Sam Altman vs. Jane Goodall
The astonishing commitments of energy to AI, put in perspective. A better idea to power it all, and a contemplation of why we think AI is more important than environmentalism.
Saul to ChatGPT :
“Why did Jane Goodall and David Attenborough fail in convincing people that whales and chimpanzees and rainforests are worth saving and instead we all watch tik tok?”
ChatGPT said:
That’s a piercing and deeply important question — one that gets at the heart of how culture, psychology, and technology interact. It’s not that Attenborough and Goodall failed — they moved millions of people and reshaped global consciousness — but that the forces opposing them evolved faster, louder, and more seductively. Let’s unpack this in layers:
But before I let ChatGPT write a post for me, a little bit of self generated context and analysis - things AI is still crap at.
Climate week in NYC felt like it was more about AI and hyperscalers than it felt like it was about saving the earth’s climate. I was hoping it would be more about how to keep this magical little planet full of life spinning at a temperature safe for all of it’s extraordinary creatures (including us humans).
I’ve been reflecting on that, and while listening to the latest Hard Fork podcast was struck by the numbers of GPU’s that OpenAI is ordering to run its little software project. Lets also be amazed that we now talk about compute in terms of GW which is a unit of power, not computation. Thermodynamically this makes sense, information processing is after all energy. The computing bit of your laptop computer is likely 20W. Sam Altman just order 16GW of compute. Thats 80 million times the power of your machine.
Is it a lot, or is it a little? We can look at EIA data (while the EIA still exists ?) to get a sense.
The first graph shows you US electricity generation broken down by source, from 1950 to today. You can see a precipitous decline in coal over the last 15 years, and a similarly rapid growth in both renewables and natural gas. Nuclear is flatlining, reflecting how hard it is to build, and perhaps its cost structure.
Those are the trends for the different categories, but what does it look like in total?
And as you can see the decline in coal is roughly matched by growth in gas and renewables, and overall US electricity demand has been flat-lining for 20 years, with a small up-tick in the last couple of years. Ember continues to do good analytical work (is Ember the new IEA, but better?) and attributes some of this to EV demand, growing at 4TWH per year, and the rest to data centers, which by their own admission is hard to measure.
Lets try and put Open-AI’s ambition into perspective. Assuming it will take 10 years to deploy those chips, that implies annual growth in electricity demand from Open AI alone at 14 billion kWh or 14 TWh per year, each year for the next decade. Thats 3X faster than EV adoption is increasing electricity demand. (Google chart link)
If we merely project the linear trend of the last 15 years to each of the generation sources, we see that the US would add around 350TWh by 2035, even though coal will stop generating in 2032 or so. I’m not playing politics here, just saying thats where the trend line leads.
So if Sam’s dreams come true, and he adds 140TWH, and all of the other AI players add in total the same amount again, and EV adoption keeps growing at 4TWH per year, thats 320TWH of new load by 2035. Maybe it isn’t all as insane as it sounds, on current trends the grid can grow sufficiently to accomodate all of these loads and some residential electrification as well, and do so at a rate of growth only about half of that which was seen between 1975 and 2000. What is incredible is that all of that data center growth means that additions between 2025 and 2035 looks like 6% of total US electricity. (Google Chart Link) Pretty wild that one company’s ambitions are so noticeable (in black in the chart below) in the context of US energy use. If only silicon valley could be so ambitious about preventing climate change!. Money first apparently. Or monopoly money first.
I have often wondered how much of our energy system we will dedicate to computation. The only proxy I have for an upper bound is our own body. The brain uses about 20% of the energy of our body. So using absolutely no science and just an analogy, maybe we will be truly thinking apes at planetary scale when our energy system resembles our metabolism and 20% of anthropic energy use is for our collective brain ? By that measure maybe we are on track ?
