Tuesday, July 28, 2009

Peter Cochrane on evolving networks and the future of machines

Peter subtitles this - when the machines cme to our rescue. He certainly provides us with an interesting view of the future relationship between man and machine.

Part 2

1 comment:

Tom Foale said...

A very thought-provoking talk. I saw Peter at Martlesham talking about ants and complexity, and it has figured in my thinking ever since.

I think swarm intelligence may be the first autonomous 'thinking' system to appear - simple programs communicating and taking actions based on requests from other machines, leading to complex interactions that no human can follow. The benefits for mankind are incalculable.

Imagine this scenario - a computer used for planning plots demand for screws and sees a shortfall, so it issues a purchase order. A computer in a screw manufacturer receives the order and checks it against capacity. The company does not have enough, but the computer has a 'unit of production' - an automated, 'dark' factory - that it can commission, so it responds to the bid and wins. The factory is pre-built, can't even be entered by humans, and the specialised production machines are entirely maintained by specialised robots. Raw steel, power and lubricants go in, screws are produced and wastes go for recycling. So mankind doesn't even need to be aware that the factory exists.

The real fun is in the economics. The manufacturing cost of a screw in this scenario is a share of buying the land and financing the capital cost of the factory, plus the cost of input materials, maintenance, shipping and waste reprocessing - no human needs paying. Now extend that across the value chain - raw products like iron ore are extracted, processed or manufactured by machines under the same sort of instruction, factory components are built the same way, goods are moved by automated vehicles, waste reprocessing is automated. The only cost in the screw becomes whatever landowners charge to mine the raw resources or for the land to place the automated factories, plus energy costs. Oh, and the cost of development of the next generation of more efficient machines, built into the cost of buying the plant.

So the price of simple components manufactured from simple resources, like screws, drops. Robots are simple but expensive things, well suited to repetitive tasks and long production runs, so initially the capital cost of these plants will be high and they will produce simple, high-volume products. Over time, as more and more of the components of the robots are produced by robot factories, the cost of robots will decline as well. We will have a world full of goods at prices that no human enterprise could match, and all produced with no need to invoke intelligent complex robots or AI - just supply and demand driving robot production facilities.

A factory can be mothballed when demand drops and re-activated as it increases, with no lay-offs or redundancies. Jobs will be scarce, but with goods being cheap we will be able to job-share in work that does require intelligence or, in particular, purposeful creativity. Some goods will still be expensive - there is not enough gold in the world for everyone to have a solid gold house, as an extreme example - and scarcity will still make resources such as land expensive. But on the whole, we will be a lot better off and we don't need to understand the complexity of the whole system at all - it just follows the rules of economics. This is just one of the benefits of the semantic web.