When I was a lad, I was easily excited by technology, and one of my biggest enthusiasms, back in the 1970s, was Content Addressable Memory (CAM).

Last week, I met another youthful enthusiast, who had discovered that there’s a research project in train to make self-healing computer circuits, using chemical wave refraction phenomena. The idea seems to be that Beluosov-Zhabotinsky (BZ) reagents can be used to solve quite complex problems – the “travelling salesman” problem being one, paths through mazes being another. And also, you can use BZ devices to emulate logic gates.

The thing about these two apparently unconnected lines of research is that one researcher believes it’s theoretically possible to construct computers of comparable complexity to standard Von Neumann engines using chemical waves. And, he suggests, they’d be much more robust than silicon.

My impression is that the researchers are not expecting to have a commercial product any time in the next decade, and have no idea whether it would be a general-purpose device like a digital computer, or a specialist function box like an analogue computer. But my enthusiastic friend felt that it was the dawn of a new era because, as he said, “Moore’s Law must be about to run out of steam, and making silicon is incredibly costly, and complex. And if we want to develop true artificial life-forms, it has to be using a base technology which could occur in nature and evolve.”

Well, perhaps. But if this technology hopes to get off the ground, it has to compete with other computer products at some point. A radically new technology doesn’t stand a chance unless it goes a lot further than just “can compete”. It has to be about 10 times better, or more. And that, of course, is why we don’t have CAM.

Standard RAM (as ICL’s former head of R&D observed when ICL announced the Distributed Array Processor) coupled to a fast cache to a fast processor is how all our software tools work. And, it turned out, he was right: the DAP is history.

And yet, aren’t we getting to the point where a distributed processor architecture, including processing power on the RAM chip, might reduce the bottlenecks?