The big winners of the last 200 years of the Industrial and Information revolution were ones who created previously unimagined scale to win.
We went from many monks and scribes hand writing manuscripts to the printing press and eventually to the reel to reel newspaper presses. The circulation of daily weekday newspapers in the United States peaked in 1987 at over 62.82 million. There would be hundreds of large pages in each major newspaper.
Cars went from dozens of handbuilt cars to factories producing millions of cars per year.
OpenAI and Google scaled large language models by 1 million times over the past 6 years. The latest LLM have 100 billion and even a trillion tokens.
OpenAI exploded into the lead in AI by having silicon valley companies and venture capital fund a multi-billion bet to scale the compute behind the LLM.
Cloud computing became a multi-trillion business by successfully making big scaling bets. Much of the success and capabilities with computers and semiconductors involved pushing to lower marginal cost with immense scaling.
Tesla FSD version 12.X has gone to pure neural nets. It will be able to scale and improve performance purely with more data, training and compute. More compute will need more energy. 1000 AI Exaflops will likely need about 1 gigawatt of power.
2023 has been awesome for Optimus.
We’ve moved from an exploratory prototype (Bumblebee/cee) to a more stable, Tesla-designed platform (Optimus Gen-1).
We’ve improved our locomotion stack, frequently walking off-gantry without falls and with a faster, increasingly more… pic.twitter.com/yCnD0SThBd
— Milan Kovac (@_milankovac_) December 31, 2023
Winning new multi-trillion markets is just the beginning by pushing LLM, humanoid robots and self driving car and trucks to complete success.
The industrial revolution grew the world economy by 100 times over 150 years from 1 trillion to 100 trillion. An AI, humanoid bot and self-driving vehicle revolution can increase the world economy from 100 trillion to 10 quadrillion in 40 years or less.
The AI race will be all that matters.
The industrial revolution started with people using whale fat to light candles. There was no concept of powering factories because factories did not exist. The first factories and the trains that fed the supply chain used wood and coal power.
Coal power was mostly displaced with oil and natural gas and electricity.
The AI race and technology transformation will settle the question that Peter Thiel raised about the lack of true technological progress. The stalling out of technology for the last few decades is going to be swept away with rapid change enabled by AGI/ASI, trillions of humanoid bots and self driving cars and trucks. There will also be mass produced spaceships. This new technology age will need thousands, millions and billions of zettaflops of compute. This will need multiples of the current 30 terawatt hours of electricity.
More Energy ==> More Compute ==> Better AI ==> More demand for better LLM, humandoid bots and self driving vehicles
More Energy, more money, more technology ==> Scaled development and colonization of space starting with the solar system
True societal and technological change will be visible in a new 100X, 1000X scale of energy and compute infrastructure and society.
People look at movies about different periods. 0 AD, 1700, 1800, 1900 and think that people just wore different clothes and had different tools. But the scale of civilization was different. 190 million people in 0 AD and 600 million around 1700. In 1776, the US is formed and the whole world has about 800 million people.
WW2 used the amount of oil in a year that we now use in a few weeks or a month. The US was producing about 80% of the world’s oil in WW2. The Nazis, UK and Soviets were fighting with about 10-20% of the worlds oil. The total 1940 oil was about 15 times less than we use today. The Nazis, UK and Soviets were using less than 1% of the oil we use today. The critical battle of Stalingrad was to get a relatively tiny amount of today’s oil.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
AI development being power constrained is the same as being compute constrained. Unfortunately we live in a time where it is difficult to produce power.
I really wish Texas would commit to making 10x BWRX reactors to shore up their grid and to enable future expansion. Batteries are nice but you can’t keep growing your population by 20% and compensate with batteries.
Many future tech industries are power limited. BTC. AI. If you can produce power cheaper than other states then you can draw these industries to your state.
