The omega machine may be regarded from the computer science perspective as an embedded system. As in the field of robotics, the concept of there being natural and unnatural divisions of labor between a physical system and a computerized control system is important because the omega machine has real time functionality requirements, in order to fulfill its role as a programmable local combinatorially complex complementation system. In particular, the concept of embodiment , is central to achieving this performance. The need for physical embodiment, apart from its AI semantic foundation, stems from the fact that it is not only efficient but essential for real time distributed functionality to make strong use of the integrated real-time computational problem solving power offered by physical systems in direct contact with the environment of interest. An example of an embodiment principle in the omega machine is the electrical capacitance of the microfluidic cell, which can be utilized to allow multiplexed sequential addressing of many electrodes with fewer control signals. Another example is the innate regulatory nature of chemical reactions: it is often not necessary to provide delicate tuning of a concentration level of a chemical in the microreactor, because it can be regulated naturally in a gradient field.
Realizations of the electronic omega machine deal with video frequency fluorescence data in a million pixel channels or more which must be processed in real time and used to control the local environment of up to a 1000 candidate artificial cells (30x30 pixels per cell) via on the order of 10000 electrodes (of the order of 10 per cell). Future developments may allow these numbers to be increased to the level of a million cells or more. Of course experimentation must first deal with the few cell (1-10) situation involving several hundred electrodes before scaling up to this level. The computational requirements of this first prototype are then much reduced compared with the later versions required for sufficient combinatorial search capability. Multichip reconfigurable electronics boards with a standard computer interface, such as that developed in my lab e.g. NGen52, Meregen using multiple Xilinx FPGAs each connected to multiple SRAM chips, have been shown to provide the necessary computational capability to deal with the full data processing, feedback and control tasks. Note that FPGAs allow modular control electrode group responses, involving complex coordinated Hz to MHz pulse trains on multiple electrodes, to be setup in digital electronics programmably. They also can be used to do feature extraction from the local portions of images, which have already been made available to a single board reconfigurable card at the full rates described above. The role of the standard Intel processors and operating system is then to provide a framework software environment for reconfiguration and non-time-critical software-intensive monitoring and recording tasks.
Because the computer system can deal simultaneously with the individual cells, it can also build up and retain information specific to each of them. This information can then treated as extended genetic hereditary information for the cell: i.e. an electronic genome. Because the information can be either in the form of data (e.g. control parameters) or electronic configurational information (defining the local FPGA electronic processing for feature extraction, control loops or pulsed electrode actuation), this genetic information has considerable functional flexibility in interaction with the cell. Online reconfiguration can further enhance the computational capabilities as variants of control modules are downloaded to the FPGA chips. The information can thus be proliferated both within the FPGA board and in the microfluidics system (where electrode configurations at different sites can be connected to copies of the same control logic), allowing daughter cells to inherit the same electronic information as their parent cells. In an extreme case, this may be the only significant genetic information in the early artificial cells. It is clear also then, that this electronic genome can be used to allow the Omega Machine to support an evolutionary process in artificial cell design and functionality. To avoid misunderstanding with “in silico” artificial cell ideas , it is important to stress that we are describing here true “wet” artificial cells, with real chemistry, but in online interaction with an electronic control system.