Caleb Scharf

The Ascent of Information

2021 003.54 SCH Tigard Library

I wanted to like this book. Rather than surprising, it was prolix and sometimes inaccurate.

• p37 planetwide rate of energy production, early 1800s 100 GW, year 2000 "in excess of a fearsome" 17 TW

  • Source https://ourworldindata.org/energy-production-consumption

    • Shows US power consumption decreasing by 1% in 2019; efficiency is improving, and digital controls help with tht
  • The earth absorbs and re-emits perhaps 1 kW/m2 solar, the sun-facing disk is about 1e14 m2, so the Earth radiates perhaps P=1e17 W at a top-of-stratosphere temperature of 200K or so. Emission is proportional to T^4, the derivative ΔT/ΔP ≈ T/4P ≈ 200/4E17 = 5E-16, so the heating from the "fearsome 17TW" is 5E-16*1.7E13 = 0.0085 Kelvins.
  • The actual "fearsome" question is how the energy is created; CO₂ from carbon combustion adds perhaps 3000x long term global heating compared to the technically-useful energy produced. Nuclear would be vastly safer, especially if the fuel was decommissioned weapon fissionables intentionally adulterated with U238 or Pu240.
  • hence ... fooey

• p37 47 GW for computation ... growing by 40% per year
  • Source: Nature 561 9/2018 163-166 doi:10.1038/d41586-01-06610-y

  • https://www.nature.com/articles/d41586-018-06610-y

  • shows graph from Anders implying data center energy (eyeball) 670 TWh/y (76 GW) in 2020, 2980 TWh/y (340 GW) in 2030, 4.44X growth, 16% growth per year.
  • the Green Growth graph shows ICT purchased half of corporate renewable energy (2GW) in 2016.
  • HOWEVER, the last actual datapoint was 2017, so everything after that is P.O.M.A. ... Pulled Out of My Ass
  • This text quote is more accurate: "Last year’s IEA report estimated that although data-centre workloads will shoot up — tripling 2014 levels by 2020 — efficiency gains mean that their electricity demand might sneak up only by 3%"
    • 3% in 6 years is 0.5% per year, not 16%.

Here's the cited Anders article: https://www.mdpi.com/2078-1547/6/1/117

• Figure one shows (expected) total consumer device power usage dropping from 1050 TWH/year in 2010 to 670 TWH/year in 2030, or 120 GW to 76 GW. Customers want battery life, not hand heaters.

What the graphs are actually measuring is the rapid globalization of internet access, combined with the rapid drop of energy per customer as electronic efficiency increases. Most of the world has some form of internet access in 2020; unless we train ants to use "smart" phones, we are approaching peak power, not exponential runaway.

• p040 "But the smaller the chips get, the more inefficient it gets to push electrons around all that increasingly narrow material" ... BULLSHIT
  • scaling: length goes down, wire side area goes down, voltage goes down, capacitance goes down, and energy is CV²

  • n040 p305 Hasler Marr 2013

• p040 Single bit of neural information, 10,000 ATP -> 3200 eV -> 5e-19J = 0.5 aJ, CMOS a few aJ

  • n040 p305 The metabolic cost of neural information 1998

  • n043 p305 S. Nakamoto Bitcoin paper

• p047 Bitcoin 7.7 GW

  • n047 p306 2021 78TWh/y = 8.9GW

  • n048 p306 Beyond the Bitcoin Bubble contradicts Scharf

• p051 Yann LeCun in 2025 VP Chief AI Scientist at Meta

  • p055 US consumed 200 billion punch cards in 1967

  • p071 misdefined Nyquist rate as signalling rate - no, multilevel can add bits inversely proportional to logarithm of noise

• p096 Generative Adversarial Networks

• p104 'When Darwin throws "mental endowments" into the mix, our modern alarm bells also go off.' PC-BS
• p108 'Any failure to be accurate, precise, or quick will not turn out well'
  • also BS, Darwin made many many mistakes and LEARNED from them. No risk, few gains.
• p111 mirror image perceived same thing, except letters like b and d, we override or undo mirror invariance
  • BS; knowing left from right essential to navigation