Both Brian and I should live long enough to see human population triple in our lifetimes, by ~2030, making the big assumption that expanding wars, plagues, or environmental/food production collapse doesn’t drastically curtail those projections. Both long and short term history show that rosy projections are unrealistic; something always intervenes, whether the Black Plague or Covid (American lifespan has declined by 2 years and may not resume upward trajectories, given lasting cellular damage + things like drug addiction sped up by Covid related lockdown policies & other stressors that have proven very resistant to techno-optimmism). Solutions may be there, but simply increasing population & throwing money at the problems won’t be enough. Too much human potential is being lost already, but those humans still eat & use other resources. Maybe 90% of humans are progress neutral – just subsist & add nothing – or even progress negative – are criminals, war-makers, etc.
[ ‘The AI race will be all that matters.’
(Who?) Define ‘work'(&art&cultural progress& ) within one generation of humans and salary/wage (IQ&EQ) with these changes?
Being&Taking part on these developments (cost of ‘AI’ hardware, data connection, electricity) being a cookshop worker in India/Africa/Asia? ]
How long until an artificial womb begins to produce humans beyond replacement needs?
Future generations may not understand the concept of ‘mother’ or ‘father’.
Spooky.
How long before governments and/or corporations begin clandestinely “editing” the brains and bodies of the humans in those artificial wombs? Perhaps a totalitarian state would want pliant sheep to keep the economy running. For that matter, who says one of these crockpot humans need be conscious at all?
Prediction- there will be a major shortage of ultra high end computer chips and that will slow AI adoption.
One trillion Optimus would require 125 trillion pounds of materials.
And, all the equipment that is required to mine and process that material.
And, the massive factories necessary to bring the parts together.
And, the power they need to operate. ~2.3 kWh per charge, times 1 trillion = 2.3 quadrillion kWh per day.
America used about 4 quadrillion Wh total in 2022.
500 times too little.
If Optimus wants to bulk up its numbers like that, AI is going to need to make some massive tech breakthroughs.
[ ‘125 trillion pounds of materials’ ~57km³ of material
That’s about processing(extract,convert,transport,repair/recycle,deposit) 3times Utah’s Great Salt Lake water volume (~19km³) or about half of Qinghai Lake (~108km³)
Volume of planet Earth ~1.08321*10¹² km³ (~1 am. trillion km³)
2022’s worldwide crude oil demand is ~5km³ (guesses on reserves left are about ~205-270km³ (~2015, 1.3 trillion barrels, optimistically including new fields ~1.7 trillion barrels, ~40yrs @ 5km³/yr) vs. ‘In its 1995 assessment of total world oil supplies, the U.S. Geological Survey (USGS) estimated that about 3 trillion barrels of recoverable oil originally existed on Earth and that about 710 billion barrels of that amount had been consumed by 1995. The survey acknowledged, however, that the total recoverable amount of oil could be higher or lower—3 trillion barrels was not a guess but an average of estimates based on different probabilities.’, crude oil production since beginning ~221km³ )
‘Known reserves of petroleum are typically estimated at around 190 km3 (1.2 trillion (short scale) barrels) without oil sands, or 595 km3 (3.74 trillion barrels) with oil sands. Consumption is currently around 84 million barrels (13.4×10^6 m3) per day, or 4.9 km3 per year.’
crude oil energy content (1 barrel, 42gal, 159l) ~1.7MWh, ~10.7kWh/l
1/5 of crude oil energy is consumer ready fuel energy content (‘https://ars.els-cdn.com/content/image/1-s2.0-S0301421513003856-gr1.jpg’)
190km³ crude oil reserves (with 1/5 available for global development&daily demand, with preprocessing and infrastructure maintenance done) ~400000TWh(?)
EROI for fuels (coal, oil&gas, from tar sands or biomass) ‘https://ars.els-cdn.com/content/image/1-s2.0-S0301421513003856-gr2.jpg’
Boston Dynamics Atlas is at ~9kWh on average duty for all day capacity (battery ~3.7kWh, mixed mode ~90min/568Wh, 180min(-4hr) on standby.
