Technocrat

Making Robots More Human

zogger
Tue Jul 15 19:42:00 -0700 2008
Science.
manage

According to a researcher in Washington, the secret to making more human like robots is to first understand how that human part works. For instance, in trying to make a robotic hand, they had first made the finger bone joints smooth. It turns out they worked better when they really copied what a finger bone looks like, with a bumpy and rough surface.

One of the first things Matsuoka and her team discovered upon taking this more anatomically governed approach is that the rough surface of the bones in the fingers was an essential, functional feature rather than an unimportant side effect of bone development.

ed.z.: Ha! The modern business idea is the reverse of that, try to make the humans be more robotic! My guess is, a few thousand years from now, given we don't nuke or biocootie or pollute ourselves to extinction, it will be more-borg like than not, a full blend, androids.

Making Robots More Human
harry666t
Wed Jul 16 07:08:42 -0700 2008
 manage

Hm, interesting. I think that this also means that if we really would like to create a "real" AI, we should study (and understand) the nature of human mind a little more.

Making Robots More Human
President4242
Wed Jul 16 07:54:16 -0700 2008
 manage

Isn't that what the neural net based question-and-answer learning methodology is all about?  Been a while since I messed with artificial intelligence, so I'm not up on all the latest names, but even back in the early 1990s weighted neural nets that asked questions were a huge research direction.

Making Robots More Human
David Leppik
Wed Jul 16 10:50:02 -0700 2008
 manage

Neural nets have always been, at best, a crude approximation of brains. The algorithms used in computer science involve "input" and "output" neurons with feedback loops, often with one intermediate layer. Other than that, they are basically unstructured; the learning phase produces all the structure it's ever going to get. The brain, meanwhile, is highly structured and somewhat modular.

When I was studying computer science in the mid 1990s, I took a few courses in cognitive psychology and was surprised to find that they were using a distributed network model to describe the mind, similar to neural networks but explicitly not neuron-based. (Particular facts are modeled as emergent from a network, rather than found in a particular network node.) Even there, you ended up with structures which limited or shaped feedback loops.

And that's without considering the fact that brain researchers today are finding that stuff other than neurons are important to how the brain works.

Making Robots More Human
harry666t
Wed Jul 16 13:32:07 -0700 2008
 manage

And that's without considering the fact that brain researchers today are finding that stuff other than neurons are important to how the brain works.

For example? Does any of this new stuff explain some of the paranormal abilities of human minds? (telepathy, telekinesis, precognition, etc) Will the AI researchers also need to somehow replicate these properties to create an artificial consciousness? I mean, I think that all the religions of the east are talking about the "third eye" for a reason...

Making Robots More Human
David Leppik
Thu Jul 17 07:40:53 -0700 2008
 manage

No, sorry. It's just mundane stuff such as fat cells which affect the timing of neural firing (which turns out to be extremely important for hearing), plus the effects of hormones and neurochemicals. The neurons are still the main actors, but there's a large and extremely important supporting cast.

Making Robots More Human
Petr Mareš
Wed Jul 16 12:33:40 -0700 2008
 manage

I think that if we really would like to create artificial intelligence, then we should make clear what we think with the term "intelligence" or "artificial intelligence".

There are some definitions, but even the best defs depend on notion of success, which is something that must be set by another intelligent being. So it really boils down to AI being something we deem intelligent.

To be precise, the engineering task of creating an AI didn't even start, for we haven't set an objective. We could not start an analysis.

All the humanity has is optimizing expert systems. People suck in thinking so much (in general) that they don't mind when they don't know what are they doing.

One of the greatest jokes in biology is naming the human "homo sapiens sapiens".

If some autonomous intelligence emerges somewhere, the humanity is doomed, for in real we are just a bunch of bigheaded greedy fools.

 

Do you have questions about the site? Write to bruce at perens.com . List of all articles we've ever published. Source Code Developer's Documentation RSS XML Sitemap XML News Sitemap Valid XHTML 1.0 Strict Valid CSS! [Valid RSS]