Can someone please explain like I'm five how or whether the energy needs for technical civilization in the future can possibly be met?
So from earliest childhood, I have experienced that from time to time the electrical grid becomes unavailable for use and it can take days or even weeks to restore service. I'm having trouble comprehending the scope, scale and plausibility of what changes would need to be made to increase the electrification of everything in the way that is being pushed by policy advisors.
Everyone is pushing electric cars. I think it's a great idea, but I have questions about how the grid can support it.
People tell me that the next big advancement in the workplace is going to be the incorporation of artificial intelligence. Doesn't AI require servers on a massive scale? How plausible is it for AI to reach all corners of society and economy on our existing grid or reasonable expectations for plausible improvement of the grid?
The banks seem to be lobbying for the substitution of electronic accounts for cash. Again, electric power is not always available. Also some people who need to use money don't have homes and can't reliably charge electronics. If I remember correctly the payment system went down in Canada a while ago and people without cash were out of luck.
What insight can you share with me?
Can't really say I share that experience, I think the last time I experienced a long term power outage was when PG&E had rolling blackouts because they were setting the entirety of northern california on fire, but even that ended up being about half a day.
I think a major issue with your hypotheticals is that transitions don't need to be binary. When an EV transition occurs, it's not as if one night everyone's ICEs are replaced with electric motors. It's a decades long process of having more and more new car sales be EVs.
So yes, in most of the world, the grid isn't really set up for the scenario where everyone has an EV, moreso for rural areas. And that's OK, people will still have ICEs, especially where power generation is expensive or power delivery is unreliable. It's a process, not a flip. As more people get EVs, more people will use electricity, and if there isn't more suppliers of electricity, electricity prices will spike. That will discourage more people from buying EVs until more supply is brought.
In terms of supply, renewables, in particular solar, combined with methods to store solar, is a good one for much of the world. Nuclear seems fairly dead end-ish politically in the west, even in Europe, but the majority of the population lives in Asian continent, and in many countries there nuclear is still going strong.
In terms of "AI", it's not really on the same scale. For one, the inevitable future of this matrix smashing is that it gets ever more optimized. Already, you can run LLMs on edge hardware - someone even got LLaMA to run on raspberry pi, albeit slowly.
Although the big models running on big GPUs are expensive, the scale of that energy expenditures compared to, say, cars isn't really close either. It also has the advantage of scale. The datacenters can be in areas with cheap electricity.
Furthermore, I'm not really sure those are comparable. Again, """AI"""'s usage will be a gradual increase, and I'm not sure anyone says it's a necessity like transportation.
Yeah, last time I experienced a prolonged (multi-day) outage was 20 years ago:
https://en.wikipedia.org/wiki/Northeast_blackout_of_2003
And again (to a lesser extend) in 2013:
https://en.wikipedia.org/wiki/December_2013_North_American_storm_complex#Ontario
I live in a forested, somewhat rural area now, so we get outages a few times a year due to tree/branch falls severing power lines, but they have never lasted for more than a few hours. So none of those outages, or the previous ones I mentioned, have been due to energy demand exceeding capacity.
I live in the Seattle area where our power lines go under many tall trees before reaching our neighborhood. Unfortunately there are strong windstorms a few times a year that cause widespread outages across the region so repairs are not quick. Our little section only has a couple hundred homes so it gets prioritized near last. In the past few years we have had two five day outages. This is in a suburban neighborhood, not a rural area. But tall trees grow like weeds here and shed lots of material for taking down lines. This year I purchased a portable generator so at least I don't have to throw out everything in my refrigerator when this happens. By chance, we actually had a four hour outage this morning for some unknown reason.
Oof, that sounds like a nightmare. Even being without power for a few hours a few times a year drives me absolutely nuts. So I can't imagine being without power for days at time, semi-regularly like that. I would probably lose my mind. :P
I have a battery backup for my computer, just because we do get those short outages quite a few times a year, and I don't like my computer suddenly turning off... but we thankfully haven't felt the need to buy a generator. I can totally see how it would be almost a necessity in your situation though.
Is there a reason that they don't run on CPUs? Is there a huge need for graphics, or is it just that GPUs are better for single tasks?
Most machine learning algorithms are heavily based on matrix math, specifically matrix multiplications. It just so happens that computer graphics also heavily use matrix multiplications, and so GPUs have been optimized and expanded to be incredibly good at doing massive amounts of matrix operations in parallel.
The best analogy I was ever given for this was to think of CPUs as university maths students and GPUs as primary school students.
CPUs are great at doing a few, long, complex maths calculations that require imaginary numbers, similar to a few high university students working on the equation together. Whereas a GPU is better at doing lots of short, simple, matrix maths calculations, like a primary school classroom working through a text book full of similar equations.
Cool! Thanks.
Could you say where you read about that? Texas has had some severe outages during winter and I have doubts it will be any better next time. But I was under the impression that they're doing okay this year.
Example: Solar, wind energy keeping Texas power grids running amid weekslong heat wave
I just learned the other day that Texas is the only state in the Lower 48 that does not have any interstate power-sharing capabilities at all. They intentionally isolated the state's power grid from their neighbors for political reasons, so they do not have to deal with Federal regulations.
But that means that Texas can never buy/sell borrow/lend power to/from its neighbor states. Just Texas. Every other state can and does regularly share power to balance demand, help with crises, etc. If Texas' grid crashes, their neighbors literally cannot help them.
