Shape of things to comeDARPA looking for Genesis-style AI lifeforms

The ghost in the machine. DARPA restless researchers have issued a request for “intelligent” electronic components and chemicals which can “self-organize” themselves to form complex items such as routers, fuel cells, biofuel factories, or medical drugs. Lewis Page writes that reading between the lines, it appears that DARPA is seeking nothing less than the creation of artificial intelligent lifeforms.

DARPA’ name for this initiative is Physical Intelligence, and full details were released last week. According to DARPA, humanity at present has only a dim grasp of what intelligence actually is and how it came into existence:

For the past 50 years, the dominant paradigm for intelligence supposes that the brain is the seat of intelligence and is functionally equivalent to a computer capable of executing any algorithm… the goal of true machine intelligence remains distant… our understanding of the evolution of life is rooted primarily in observations of the natural world… With some exceptions, current approaches to understanding intelligence and evolution are disconnected and often lack grounding in fundamental physical principles.

The idea behind physical intelligence seems to be to achieve a much better, hard-science understanding of what intelligence and life actually is and how it evolves as a matter of physics. Page writes that this being DARPA, this almost God-like intellectual toolkit is then to be put to use.

Although the idea that life is “a struggle for entropy” (Boltzmann) has been supposed for more than a century… applications to engineered systems are scarce. The Physical Intelligence program aspires to change this situation… The objective is to demonstrate the first human-engineered open thermodynamic systems that spontaneously evolve non-trivial “intelligent” behavior…

Specifically, bidders for DARPA Physical Intelligence funds will be invited to design one of two things: electronic gizmos or “basic units that might be described variously as ‘gates’ or ‘cells’ or ‘neurons’”, or alternatively “an open chemical environment.”

The electronic “units,” which may initially exist only in a simulated environment “comparable in complexity to simple video games (e.g., Tetris)” are expected to “self organize” and “evolve” into a complex configuration, presumably one demonstrating some non-trivial aspects of intelligence. As a starter for ten, the super Tetris-block electronic neurocells should be able to spontaneously form into “a continuously self-organizing router for Internet traffic or similarly complex application”. One should then be able to “extract the algorithm, and map it to a conventional computer” — effectively turning that computer into an intelligent lifeform.

As for the vatful of smart-chemicals, they are expected — without human intervention — to be able to form themselves into drugs, organic fuel cells, solar powered biofuel supercrops or “a similarly complex system”.

Page writes that Physical Intelligence may make DARPA gear capable of becoming intelligent life — potentially more capable life than humanity itself. The AI algorithms which evolve from the spontaneously self-organizing Tetris blocks might outclass the human brain: the fuel-celled, solar-powered, self-medicating lifeforms which emerged from the smartware vats would be physically superior to humans.