California Dreaming (And Hallucinations)
California proves the energy transition is complicated. AI and media hallucinate a different reality.
It’s Fourth of July weekend, beginning of summer…indulge me. Let’s ignore the messy politics around the world, and the rather inconclusive US inflation and employment data. I had promised myself I would not write about energy again for a while, but then I came across this article from The New Yorker, titled, “California Is Showing A Big State Can Power Itself Without Fossil Fuels.”
This is exciting, I thought. California can now power itself fully with cheap and clean renewable energy? Maybe I should move back! (I left San Francisco for Miami four years ago). Then I noticed the subtitle:
For part of almost every day this spring, the state produced more electricity than it needed from renewable sources.
Hmmm. “For PART of ALMOST every day THIS SPRING”…? That’s three qualifiers in just eight words. Not quite the same thing as “can power itself without fossil fuels.” Hey, I can go without eating for part of almost every day of every year (thankfully the human body has developed a sophisticated and efficient energy storage system).
“We’re making history”
But, credit where it’s due, California has certainly brought a lot of renewable power on line. So my reaction was, guys, tone down the hype, it just undermines credibility, and keep up the good progress.
But by now you know my weakness: I can never stop at the headline, I always surrender to the temptation to read more, look for some numbers. So I wondered, how fast is California weening itself off fossil fuels?
Source: California Energy Commission
Oops. The share of fossil fuels in California’s electricity generation last year was exactly the same as in 2017 (44%). A net improvement from the 60%+ peak of 2012-15, but absolutely no progress in the last seven years. Given that California aims to virtually eliminate fossil fuel use by 2045 (“We’re making history here in California,” said Governor Newsom), this track record does not exactly inspire confidence. Though admittedly, by definition, making history takes time—Rome was not built in a day, etc. It’s complicated.
Energy hallucinations
Then I wondered, how is progress going in the US as a whole? But it’s summer, I’m feeling lazy, and I didn’t really feel like looking for more numbers, calculating percentages, drawing graphs… I almost gave up. Until I thought, wait, I have access to a superb, productivity-enhancing research assistant: AI. I fired up ChatGPT and asked: “please give me a time series of data on the shares of fossil fuels and renewables in US electricity generation for the past 20 years.” Et voilà:
Here is a time series plot showing the shares of fossil fuels and renewables in United States electricity generation over the past 20 years. The data illustrates a steady decline in the share of fossil fuels and a corresponding rise in the share of renewables.
- Fossil Fuels Share: Decreased from 70% in 2004 to 42% in 2023.
- Renewables Share: Increased from 2% in 2004 to 30% in 2023.
Thus spoke ChatGPT. Fantastic, I thought. Superb research assistance, and wow, that’s a very impressive reduction in fossil fuel reliance, from 70% to 40% in less than twenty years. Way better even than California.
But then, my weakness again…against my better instincts I asked: “what’s the source?” And my top-notch assistant replied:
The data I used for the plot is simulated to illustrate the trends. For accurate and real-time data, you can refer to reliable sources such as the U.S. Energy Information Administration (EIA). The EIA provides detailed statistics and reports on the shares of fossil fuels and renewables in U.S. electricity generation.
Hold on a sec. “The data is simulated to illustrate the trends”. Are you saying you made up the numbers???
Would you like me to look up the most recent data for you?
Yes please
The U.S. Energy Information Administration (EIA) provides comprehensive data on the shares of fossil fuels and renewables in U.S. electricity generation. According to their reports, there has been a significant shift in the energy mix over the past two decades.
From 2004 to 2023:
- The share of electricity generated from fossil fuels (coal, natural gas, and petroleum) has decreased from around 70% to approximately 60%. This decline is largely due to the reduction in coal usage, while natural gas has maintained a significant portion.
- The share of renewables (wind, solar, hydro, and biomass) has grown considerably. In 2004, renewables contributed only about 8% to the total electricity generation. By 2023, this share increased to over 22%, driven mainly by the rapid expansion of wind and solar power
Oh, I see. So in fact the share of fossil fuels declined by 10 percentage points, not 30; and the share of renewables increased by 14 points, not 28.
Source: US Energy Information Administration; ChatGPT’s own hallucinations.
Much like The New Yorker, ChatGPT hallucinated. It’s a bit depressing—wherever you turn, people and machines now are just making stuff up.
Resigned to doing the research myself, I pulled up the data, and here we are: progress in reducing the share of fossil fuels in US electricity generation is ongoing, but rather slow. From about 80% in the 1960s and early ’70s we are down to 60%. We’ve achieved a first 10-point reduction between the mid-1970s and the late 1990s, and a further 10-point reduction between 2010 and 2020, but progress seems to have stalled in the last four years.
Source: US Energy Information Administration
Much like The New Yorker, ChatGPT hallucinated. It’s a bit depressing—wherever you turn, people and machines are now just making stuff up.
It ain’t cheap yet
Tired but with a better sense of the numbers, I realized I had not yet determined whether I should be moving back to California. There has been no progress in reducing the reliance on fossil fuels in the last seven years, but still, California is getting half its electricity from renewables. Electricity must be cheap, no?
No, no, don’t look at the numbers, just ask ChatGPT!….
But I can’t help myself, and my confidence in my AI research assistant wasn’t quite the same anymore, so I did go look at the numbers:
Source: US Energy Information Administration
Ouch. California has the second-highest residential electricity prices in the country, and double the US average. Second-highest also for commercial use (a bit less than twice the national average) and fourth-highest for industrial use (over twice the national average). I’ll spare you the identical-looking charts.
I guess I won’t be moving back to California just yet.
I do look forward to a future of abundant, clean and cheap energy. Knowledgeable friends assure me the acceleration is just around the corner. Keep the faith, and keep up the good work—but beware hype and hallucinations.
P.S.: Avoidance
I noticed the New Yorker article because it had been shared on LinkedIn by a California-based tech company, and a friend (a knowledgeable optimist about the green transition) commented on it and tagged me to invite my reaction. So I posted the first two points above in the comments (the difference between the New Yorker headline and reality, and the lack of progress in the past 7 years). The comments have since been deleted. The tech sector, after all, has a well-know aversion to open debate on certain issues… And when hallucinations are so much better than reality, who can blame them?
Ciao Marco, and thanks again for sharing your process thought in analyzing data trend. we (coming from the oil & gas industry) can appreciate the high level of complexity in the value production of transforming useful energy and im always dubbed as the dirty guy when i attempt to explain the leap is more than a simple step into a new era and sometimes i just give up. but where i never give up is regarding the usage of AI as your assistance, i published on LinkedIn the paper 'Do you believe your questions always have intelligent answers?' in which i faced your exact surprise of HAL 9000 sindrome... I truly believe that we can harness AI full trusted power once its feed with the data set you want to analyze not the stochastic vast data of the internet. My father has terminal cancer and we are submerged with documents with diagnosis and recommendation that is truly difficult to extract what has to be done, once AI feed with all the 'cartelle cliniche' it summarize what, who, when, how. i will never ask for a medical diagnosis! maybe at school the new topic is how to use AI. take care. paolo.
Thanks Marco, much appreciated, I read from your blog many thoughts i share from Musk biography that i just downloaded to energy and AI to a bit lesser GE old days glory, again i loved the old one GE and its powerhouse it was amazing. I believe in rebirth or second chances for the new GE as phoenix from the ashes. thanks. stay in touch. p