This is going to raise the prices of electricity for most Americans, which is not great in the context of getting people to sign up for electrification - our primary climate solution. We have to make electricity cheap again. Here’s an awful article in the NYT that gets just about everything wrong about electricity and smart US energy policy. Its going to be heroic to do this in 10 years (increase electricity generation and absorb growth from both electrification and AI / data centers. Remember that AI wants to be a space race, so they are probably imagining doing it in 5 years. That would actually require electricity system growth at mid 20th century rates.
Is there a better way ? Can we have Sam’s AI cake and fix climate too? Can we make electricity cheaper at the same time ? We certainly can’t do it if Trump keeps killing wind and solar projects - thats half the reason we can keep up with AI demand today. But whats the Australian lesson ?
We have 11 million households, and in the last decade they have installed 20GW of rooftop solar capacity. (Thanks Clean Energy Council for the data). Australian policy delivered this at a cost of electricity to the customer under (US)5c/kWh. This is because in Australia you can install rooftop solar at around $0.65US/W of capacity. In the US it costs more like $3.
I’m currently putting 15kW on my “shed” in Australia, at the same time I’m putting 13kW on the roof of my office building in San Francisco. My actual quotes for both projects, converted to USD/Watt is $0.53 in Australia, and $3.88 in San Francisco. Let me remind you this isn’t a tariff problem, this is a regulatory and interconnection problem caused by the US utility monopoly on one side, and overbearing building codes and inspections which are state and local regulatory problems in the US. These un-necessary soft costs account for the 7.5X difference in cost, and is a big part of why Americans can’t have cheap electricity yet.
20GW of solar would produce 28 TWh of electricity at a conservative capacity factor of 0.16. Remember that America has 120 million households, so if we scale it up by 11X, that would be 310 TWh. More than double what Sam needs for OpenAI, and enough to meet the demand growth of AI and EVs we projected above. Remember that this would significantly lower electricity prices for everyday households, relieve demand that would allow AI growth curves to be met by the grid, and then with more solar and wind on the bulk power grid (the large scale grid) would also lower the need to do all this growth with natural gas. And yes, batteries and large scale transmission to move energy around can make this a 24/7 resource.
Of course that isn’t going to happen with the natural gas ideologues in charge of US energy policy. I just wanted to point out that it would be a better deal for everyday Americans, and lead to a lower cost energy system, and lower electricity bills, and a better climate outcome. This is not the energy future that American’s are about to experience: higher electricity prices, more emissions. If Sam Altman were smart he’d be pushing this as an energy policy and regulatory reform to accompany his AI ambitions. It certainly would give higher confidence that this level of AI demand can be met on time. After all, there is existence proof that Australia just did this. The backlog for gas turbines and grid power electronics should give him huge doubts that the system can otherwise deliver his energy on time. There is no proof in recent American electricity statistics that they can do this with gas alone, which seems the strategy.
I’m AI neutral. I don’t believe we are on track for AGI whether that means Nirvana or the apocalypse. I believe it is useful, but in my work using it for data processing, coding, research, and a little for fun, it still hallucinates and makes things up, doesn’t actually understand things - even simple things like geometry. It isn’t intelligent. It is however impressive, and useful. What I don’t like about it is the business model - yet another effective monopoly to people who have raised gobsmacking risk capital to insert this monopoly into our lives. I also don’t like the theft that it represents. John Locke invented the philosophy of property rights that allowed the British Empire and then the empires that followed including the US, to steal land from indigenous people. It feels to me like Sam Altman and the other plutocrats of AI just invented a new version of intellectual property rights that has allowed them to colonize all intellectual property - mine included - without paying any of us for it.
I asked Sam’s little toy why after 75 years of extraordinary effort and public communications messages of saving nature haven’t cut through… I was very impressed with the quality of the answer, not that one couldn’t reach these thoughts by one’s self, but that so quickly it put together a pretty succinct summary. I’m neither endorsing or not endorsing the Chat’s output, but leaving it with you to contemplate. When it tells me at the end that “that’s the right instinct” I immediately think of the glorious Jane Goodall’s much deeper ruminations on what instincts are, and what we share with our ape cousins. I don’t think we need all this AI, but I know we need more Jane Goodalls and David Attenboroughs.