Tesla Optimus ~500W for 5mph walking speed (~12kWh/day), ~100W for sitting on standby (~2.3kWh/day)
1 trillion Optimus (2023 spec), on average duty (~6kWh/day?) ~6*10¹² kWh/day (4 quadr. Wh =~ 4*10¹² kWh =~4050TWh electricity for a year 2022(?) )
Estimation from automobile manufacturing is very roughly ~41.8MJ/kg [Sato et al. (Fernando Enzo Kenta and Nakata, 2020) ‘According to their findings, the energy needed for automobile manufacturing is 41.8 MJ/kg per vehicle, with mining and material production operations accounting for 68% of overall energy usage which means 13.376 MJ/kg energy consumption just for manufacturing.’], what’s ~57kg*41MWs =~650kWh for a lower boundary estimation (high tech density car vs. robot)
1 trillion Optimus * ~1MWh =~1000000TWh (means oil reserves, including oil sands, are more or less used up for 1 (or ~1.5) trillion Optimus production number) (?) ]
I read this.
Hard to nail down exact numbers, but all the passenger vehicles built ever comes to about 2.4 billion.
~500 times short than a trillion.
Of course, if 95% of O-bots are engaged in off-world activities, that leaves a paltry 50 billion O-bots to get things done on Earth.
Still, thats ~20x the total number of passenger vehicles on Earth.
[ A passenger vehicle’s initial production energy is ~10-30MWh (above ~41.8MJ/kg =~11.6kWh/kg for a ~1000kg car =~11.5MWh), what’s for 1.4-1.7-2.4 billion ‘cars’ was a total energy investment of ~14000-72000TWh (or 1.3-6.7km³ crude oil equivalent, what’s up to 1/2 year global energy consumption on todays scale or 1/2-3yrs all of todays electricity)
Another measure for consumer goods average energy content is ~1kWh/$1 spent (what fits mostly with average passenger vehicle prices).
If a mass production Optimus ~$20k-10k-7k-5k, that would allow energy requirements being estimated ~5-20times that above number for a trillion, with ~1 million O-bots then roughly ~(1-)5-20TWh (times 50000 for 50 billion, that’s ~20x average energy requirements what passenger cars production was, also and ~2-6yrs global(2022) primary energy consumption equivalent)?
yr1978 global primary energy consumption ~290EJ (~81000TWh)
yr2000 global primary energy consumption ~400EJ (~110000TWh)
yr202x ~600EJ (~160000TWh, unprocessed crude oil equ. ~15.5km³), electricity ~25000TWh
‘https://ourworldindata.org/grapher/global-primary-energy’
2022_160000TWh ~18.3TW (7.963 billion people) ~2.3kW
1 billion O-bot production ~20000TWh ~2.3TW ~2.3kW_each_O-bot power requirements for production and ~0.1-0.5kW for in service(?) ]
Your numbers are garbage.
“…One trillion Optimus…”
We don’t need a trillion, so strike that right off.
“…And, the power they need to operate. ~2.3 kWh per charge, times 1 trillion = 2.3 quadrillion kWh per day…”
Nope. Humans are relatively inefficient. An electrically run device could be far more efficient.
https://en.wikipedia.org/wiki/Human_power
“…Normal human metabolism produces heat at a basal metabolic rate of around 80 watts…” (Note: Heat not work.)
“…Over an 8-hour work shift, an average, healthy, well-fed and motivated manual laborer may sustain an output of around 75 watts of power….”
“…During a bicycle race, an elite cyclist can produce close to 400 watts of mechanical power over an hour and in short bursts over double that—1000 to 1100 watts…. An adult of good fitness is more likely to average between 50 and 150 watts for an hour of vigorous exercise. Athlete human performance peak power, but only for seconds, 2,000Watts…”
Combining answers to the next post,
“…Tesla Optimus ~500W for 5mph walking speed (~12kWh/day), ~100W for sitting on standby (~2.3kWh/day) …”
The above numbers on actual human power show these numbers to be absolutely absurd. When it’s in standby it could go into low power mode. Microcontrollers, right now, use micro watts on low power mode. You telling me we are too stupid to do this with bots? No.