• p113 intrinsic forgetting

• p121 old weather from ship's logs

• p121 Digital Humanities Laboratory

• p123 Milkweed emetic for birds, monarch butterfly caterpillars consume it to make birds vomit

• p124 evolvability

• p134 altruism propagates shared genes in relatives, easier than having children

• p135 kin selection Hamilton's Rule

• p138 horizontal gene transfer between species by bacteria

• p138 1600 gene regulation molecules in humans

• p139 epigenetic changes persist over generations

• p141 Tripitaka

• p142 16S rRNA

• p143 Julius Rebek synthetic autocatalytic molecules

• p143 quines software replicators

• p144 De novo gene birth evolve from noncoding DNA stretches

• p144 perhaps remnants of earlier genetic experiments

• p148 Holobiont symbiotic ecological unit

• p148 dataome

• p149 A symbiotic view of life: we have never been individuals

• p150 It's the song, not the singer holobiosis

• p152 dataome increases energy use but changes planet. "conflict" or opportunity

• p150 ASIP and MC1R genes, prey with lighter colored coats in lighter environments

• p154 corg author's neologism for "core algorithm" metaphor for convergent adaptations

• p159 predictable versus Poincaré mathematical "chaos"

• p160 Evolutionary Algorithms

• p164 evolutionary biologist David Krakauer

  • n164 p319 The Information Theory of Individuality

• p169 C₆H₆ 217 isomers

• p171 bacteriophage lytic or lysogenic reproduction

• p172 viruses switch modes based on signalling arbitrium peptides

• p174 poliovirus Sociovirology

• p174 extracellular vesicles

• p177 information ratchet

• p177 Sara Walker Az.S.U.

• p178 Jim CrutchfieldSteen Rasmussen Takashi Ikegami Eric Smith

• p181 Hokusai The Great Wave off Kanagawa 1831 Kanagawa

• p190 Toyabe et al . Information heat engine - 2010

• p191 computational bit energy less efficient than biological brain

• p192 Landauer Limit 3e-21J at 300K

• p192 blowfly visual system 1 bit is 1e-14J

• p195 Activity-regulated cytoskeleton-associated protein ''Arc'' Gene synaptic activity to long term potentiation

  • 1995 https://www.pnas.org/doi/abs/10.1073/pnas.92.12.5734 Worley

• p198 Aleutian Islands 1200 mile arc, 44 active volcanos

  • n199 p322 "Constructor Theory" Synthese, 2013

• p203 predicts (in 2020) 20 billion Internet of Things devices this decade. Statista projects 75 billion in 2025

• p209 Joan Clarke Turing codebreaker colleague

• p211 Von Neumann 1957 (age 53 cancer ), Turing 1954 age 41 suicide)
• p212 machine learning, not yet artificial intelligence
• p213 "planet shadow tube of darkness to infinity" ... no, a conical dark umbra with a slightly dimmer penumbra
  • fix AI L1 nomenclature to "umbra"

• p222 Hox genes body plan

• p225 65 billion chickens slaughtered per year (age 6 to 12 weeks)

• p227 c. elegans nematode evolves to bisexual in 70 generations

• p231 Bohr to Pauli: "We are agreed that your theory is crazy. The question that divides us is whether it is crazy enough to have a chance of being correct."
• p234 ... the Great Dying 251 MYA (the Permian-Triassic extinction

  • Michael Benton, When Life Nearly Died: The Greatest Mass Extinction of All Time, 2003

  • Yadong Sun et al "Lethally Hot Temperatures During the Early Triassic Greenhouse", Science Oct 2012: p366-70, doi:10.1126/science.1224126

• p234 251 MYA Permian-Triassic extinction

• p234 96% of marine species, 70% of terrestrial vertebrates griffinfly

• p235 plate tectonics recycles ocean floor in 200 million years

• p235 Emeishian Traps The Great Dying?