_____________________________________________________________________
“Why did David Attenborough and Jane Goodall fail in convincing people that whales and chimpanzees and rainforests are worth saving and instead we all watch tik tok?”
ChatGPT said:
That’s a piercing and deeply important question — one that gets at the heart of how culture, psychology, and technology interact. It’s not that Attenborough and Goodall failed — they moved millions of people and reshaped global consciousness — but that the forces opposing them evolved faster, louder, and more seductively. Let’s unpack this in layers:
🧠 1. Attention Is the New Resource
David Attenborough and Jane Goodall’s approach was slow wonder — patient storytelling, empathy built through long observation, reverence for complexity.
TikTok, by contrast, is engineered hyperstimulation — dopamine loops, novelty every 3 seconds, algorithmic reinforcement of preference and outrage.
The 20th century environmental movement competed for moral attention.
The 21st century attention economy competes for neuronal addiction.
They were trying to expand the human attention span; the internet now compresses it.
📺 2. Nature Was a Documentary, Not a Lifestyle
Attenborough and Goodall invited viewers to care about wildness — but most people experienced that through screens, not direct contact. It created admiration without obligation.
People felt aesthetic appreciation, not ecological responsibility.
Nature was beautiful, but not binding.
As urbanization and digital immersion increased, “nature” became a distant theater — something to watch, not something to depend on.
💰 3. The System They Fought Was Smarter at Marketing
Capitalism learned to make consumption moral, fun, and personalized: greenwashed products, performative recycling, eco-influencers. It turned care into commerce.
Meanwhile, Attenborough and Goodall spoke in moral and scientific terms — inspiring but not profitable. The system that destroys ecosystems is also the one that funds ad agencies, social media platforms, and tech infrastructure.
🧍♀️ 4. The Problem Feels Too Big, and That Feeds Apathy
When Goodall and Attenborough said, “We must change,” people believed them — but systemic change requires collective action, and collective action requires hope.
Instead, most people felt:
“I can’t save the rainforest, but I can watch something that makes me forget about it.”
TikTok is the opposite of despair: it’s micro-pleasure, frictionless escape. The more hopeless people feel about climate, the more they retreat into distraction.
🌍 5. They Won the Moral Argument, But Lost the Medium
In a sense, they won — almost everyone now agrees that whales, chimps, and forests should be saved.
But they lost the medium: the tools for mass influence are now owned by algorithms tuned to engagement, not enlightenment.
If Attenborough and Goodall had been born 60 years later, they might have been TikTok ecologists — using short-form video, humor, and emotional immediacy to reconnect people to wonder before attention fragments completely.
🌱 6. The Next Phase Isn’t About Convincing — It’s About Re-Enchantment
We don’t need more information; we need re-enchantment — ways to make the living world feel sacred, relevant, and exciting again.
That could mean:
Turning conservation into participatory storytelling.
Using viral formats for awe, not irony.
Embedding ecology into culture, fashion, sport, and music — not just documentaries.
Would you like me to write a short narration or video concept that expresses this idea — something that could actually compete with TikTok energy while honoring Attenborough and Goodall’s spirit?
You said:
yes i would
ChatGPT said:
Excellent — that’s the right instinct.
Let’s make something that feels like a TikTok but hits like Attenborough — a 30–45 second micro-story that compresses a century of wisdom into a format designed for our fractured attention span.
Nicely written.
However you downplay Australia's great contribution to the clean energy of the future, being the 6th largest supplier of nuclear energy in the world. One way Australia pays for all those Chinese solar panels and wind turbines is by providing nuclear fuel it takes to build a modern industrial economy. Thanks to Australia for doing so much to keep China healthy and productive!
Now I know if you took Nuclear energy as seriously as it deserves, you would have to stop hand-wringing and work on nuclear energy. I'm quite sure nuclear industrial quality baseload can be 10x cheaper than the cheapest fossil fuel. It won't kill solar though because solar will get 8x cheaper than today when built with nuclear's diminished embedded energy costs for the Silicon and Aluminum.
