Freitas' 2000 analysis of self replicating nanotechnology found that the fastest replicators could be about the size, speed, and power consumption of bacteria. Replication speed is limited by the energy cost of pulling atoms apart to move them to where you want, on the order of 10^-19 J per atom or 100 MJ/kg. https://lifeboat.com/ex/global.ecophagy
Although biology is optimized for energy efficiency, it only uses 300 of the 90,000 TW of available solar energy at the Earth's surface. We know this because the atmospheric carbon cycle is 210 billion tons C per year at 4 Kcal/g of CH2O carbohydrates. Meanwhile we already have solar cells with 20-30% efficiency. This puts DNA based life at a severe competitive disadvantage to nanobots. A straightforward projection of Moore's law puts this around 2150, assuming global storage capacity, currently 10^24 bits, continues to double every 3 years until it surpasses the 10^37 bits stored as DNA in the biosphere. But Moore's law for transistors has ended. We can put 10^11 transistors on 1 cm^2 of silicon but not much more. Assuming half of the chip area is wiring, then transistors are about 20 nm across, or 180 silicon atoms. P and N type silicon is doped to concentrations of about 10^-6, spaced about 100 atoms apart. DRAM capacitors hold about 8 electrons. Any less and the bit error rate is too high We can increase chip density by stacking layers, but this does not solve power consumption. The only ways to reduce power is to make components either smaller or slower. A human brain sized neural network with 10^14 parameters running at 10 petaflops needs about 1 MW of electricity. But the brain uses only 20 watts by transmitting signals at 100 m/s instead of almost the speed of light. Cellular chemistry is likewise efficient. Your body has 10^13 cellular nuclei storing 10^23 bits of DNA transcribing 10^16 amino acid molecules per second at 100 W resting metabolism. In both cases, the power efficiency comes from moving slow, heavy atoms and molecules instead of light, fast electrons. A DNA based pair weighs 600 Daltons, the same as a million electrons, but is a million times slower. It takes 10 hours to copy a fertilized human egg, which is 1000 base pairs per second for the largest chromosome 1 (250M base pairs). Spudcells are the first artificial cells built from scratch capable of self replicating. https://biotic.org/research/spudcell/ https://www.science.org/content/article/lab-created-spudcell-marks-major-step-toward-building-life-scratch They are far from being capable of living in their own. They lack ribosomes for synthesizing protein, so they require borrowed ribosomes from E. Coli, which degrade after 5 generations. They lack proteins for metabolism, so they "feed" by merging with lysosomes, smaller water bubbles surrounded by a layer of fat. They lack the proteins for a cytoskeleton, so they need help dividing by squeezing them through tiny holes, and then the DNA often doesn't end up evenly divided. After 5 generations, only 30% of cells have all their DNA. Spudcells have only 90K base pairs, compared to 40M for chlorella algae, the simplest organism that needs only sunlight, water, CO2, and some minerals to survive and reproduce. For now, nanotechnology is the only way that Moore's law can continue. I expect further progress. -- Matt Mahoney, [email protected] ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf0bc7319c7107249-M614faaf8cc79795b4d3f1e9f Delivery options: https://agi.topicbox.com/groups/agi/subscription
