BARBELITH underground
 

Subcultural engagement for the 21st Century...
Barbelith is a new kind of community (find out more)...
You can login or register.


Underground > Laboratory >

The autonomic network?

  
 
The Return Of Rothkoid
10:49 / 19.10.01
BBC writes about IBM's plan for a self-monitoring/configuring network based on a biological model. quote:Bringing computers to life

Soon the net could be healing itself

IBM has unveiled an ambitious initiative to develop technologies that share the basic biological abilities of living organisms.

Senior researchers at the company said the growing complexity of computers and networks demands that the technology does a better job of maintaining and healing itself.

The researchers warn that without these efforts there is a danger that networks will soon become unmanageable.

The company is backing its initiative with its own research program, a series of grants to universities and efforts to make other computer companies sign up.

Grand challenge

This week IBM is sending 75,000 copies of a manifesto written by Paul Horn, senior vice president of IBM Research, that details the aims of its Autonomic Computing initiative.

Mr Horn warns that humans are losing the battle to manage the increasing complexity of computer systems and networks.

This complexity is only going to increase as computer technology shrinks and finds its way into ever more devices.

If the current rates of the expansion of digital technology are maintained, soon there would not be enough people to keep the world's computer systems running, he said.

He called finding ways of handling this complexity the next "grand challenge" facing the technology industry.

In the manifesto he said: "The growing complexity of the [technology] infrastructure threatens to undermine the very benefits information technology aims to provide."

Ideally future networks should resemble the autonomic nervous system which maintains and monitors many basic bodily functions without conscious help.

Global power

The autonomic nervous system maintains blood sugar and oxygen levels and monitors temperature. It adjusts the body's heating and cooling systems to keep body temperature hovering around 37C.

What is needed, argued Mr Horn, are computer systems that do a much better job of configuring themselves, can work around disruptions, heal any damage they suffer or fight off potential problems.

IBM is planning its own research programs to create technologies that can turn relatively dumb networks into smarter alternatives.

It is also planning to spend millions over the next five years funding 50 research projects at universities to take on the complex challenge.

The likely outcome of the project is a series of software standards that define how to build software or hardware that has these more biological properties.

IBM is working closely with the Global Grid Forum. This industry body is driving efforts to turn the disparate computing and research capabilities of the world's science labs into a shared pool of resources that anyone can plug into.

This effort is already driving the creation of software that hides the individual quirks of individual machines and instruments behind common interfaces. IBM research page.
 
 
grant
15:03 / 24.10.01
There's a department or lab at MIT - besides the robotics lab - which is working on individual chips & circuits based on neurological/biological models.

And in Israel, AI researchers have created a bit of software which is set up to learn from people like a child does -- reading and arithmatic lessons. It's at about the 5yr old level right now.

Can't put my hands on sources for either right now, but it's interesting that machine dudes are looking to bio hardware for inspiration.
 
 
Magic Mutley
15:48 / 24.10.01
There's a history of this - in the 80s an AI researcher called Igor Alexander was doing experiments in simulating neurons using memory chips - he wired together a network of them, and made a machine that could recognise faces. The bit that interested me was that if you turned off the camera, the network would tend to oscillate, throwing up it's own images - kind of like 'dreaming'

I've also heard suggestions of applying models from nature to telephone networks - using electonic 'insects' to find the fastest way across the network, in the way that ants make paths to a food source.
 
 
belbin
15:01 / 26.10.01
WG> "Swarm Intelligence" - http://www.hbsp.harvard.edu/hbsp/prod_detail.asp?R0105G (don't think I still have a copy of the article sadly).

The key idea here is rather than attempt to solve a problem by producing a dirty great equation or set of complex operations, you create a large number of autonomous agents equipped with a few simple rules and let them come up with the answer for you. It's about being small and numerous rather than big and clever.

I like the concept that swarms of tiny agents can be collectively cleverer than one big agent. So therefore AI should be getting down and dirty with insects, bacteria, plants and viruses rather than attempting to model something like the human brain straight off. We tend to think of familiar models of 'intelligent' (i.e. problem-solving) behavior. Where as the best ones might originate somewhere completely alien to us.

Back to IBM: Presumably you'd have to build in more feedback loops and homeostatic controls into your system. Odd becuase biology is becoming more like computing (modelling population behaviors, modelling genetic material as information) and computing is becoming more like biology.

My guess (and it's a bit obvious I concede) is that biotech (I don't just mean genetic engineering here) will be the 21st century equivalent of what the transport, telecomms and software sectors were to the 20th century.

Further thoughts: now what does the concept of a self-regulating computer network imply about vulnerability to viruses and the kinds of reproductive code we can see spilling about?
 
  
Add Your Reply