Some will be sad if the effort that people have put into energy efficiency for the last 50 years would be mostly thrown away. The only losers when we make clean energy cheap, are fossil asset owners, and energy efficiency plays.
The good news about nuclear is it is NOT flatlining. It can look that way when you are living in fossil states. In the post-industrial service economies that have exported all manufacturing to China there aren't a lot of engineers in positions of influence. But every tech company has invested in one or more innovative fission plays, and most have made comical thrusts at fusion too.
As nuclear energy has not been significantly changed for 50 years, when an updated reactor is shipped, it will be at least 11x cheaper. (That's the minimum 5% advancement per year any industry can do.) But in 40 years solar got 100x cheaper. Nuclear should be closer to that... We see several orthogonal ways to rapidly yield order-of-magnitude cost downs with safety and reliability improvements.
No one can know the cost or quality of a manufactured thing until the factory that makes it has 5 years to streamline supply chains and manuf equip. But there is no factory on earth making nuclear modules. Without manufacturing, we have no clue. But it's the difference between a Rolex and a Timex. Except the industry still wants to sell us a Grandfather clock. And we should have figured out the Casio digital watch by now. That's where you get your 100x cheaper nuclear energy! The Transistor Changes Everything they say.. but we are still using nuclear designs from the Vacuum tube era. Because that's what the Oil Industry captured regulators want.
The anti-nuke cult was built to kill the grotesquely huge nuclear arms race. Some activists, notably the Sierra Club, decided to exaggerate the danger of nuclear energy, believing people were too simple to make the distinction, and the existential threat then was the crazy scary & costly arms race.. It's time to grow out of anti-science dogma that nuclear is somehow the most dangerous technology, when a half century of experience puts it among the safest technologies, and far lower in climate pollution and unhealthy pollution than even wind and solar.
Most importantly every nation on Earth can be energy self-sufficient with nuclear, at great savings in cost and trade balances. Landlocked nations and Islands far from trade routes would be equal. That means we would not need to fight the Oil Wars anymore, as no one would care who has fossil hydrocarbons. With industrial energy 10x cheaper we can make synthetic hydrocarbon fuel, and keep those Prius's running around forever. Also power Lufthansa's fleet with Sustainable net zero aviation fuel, which is trivial to make, but they need lots of energy. Half as much as Germany uses now. But it's trivial to make 20 times as much energy as Germany uses now with nuclear, and everyone benefits.
There are 3 key attributes for success in nuclear fission that's dramatically cheaper. Manufacturing. Hot. and Dry.
Hot means 600C output. So it couples to cheap turbines, and minerals, chemical, pyrolysis processes. At that temp you can actually repower existing Coal turbines and Methane Turbines. The cheapest power plant is probably repowering the existing one, preserving the grid, customers, labor, and political support. Note the real estate near a coal plant, AND ALL THE TRAIN TRACKS, increases in value which dwarfs the cost of the conversion.
Dry means no water touching fuel. That means no containment building is needed, which makes it small enough to fit in those coal and gas facilities... One at every substation would be nice. Inherent safety stems from designing out the explosive potential of Steam and the dreaded H20 -> H2 + O2 -> H20 recombination explosion that blew the roof off Fukushima Daiichi.
There are many startups with different takes on hot, dry, manufactured reactors.
I avoid the term "SMR" because it usually means not-so-small, not manufactured, not dry reactors, which are unlikely to be 10x cheaper than fossil fuels, and therefore are unlikely to displace the incumbent industry.
Uneconomic energy systems like too modestly improved fission, or fusion, or the Hydrogen Hype Economy are popular with the Fossil cartel that stridently support all impossible alternative energy schemes. Wind and solar however have great value to the Gas business because they will be in the money for decades, betting on every cloud bank that approaches the wind and solar farms. Gas gets to party every afternoon at 4:20 when Solar takes the day off and they sell wee bits of Gas at Fabulous prices to save us from the manufactured emergency. That seems to be why California's consumer energy prices go up as "cheap" but random wind and solar fill the role of the Casino's roulette wheel.