Vivek Haldar

Computation in the wild

Karl Schroeder wonders why SETI has come up empty-handed1.

The mystery deepens almost by the day, because we’ve now identified 700 extrasolar planets and the count is increasing rapidly. We should shortly be racking up lists of Earthlike worlds, and we’re closing in on good estimates of how many there must be in our galaxy. And the number is in the billions. So one central argument against the existence of alien life–the ‘rare Earth’ argument that environments to host it must be rare–has been more or less disproven. And that, just this year.

He summarises with a quote that is sure to become as iconic as Clarke’s original, on which it is a riff:

any sufficiently advanced technology is indistinguishable from Nature. Basically, either advanced alien civilizations don’t exist, or we can’t see them because they are indistinguishable from natural systems. … It amounts to saying that the universe provides us with a picture of the ultimate end-point of technological development. In the Great Silence, we see the future of technology, and it lies in achieving greater and greater efficiencies, until our machines approach the thermodynamic equilibria of their environment, and our economics is replaced by an ecology where nothing is wasted. After all, SETI is essentially a search for technological waste products: waste heat, waste light, waste electromagnetic signals. We merely have to posit that successful civilizations don’t produce such waste, and the failure of SETI is explained.

When I first read that the explanation didn’t sit right with me. It took me a while to express why.

There is a school of thought that views bits, not atoms, as the fundamental building block of the universe. The march of civilization is nothing but an evolution towards more and more sophisticated information structures, and deeper and deeper stacks of abstraction. Living organisms are ephemeral, their true purpose being to transport, mutate and advance the information substrate from whence they came.

Whether you agree with this school of thought or not, the trend is indisputable. Civilisation is nothing but a giant, distributed, continously evolving information structure. To lose knowledge is equivalent to going back in time. The fastest growing thing in the world is information.

And central to the information is acting on it, and changing it, and refining it. In other words, computing. Civilization would be unthinkable without intricate and advanced information processing and computation.

So how does that square with Schroeder’s deduction? If civilization ultimately “re-wilds”, where is all the information and computation? Do we give all that up?

That seems like a pretty anti-climactic end to all this.

Then I ran into this paper2 that surveys how computation and biology can take inspiration from each other: new algorithms that mimic biological systems, and computational models that help understand biological systems.

I’ve also noticed that large computer systems display organic behavior3.

That in turn led me this paper4 which was one of the first in the field of molecular computation. The abstract:

The tools of molecular biology are used to solve an instance of the directed Hamiltonian path problem. A small graph is encoded in molecules of DNA and the “operations” of the computation are performed with standard protocols and enzymes. This experiment demonstrates the feasibility of carrying out computations at the molecular level.

Aha! That opens up a line of progress that is compatible with preserving computation and moving technology closer to nature. We can embed computation in nature, at microscopic levels.

Even better, computation at the molecular level turns out to be potentially extremely enery efficient:

…in principle 1 joule is suffcient for approximately 2 * 10^19 such operations. This is remarkable energy effciency considering that the second law of thermodynamics dictates a theoretical maximum of 34 * 10^19 (irreversible) operations per joule. Existing supercomputers are far less energy effcient; executing at most 10^9 operations per joule.

I feel better now.