300 packages at $29.95 + $3.00 shipping. Nobody asked for their money back.

[theFIRMbss]

A guy also wrote code that creates some very impressive pictures.

[r9etb]

Interesting article, though I think it's probably a bit heavy on how well it all works -- the lack of "easy off the shelf useability" says they're still a long way off.

"None of his computers can do ballet."

True -- and none of the computers have human morality, either. If we extrapolate the capabilities of this out to a set of really capable "thinking machines," there does not appear to be anything that will stop a computer from dealing with humans as humans deal with, say, cows or horses.

More believably, I think there's probably a great deal of utility in these things as "optimization engines," but note that Thaler's "critic networks" still require some sort of rules for what "best" means -- which means that humans aren't off the hook for understanding the problems they're trying to solve.

[r9etb]

but has none of the weaknesses that we have

Do you mean "weaknesses" or "morals?"

[avg_freeper]

I agree. Evolutionary systems that pit slightly mutated or pertubated variants of neural networks against each other have been around for some time. This seems to be very similar with some smoke and mirrors thrown in as decoration.

[null and void]

At last, the explanation for a certain freeper, now banned.

Who? That doesn't narrow it down much...

;^)

[N. Theknow]

Stephen Thaler invented Al Gore.