(Some random thoughts on self-replicating machines.)

One neat thing about the hexapod is that any errors in manufacture (except for play/slop/lash) can be calibrated away by using software. The accuracy of movement depends on the accuracy of the reference standard rather than the accuracy built into the machine. This means that the thousandth generation of hexapods is just as accurate as the first.


With enough wangling of the terms, just about anything can be defined as self-assembling. An apple falling on the ground is self-assembling because it required no input of control to ensure that the apple ended up on the ground.

Replication should end up with something useful. A prion or a crystal or a flame end up making copies of themselves, but they do not create functionality. Viruses have subtle side effects such as horizontal gene transfer, and this is probably why they are allowed to exist. A basic unit of functionality to consider as a goal for replication is Turing's Logical Computing Machine, or simply a Turing machine.

It turns out that the above challenge isn't as difficult as I had originally thought. A self replicating computer in N easy steps:

  • Generate (in vivo) DNA origami tiles with piggybacked carbon nanotube semiconductors that will self-assemble into an FPGA *1
  • Program said electronic FPGA to act like a computer, using an ion channel leading outside the cell to provide the necessary voltage to power the inorganic circuitry.
    • by using a modified ribosome to translate an mRNA sequence into an FPGA bitstream signal
    • Alternatively, use optical signalling between {fluorescins, gold nanoclusters, nanowire antennae, quantum dots}*2 in a simple serial protocol to map out errors in the newly assembled individual FPGA's physical structure. The other computer then compiles and transmits a customized bitstream that gracefully works around errors in that particular FPGA's structure.
    • if you really really want a perfect FPGA structure for some reason, the origami tiles could have an electronically programmable hasp in the inorganic potion, which would make them more resistant to coming apart. Then there would be temperature cycling (like in PCR) to give the tile matrix multiple chances to reassemble correctly.
  • alternatively, make a (very slow) 100% DNA computer with transcriptional switches. *3 In a completely amorphous computer, these have the annoying problem that too many switches means a combinatorial explosion in the 'address space' consisting of toe-hold sequences and resulting cross-talk. in order to ensure low cross-talk, the sequences are hard to compute for complex circuits with many addresses, even with fancy modern computers.
    • if two trascriptional switches a certain distance apart have the same address, they will interfere with each other.
    • as the distance increases the signal to noise ratio increases. we can tolerate a certain amount of cross-talk, so if the DNA sequences are laid out properly on a grid such that the cross-talk (noise) level is far below the switch threshold, it shouldn't matter if they do interfere because the switch won't turn on.
      • from this we can simply build a grid of DNA origami tiles with dangling transcriptional switch genes.
  • cell divides, process repeats

Hierarchical macrostructures appear from many non-periodic tiling sets. This particular macrostructure's "Ammann bars" resemble the internet-0 radio antennae used by MIT fablabs. Antennae self-assembled from DNA tiles could serve as a link between the nano-scale world and the "real world".

*2 all this inorganic crap can be injected into living cells by using liposomes, but an entirely biological mechanism is best since it would require no intervention

Front page   Diff Backup Reload   List of pages Search Recent changes   Help   RSS of recent changes
Last-modified: Sat, 21 Jun 2008 12:22:57 GMT (934d)
The Gingery Machines wiki is published under the
Creative Commons Attribution-Share Alike 2.5 License Creative Commons License