Slime Mould intelligence points to a new model of AI

slime-mould.jpg

Earlier this year, a group of Japanese scientists reported that with appropriate training, the true slime mold Physarum polycephalum can anticipate the timing of periodic events.

That’s more than some politicians can manage and P polycephalum is only a single-celled amoeba, albeit a talented one. A few years ago a Hungarian team showed that slime mold was able to find the shortest way through a maze.

Clearly, primitive intelligence has cellular origins but how might this work?

Yuriy Pershin at UC San Diego and pals think they know how. They say that this kind of behaviour is identical to the way a simple electronic circuit reacts to train of voltage pulses. The circuit consists of an inductor, capacitor and a memory-resistor, or memristor.

Interestingly, this learning behavior comes from purely passive components. This can easily be reproduced in the lab and the San Diego team say it may turn out to be a useful way to build passive circuits that learn.

Link several of these passive learning circuits together and you might be able to knock up a simple neural net. Suddenly, you’ve got a new kind of AI on your hands and the origins of cellular intelligence don’t seem so obscure, after all.

Ref: arxiv.org/abs/0810.4179: Memristive Model of Amoeba’s Learning

4 Responses to “Slime Mould intelligence points to a new model of AI”

  1. says:

    [...] http://arxivblog.com/?p=684 By Ben, 10/26/2008, 12:46 am o’clock [...]

  2. Zephir says:

    Well, by AWT such intelligence is quite common even for non-living matter due the memory effect of nonlinear behavior of Aether foam. The dense chaotic field exhibits the foamy density fluctuations, the density of which increase with energy density, by the same way, like the soap foam which gets dense under shaking – it has an rudimentary memory.

    After then, every isolated wave (a soliton), which is spreading through such foam is behaving like less or more dense blob, i.e. wave packet or particle. And this particle focuses the neighboring energy wave into itself like food, being attracted by density gradients of Aether foam (a gravity field) like bacteria by sugar (chemical energy) concentration.

    Such system furthermore intensifies its energy exchange by formation of nested foam of density fluctuations by the same way, like the slime mold if forming a branched colonies. And the energy undulates along surface of the resulting quantum foam by the same way, like protoplasm is changing direction inside of slime mold colony.

    By such way, the slime mold intelligence is exaggerated case of quantum foam intelligence – in fact it’s just another phase of it!

  3. [...] The mathematical modeling of memristors have started its bleedthru into other fields: arXiv.org is reporting in the Quantitative Biology & Cell Behavior a new paper titled Memristive model of amoeba’s learning, by authors Yuriy V. Pershin, Steven La Fontaine, and Massimiliano Di Ventra. Neural nets have been around a long time, even though seemingly of primary interest to stock market marketing schemes :) … Earlier this year, a group of Japanese scientists reported that with appropriate training, the true slime mold Physarum polycephalum can anticipate the timing of periodic events. That’s more than some politicians can manage and P polycephalum is only a single-celled amoeba, albeit a talented one. A few years ago a Hungarian team showed that slime mold was able to find the shortest way through a maze. Clearly, primitive intelligence has cellular origins but how might this work? Yuriy Pershin at UC San Diego and pals think they know how. They say that this kind of behaviour is identical to the way a simple electronic circuit reacts to train of voltage pulses. The circuit consists of an inductor, capacitor and a memory-resistor, or memristor.[article] [...]