For reference, a good horse working at a good constant rate works at 746Watts.
As for oil usage, this is not a problem. The cost of solar is very, very, very close to being cheaper than drilling for oil or mining coal.
https://rameznaam.com/wp-content/uploads/2020/05/Solar-Costs-2010-2020-vs-Fossil-Fuel-Cost-Range.jpg
“…As an example, using only desert military bases in California and Nevada, which receive a lot of sun, would produce enough power to supply the entirety of North America. In practice, a mix of rooftop solar and utility farms in sunny areas is the most robust approach.
It’s instructive to consider areal land uses for other forms of energy production. In Australia’s picturesque Hunter Valley, there are several enormous open cut coal mines. I computed that a solar farm operating for 20 years will produce energy equivalent in value to a coal seam 3m thick. That is, even if the coal is at the surface, and it’s not, if it’s less than 3m thick it’s better to use as a foundation for solar panels than to dig it up and burn it…”
https://caseyhandmer.wordpress.com/2018/10/01/the-program-or-a-strategy-for-universal-prosperity-in-the-twenty-first-century/comment-page-1/#comment-9926
There’s also plenty, plenty by far, enough energy, land etc. for all the food anyone could ever want. There’s more than one way to do this. Jim Bowery does the numbers on building “vorex” Islands. By using vortex power stations in the tropics we could build enough islands, with condo sized apartments, that also produce food, housing and materials for ALL the total human race in 15 years.
http://jimbowery.blogspot.com/2014/05/introduction-extinction-of-human-race.html?view=sidebar
Other ways are to stop using the super inefficient farming we do now. There are already people building food products by using electricity to produce sugars and energy to grow genetically engineered yeast and bacteria. No, we don’t have to eat bugs. There’s no reason that we can not engineer the same yeast and bacteria to produce prime rib, lobster, corn or whatever you desire. It’s only engineering. No vastly new science is involved. The efficiency of plants is a dismal less than 2% and that’s for the whole plant. The actual portion that we eat is way less. I would be surprised if it were 0.5% efficiency in a sunlight per actual food basis. With electricity based food production, some of the numbers they are getting right now, food per watt is 200% better.
There’s a company that takes electricity , air and a few minerals and grows, with fermentation like beer, protein and carbohydrates.
“…We feed the microbe like you would feed a plant, but instead of watering and fertilising it, we use mere air and electricity. With our current process, this is 20x more efficient than photosynthesis (and 200 times more than meat)…”
https://solarfoods.com/science/
So for a rough approximation, let’s assume a plant covers all the space that sunlight falls. Each meter of space is about 1,000 watts. If the plant uses every inch of it, it doesn’t, then we have at 3% efficiency or 30 Watts worth of actual plant matter. Of course most of the plant is not edible but for comparison we assume it is.
With the new system fermenting food instead of 30 watts worth of food energy we get 20 times more so it comes in at 600 watts.
If solar panels are used at 20% efficiency then we get 200 watts. Taking off another 25% losses in the process, we could say that the fermentation process for food could get us 150 watts of food per square meter.
Let’s assume we only get 8 hours of sunlight a day to cover all the losses and inefficiencies. We come up with 150 watts per meter times 8 hours equals 1,200 watts a day per meter.
Now we divide the total needed watts to feed everyone every day by the amount per square meter to get the amount of square meters needed every day.
136,000,000,000,000/1,200=113,333,333,333. square meters
113,333,333,333 square meters=43,758 square miles
So a desert area that’s 209 miles by 289 miles could feed everyone on planet earth with a large amount of food every day. Maybe it would need to be a little bigger but I overestimated constantly so it wouldn’t be that much higher. With nuclear power the area could be way smaller.
