Now that I have retired from a full-time commitment to engineering design work, I have the time to return to thinking about agricultural robotics. Much of the current field of agricultural robotics is focused on enhancing, with robots, those aspects of the farming process that have already been automated to some extent; for example, things like taking a tractor and replacing the operator with an AI system.
In other words, the commercial potential in agricultural robotics currently lies mostly with leveraging existing technology with the addition of robotic components.
But the real challenge in agricultural robotics, and the aspect of this field that intrigues me the most, is the challenge of replacing the individual field worker responsible for picking our food crops. This challenge is not unique to the farm industry. Any industrial working environment which takes place out-of-doors, like logging, fishing, mining, or construction will present the same challenge, that is, there is no fixed working situation that is repeatable or predictable over time. Whatever tasks robots engaged in such work will be required to do, they will be required to constantly modified their programming, on a real-time basis, and adjust their activities accordingly.
So the key challenge replacing the individual field worker is one of creating a system that can learn and modify its programming while at the same time running the tasks it’s been expected to do. With this in mind, in the coming months I want to tackle issues related to machine intelligence and learning systems.
This goes back to a project I started several years ago but had to put on hold; that was the very large asynchronous array of simple processors.
Since I don’t have a job anymore with an income stream, building hardware cost too much money to keep going in that direction. So I’m renewing the project but in simulation. So my goal in the coming weeks is to start putting together a simulation program to create a very large array of simple processors.
Such an array becomes an ecosystem, so to speak for cellular automata to exist; much like John Conway’s Game of Life. But the twist I want to explore is the possibility that the kinds of stable structures that form in such an environment might actually be able to act as computing agents. I am not sure if anyone has explored this option. So one of my to-do’s is to get up to the library at UC Santa Cruz, start reviewing literature, and see if anyone are worked on this possibility.
This is the project for the hardware engineer in me. But for the physicist in me, I want to pursue the deeper question of whether not such a computational structure could then qualify, in some abstract sense, as a life form. The starting down that abstract mathematical path, I will be spending time looking at Landauer’s principle versus Maxwell’s demon as a life form.