Oh, the unintended consequences!
Read or skim this, then let's play it forward.
The AI Revolution Has a Problem: Energy
ByIrina Slav- Apr 10, 2024, 6:00 PM CDT
EQT CEO Rice: growth in AI use would require an expansion of gas generation capacity.
On the one hand, artificial intelligence promises to help advance the transition on the other, it is clearly hindering the progress of that transition by posing an energy demand question that only hydrocarbons can answer.
Dutch Scientist finds that global electricity use of AI globally could exceed 85 TWh annually.
Artificial intelligence is helping the oil and gas industry boost output by accelerating drilling and boosting efficiency. It offers similar gains to other industries as well and online platforms and websites with AI-powered chatbots are multiplying fast. But there is a problem. AI consumes massive amounts of electricity.
A New Yorker article from last month cited the number half a million kilowatt-hours. Per day. That's the electricity consumption of ChatGPT to handle the two hundred million requests it gets daily. That's a lot of electricity. And that's just one AI program.
According to one Dutch scientist who has calculated the potential electricity use of AI technology globally, it could reach a staggering 85 terrawatt-hours. Annually. And that's the lower end of the range. The higher end is 134 terrawatt-hours or 134 billion kilowatt-hours.
"You would be talking about the size of a country like the Netherlands in terms of electricity consumption. You're talking about half a per cent of our total global electricity consumption," Alex De Vries told the BBC last year when a study he authored about the electricity appetite of AI hit headlines.
Skip Ad
Needless to say, this sort of additional demand cannot be satisfied by wind and solar, as De Vries himself acknowledges. "We need fusion or we need, like, radically cheaper solar plus storage, or something, at massive scale—like, a scale that no one is really planning for," De Vries told the New Yorker this March. Alternatively, gas consumption for electricity generation—and quite likely coal consumption—will rise to satisfy this demand.
Former U.S. Energy Secretary Ernest Moniz, a staunch supporter of the energy transition, put it even more succinctly: "We're not going to build 100 gigawatts of new renewables in a few years. You're kind of stuck," he said last month in comments on AI's energy needs, as quoted by the Wall Street Journal.Related: Top U.S. Utilities Prepare for AI-Driven Surge in Power Demand
It's all because of the enormous amounts of information AI programs need to process to perform a task. De Vries has calculated that if Google switches its search engine to generative AI, the electricity consumption of that search engine would shoot up to 29 TWh per year. And that's just Google, which started trialing AI-powered searches in the UK earlier this month.
"The more information they gather, the smarter they are, but the more information they gather to get smarter, the more power it takes." This is the description provided by the chief executive of Arm, a chip design company owned by Japan's SoftBank.
This amount of power, in total, could come to represent between a fifth and a quarter of total U.S. electricity demand by 2030, Rene Haas told the Wall Street Journal this week. "That's hardly very sustainable, to be honest with you."
Indeed, it is not exactly sustainable if you are planning on a shift to low-carbon energy generation that would inevitably necessitate a decline rather than an increase in energy demand. And this is exactly why the oil and gas industry is very much in board with AI, especially the gas part of it.
"Gas is the only cost-efficient energy generation capable of providing the type of 24/7 reliable power required by the big technology companies to power the AI boom," the founder and senior partner at Energy Capital Partners, an investment firm, told the FT earlier this month.
This is why growth in AI use would require an expansion of gas generation capacity, according to the company. The word "critical" was used. Or as EQT chief executive Toby Rice put it, echoing the message of Ernest Moniz, "It will not be done without gas."
All this has put transition advocates in a difficult position. On the one hand, artificial intelligence promises to help advance the transition, according to some of those advocates that work in the technology space. Yet on the other, it is clearly hindering the progress of that transition by posing an energy demand question that only hydrocarbons can answer.
The AI boom is also incompatible with the idea of degrowth that has been gaining traction in Europe. Degrowth denotes the idea that economies no longer need to aim growth at all costs. On the contrary, lower consumption is what economies should be striving for, notably including lower energy consumption. This won't work with AI data centers.
"It’s going to be difficult to accelerate the breakthroughs that we need if the power requirements for these large data centers for people to do research on keeps going up and up and up,” Arm’s Haas told the WSJ.
It is also going to be difficult for Big Tech to continue promoting the energy transition and advertising its power purchase agreements as evidence of their clean energy commitments being applied in practice. Growth versus degrowth is one thing. But what happens when the debate goes to technological progress versus the end of progress? Big Tech, this darling of green investors, might just turn into the new target of activists.
So let’s take the black-pilled approach: They decide to steal for themselves and their freakish AI Overlord all the energy necessary to run this technocratic tyranny, leaving us plebs to go feral. (Bring it on, you anti-human scumbags.)
If they commit mass murder in this fashion, who’s left to lord OVER? Part of what drives these maniacs (and by now, we know who I’m talking about*) is pure psychopathy. It’s beyond greed. Although I do believe fear plays a role.
AI is not our friend. Ever see the pile of shit that woke chatbots spew back to a query? As a copy writer, I find it offensive — and a little funny.
*Related: Check out this graphic by @fatemperor, Ivor Cummins. That’s his X account.
I work in AI. It's innocent.
AI is not sentient or conscious. It does not "think."
AI is trained on data and utilizes human made algorithms to process data until machine learning and self generated algorithms contribute to how it processes.
The flaw is in the first two parts.
All data is biased, at least when used in certain ways. Human created algorithms are biased, often unintentionally. The training data used to train AI that the public interacts with is unknown, especially in what is called a foundational model or foundational data set. Therefore, the biases are unknown until it rubs thinking people the wrong way.
The data selection/filtering and those human algorithms are what skews the AI, at least the majority of the time.
I created a business bot using a well-known LLM and I have it controlled to not delve into those rabbit holes.
We could create and train an anti-vaxx AI or an anti-psychopath AI and it would largely say what we want.
So the problems with AI relative to Ivor Cummins' graphic are largely from human interventions. The AI has been trained to be leftist and anti-human in many ways and that will only worsen and skew more if we put up with it.
Humans and psychopaths are the culprits. What a surprise. ;)
I have been randomly testing free AI-- specifically Deep Chat AI. Completely inconsistent. Some information is usable, other info is completely wrong. COMPLETELY and obviously wrong. For example, if you ask about something like resorts in a specific area, they will list properties in a different state.
As for copywriting... the marketing girl at my former job used to use it to write. Horrible. Just horrible. Strangely enough, out of five people, only 2 of us noticed the writing was bad. The other three-- one in her mid 20s, one in her 40's and one in his late 50's-- thought it was fine.
We're drifting into Idiocracy, and Our Betters are OK with things being less than mediocre, because "less than mediocre" means "no benefits or sick days or even a salary." So it's a win for them.