My point is that all these people keep telling us we need to eat bugs and there’s too many people so we should kill them all off. They are liars and evil and it can be proved. I bet we could perfect all this for far less than the $5 billion(the number has gone far higher than this) they want to pack our country with aliens. At the price we’ve spent on murdering Russians and Ukrainians, we could afford to feed everyone on earth like this for decades and let them stay where they are.
We have several companies pursuing this. It will take a while but the economics are such that it will happen. There’s a company that sells milk made this way. They have sold ice cream and whey powder. It doesn’t have to be meat or protein or milk it can be any sort of food at all. The wastage would also go down dramatically. You could send the food dry and then add water, so none would be lost in transport. It would likely add a great deal to the time it could be stored.
https://en.wikipedia.org/wiki/Perfect_Day_(company)
https://www.esa.int/ESA_Multimedia/Images/2018/10/Protein_from_air_and_electricity_only
Your numbers for the cost of materials is also WAY WRONG. We could use concrete or even better geopolymers mixed with graphene or carbon nanofibers. The price of these fibers has gone way down and assuredly will go even lower. Graphene mixed with bioplastics is SUPER STRONG. Look at this video where he puts graphene in a bioplastic, I’m fairly sure it’s casein or milk plastic. He’s been shooting thin layers of this will 9mm also, and it does not go through. He has hundreds of videos showing how to cheaply make graphene.
https://www.youtube.com/watch?v=iqOCtEsMWjs
The only thing stopping us from making life on the planet super cheap and materially much better is the assholes, psychopaths and wretched blackmailed politicians who do their bidding.
[ Less about right or wrong (with our level of information, e.g. Optimus (mechanical) standby could include computing support towards a network worth a ~100W equivalent value?), just accepted the thesis that was offered
“The stalling out of technology for the last few decades is going to be swept away with rapid change enabled by AGI/ASI, trillions of humanoid bots and self driving cars and trucks.”
Maybe it’s about going in front of the people who want&will follow Your ideas and way of life. Probably they are not all the people of the human species and very likely the others have different POV, needs and possibilities.
Yes, a trillion O-bots within a next decade is (probably) a garbage number.
~2 billion passenger vehicles is a ~century enduring production result. ]
One small correction is the first factories were water powered not coal powered. Romans used water wheels for industrial scale milling of wheat into flour. Later water powered textile mills.
One large correction is exponential population growth does not continue forever. All populations obey the same growth curve. The curve has a lag phase, an exponential growth phase, a plateau phase, and a catastrophic 90% population decline. Human population data show we are on the plateau and near the 90% population catastrophe.
We must also account inflation. 1 dollar some time ago was worth way more than now.
How much is living standard or ordinary people for the better? Can they build their houses way easily or not?
[ Me guessing, we (mankind) don’t even have available energy reserves (without fusion plants available&previous reserves of fossil fuel) for rebuilding all goods available&necessary for this preset standard of living on this planet (extended to industrial countries average for all ~8 billion humans) within one generation of human species (~25yrs)?
‘https://en.wikipedia.org/wiki/Net_energy_gain#Examples’
‘During the 1920s, 50 barrels (7.9 m3) of crude oil were extracted for every barrel of crude used in the extraction and refining process. Today only 5 barrels (0.79 m3) are harvested for every barrel used.’
‘https://en.wikipedia.org/wiki/Energy_cannibalism’
‘In the article “Thermodynamic Limitations to Nuclear Energy Deployment as a Greenhouse Gas Mitigation Technology” the necessary growth rate, r, of the nuclear power industry was calculated to be 10.5%. This growth rate is very similar to the 10% limit due to energy payback example for the nuclear power industry in the United States calculated in the same article from a life cycle analysis for energy.’