• p236 40℃ oceans

• p236 disaster taxa, pioneer organism

• p237 Triassic:first true mammals

• p237 Olivia Judson The Energy Expansions of Evolution

• p238 Heinrich D. Holland The oxygenation of the atmosphere and oceans

• p239

• p239

• p240

• p240

• p241

• p241

• p242

• p242

• p243

• p243

• p244

• p244

• p245

• p246

• p246

• p248

• p249

• p249

• p249

• p250

• p251

• p252

• p252

• p253

• p255

• p258

• p259

• p263

• p263

• n265 p327 Erling Thyrhaug May 2018 Identification and characterization of diverse coherences in the Fenna–Matthews–Olson complex

• much speculative ... blather? ... about distant future
  • My take "we haven't seen farther than future beings, because we haven't stood on their shoulders yet" Speculating on what they will see and do based on incomplete 21st century technology is wanking
• p273 Dyson's over-the-top-proposal - two meter shell made from Jupiter, capturing all the Sun's energy.
  • "Search for artificial stellar sources of infrared radiation", Science June 1960, p1667-68. doi:10.1126/science.131.3414.1667
  • Dyson's argument was for Infrared SETI, not the notion that humanity will actually disassemble Jupiter to build a "spherical shell" massing 200 grams per square centimeter (2 tonnes per square meter) at twice Earth radius (3e11 meters).
  • The gravitational forces on that shell would be enormous, and crush strength "Einsteinian".
    • The orbital period at R = 2 Re ( ≈ 3e11m ) is 2√2 years ≈ 9e7 seconds, so the gravitational acceleration is
    • a = R * (2 π / T)² ≈ 1.5e-3 m/s².
  • The mass of the shell is M = 4 π R² × 2000 kg/m². The total distributed "rim" compression force is M/4 times a:
    • F = π R² × 2000 kg/m² * 1.5e-3 m/s²
  • The compression force is distributed over the circumference ( 2 π R ) times the thickness t (3 meters?) so the pressure is
    • P = F / 2 π R t = 1.5 ( R / t ) Pascals or 1.5e11 Pascals.
  • Jupiter is mostly hydrogen, a gas at 2 Re radius. If the hydrogen was (somehow) transmuted to perfect diamond, that would result in a compression strength of 470 GPa = 4.7e11 Pascals However, the density of diamond is 3500 kg/m³, so the shell thickness t would be only 0.57 meters, and the required compression strength climbs to 790 GPa.

  • It is amusing that these numbers are in the same ballpark ... but the bottom line is that a Stapledon-Dyson shell (his term, not "Dyson sphere") is just not strong enough that close to the Sun.

  • And even if it was - over those vast distances, the shell has effectively zero bending strength; restoring forces in a gravity field are negative, and it will rapidly buckle.
  • At 10 Re, the gravitational force is 25 times smaller than 2 Re, and thickness and the areal mass density is also 25 times smaller, so the compression pressure is 25 times smaller. Diamond would work ... but would still buckle.

  • A much thinner shell can be supported by actively adjusted light pressure.

• p274 Landauer limit ... evasion if species smart enough? BS

• p278 Boltzmann Brain

  • n279 p328 Albert Einstein Physics and Reality 1936

  • n280 p328 Scott Aaronson 2016 The Ghost in the Quantum Turing Machine in The Once and Future Turing

• p283 Seth Lloyd PhyRevLet 88 Computational capacity of the universe

• p284 "our species ever-dwindling attention space" ... individuals have attention, and some waste it. Granfalloon

• p286 "Treat information like a natural resource" ... fuzzy thinking
• p286 "assigning a price to the "emission" of information ... more fuzzy thinking

• p289 Jaron Lanier 2013 Who Owns The Future Bvtn 303.4833 LAN