‘https://en.wikipedia.org/wiki/Energy_return_on_investment#ESOEI’
‘Vanadium redox battery 10
NaS battery 20
Lithium ion battery 32
Pumped hydroelectric storage 704
Compressed air energy storage 792’ ]
More silliness. The numbers have been done. We know exactly how much sun we get, we know the efficiency of solar cells. We know the rate of decrease of cost of cells. We know the exact amount of energy needed to make methane, diesel fuel or other fuels from air and water using solar. The tech for this is many, many decades old. This guy is building solar farms to produce methane and eventually diesel fuel. As soon as the solar cost comes down a little more, he will deploy in the desert in California.
“…As solar cost at the array falls below $10/MWh in the next few years, it will be cheaper to synthesize hydrocarbons from CO2 and water vapor than to drill them from the ground and refine and transport them. Synthetic fuels are cheap, local, carbon neutral, and above all scalable solutions to both fuel scarcity and CO2 emissions. Developing this tech is the primary focus of my professional effort at my start up, Terraform Industries…”
We could have boats full of solar cells, or even better, linear accelerator reactors, ride around the oceans making fuel out of air and water.
https://caseyhandmer.wordpress.com/2023/10/11/radical-energy-abundance/
https://rameznaam.com/2020/05/14/solars-future-is-insanely-cheap-2020/
Look at these numbers again.
https://rameznaam.com/wp-content/uploads/2020/05/Solar-Costs-2010-2020-vs-Fossil-Fuel-Cost-Range.jpg
In fact with the newer Perovskite solar cells, possibly combined with silicon, we may get there much faster than has presently been calculated by even the most optimistic.
“…Solar-cell efficiencies of laboratory-scale devices using these materials have increased from 3.8% in 2009[3] to 25.7% in 2021 in single-junction architectures,[4][5] and, in silicon-based tandem cells, to 29.8%…”
https://en.wikipedia.org/wiki/Perovskite_solar_cell
And BTW, I’m not in any way motivated for solar by CO2 emissions. I don’t believe we are all going to burn up in 20 years. My motivations are energy independence and anti-oligarchical de-control. With cheap enough solar people can have their energy needs met without the oligarchs controlling them. They can’t shut the power off.
[ scale production of solar cells (provided with energy from solar power) including installation/construction/grid&storage&fuels/maintenance/upgrading has energy dependent concurrency with other scale production (e.g. information&networking, space exploration, climate disasters precautions&repair, agricultural restructuring, mobility, different energy concepts/preferences/needs (e.g. nuclear, geothermal) )
What’s a number for EROEI for solar/wind/geothermal(/nuclear) power to fuels (including CO2 processing)? ~15(about coal ages energy efficiency and ‘living standards’/’social chances’/’public infrastructure’, but different priorities? comparable?)
‘https://en.wikipedia.org/wiki/Net_energy_gain#/media/File:Net_energy_cliff.gif’
(cheap fossil oil(&gas) on best EROEI was up to ~50)
If efficiency improvements are cheap, yes, otherwise it’s more about cheap&robust solar on global scale(?)
We have no controversial about independence within societies controlled by resource-accumulating&egocentric&influential groups acting to only their own advantage (even within democracies, see what’s ‘fortune accumulation’ within companies/for individuals compared to average surroundings?) ).
The technical difficulty might be with not being available, even with ‘low cost&price’ for solar energy converters, because of ‘limited time’/working force/materials/energy resources(/geo-political instabilities from lower income by fossil fuels)/accessible&suitable surfaces/areas for production&utilization?
EROEI vs. (theoretical, instant/current) cost(?) ]
[ with reference to availability of materials and work invested for a capacity factor, there’s a description close for system integration named LFSCOE(?) (levelized full system cost of electricity), enabling comparison for electricity system differences also (e.g. capacity(vs. utilization) factors or supply demand matching, grid size etc.)
‘https://www.sciencedirect.com/science/article/pii/S2352484723010569#sec5.5’
The Changing Nature of Energy Return over Time
(&Figure 1. The Historic Development of Energy Expenditure in the United Kingdom)
‘https://www.sciencedirect.com/science/article/pii/S2590332219302209#sec1.5’ ]