This is not true. Most of Texas is on its own grid, the Texas Interconnection, but there are two DC ties from that grid to the Eastern Interconnection. This is not anywhere near enough to supply all of Texas, of course.
https://en.wikipedia.org/wiki/Texas_Interconnection
"The Texas Interconnection is maintained as a separate grid for political, rather than technical reasons, but can also draw some power from other grids using DC ties. By not crossing state lines, the synchronous power grid is in most respects not subject to federal (Federal Energy Regulatory Commission) regulation."
Yes, Texas has a couple of connections to the Eastern grid (and also, apparently, to Mexico grid?) ... but it is still designed and maintained as its own, separate electrical grid.
Yes, but your claim was that the state "does not have any interstate power-sharing capabilities at all". That isn't true. It runs on a separate grid, but it isn't totally disconnected from the rest of the country.
Yep, that bit them pretty hard a couple of years ago. That outage, incidentally, was one of several causes of the chip shortage.
I just realized that the last time I remember the electrical grid going down was more than fifteen years ago, and it was a matter of minutes. So just speaking to your first point, the reliability of the electric grid varies drastically depending on where you are.
Electric cars and their charging - there's multiple ways how that can be handled. Right now, as we are still in the early stages of electrification of personal transportation, cars just plug in and use existing grid reserves. What the plan is for the future depends on where you are. The strategies range from grid extension, heavily subsiding local generation (and potentially storage), peak load shifting (basically loading cars at night, industry consumers for example often pay less for consuming off peak) to offering charging stations that are in grid-convenient places.
And with regards to AI tools - you're right that simulating large neural networks, the heart of what people call AI today, is energy intensive if you compare it to simpler programs. But compared to having one office worker more that drives to work and uses the lift and wants air conditioning and a PC etc it really is not a big amount of energy. You might compare it to building roads - it's an energy and labor intensive process, but it makes transportation much more efficient.
Of course all of this has downsides, too, like the virtualization of currency. Most companies today are just as dependent on the Internet as they are on the electrical grid. Many shops are dead in the water if their online services are down because their registers stop working, for example, so having cash night not help you in that situation.
This enbrittlement of civilization has been going for centuries now (what is a city without food deliveries?) and is the flip side of infrastructure and increased productivity.
This reminds me about a conversation I had with a friend some years ago regarding streaming services. He was convinced it won't catch on because so many people have spotty internet connection, data caps and so on. Nothing beats having all your music on a hard drive! And well, look what happened. Similar arguments are made today about game streaming.
The reality is that technological development doesn't wait for everyone to be 100% covered or for nobody to ever be inconvenienced. Maybe you live in a place where electricity goes out for days, but most of us don't.
I think technology access is a new kind of river/coastline access for people. For example which ISPs were available and what speeds they offered has factored into my home buying decisions in the past. I had a friend completely walk away from a home buying deal because the internet speeds they'd be able to get were unacceptable.
I'll point out a different take: We will eventually hit up against physical limitations of our energy use.
There's going to be a reckoning, if not in my lifetime, certainly my kid's.
Am I a freak for finding immense comfort from this post? That, within a very short timeframe we're going to run up against a palpable brick wall?
Steady economic growth is all we've ever really known. We need to get off this bombed strapped bus, like, yesterday, but actually what alternatives are there?
The energy needs for our technological society will easily be met, especially after a few more decades of science advancing. There is a huge amount of energy that comes to Earth (in the form of light from the sun, which we harvest using solar panels) and is generated by the earth (wind, power from dams, from geothermal wells, from the tides of the ocean etc). The amount of energy we could harvest is WAY WAY more than we use right now.
The thing is, we didn't know how to harvest much of this energy in the past. Many hundreds of years ago we figured out how to get energy from water flowing downhill, using things like water mills which could grind wheat into flour. Primitive windmills were being used a couple of hundred years ago as well.
We need to use MUCH more energy than we used to at this point, but we now know a bunch about how to harvest it. We are right in the middle of the period of time where we are developing this technology. It was primitive a hundred years ago, and a hundred years from now it will be very easy for us. We are also getting really good at storing energy in batteries and other stuff, and that will also get WAY better over the next few decades.
The problems we are going to face is mostly other stuff. For example, wars, climate disaster, and evil rich people trying to stop everyone from switching from burning fossils fuels to renewable energy.
But overall, the good news is that we already know how to get the energy. It's become very cheap to harvest, just in the last couple of decades. If we eventually need 100 times as much energy as we currently need, that will still be something we can do, as we keep improving tech.
I'll let other people address the other stuff you mentioned, but the energy issue is mostly solved in theory already. We mainly need to just build a whole bunch of energy harvesting stuff, the storage to store it so that things like electricity going out becomes super rare, and the stuff it takes to move the energy around, so that everyone's home and work place etc all have as much power as they need.
I would also assume the needs are going to go down for a lot of things.
If we look at trends, many people switched from incandescent to LED. My phone is more powerful than my PC of 15 years ago, running from a small battery. A low wattage Raspberry Pi is as powerful as my desktop from 10 years ago.
Heat pump usages looks to be increasing, taking less power for heating and cooling. California is seeing most of its mid day needs met by solar (https://www.eia.gov/todayinenergy/detail.php?id=56880), now if that curve deepens into a surplus, we just need to learn how to store it and we have power to put into vehicles.
Technology Connections has an entire series on heat pumps - I believe this is the first one. They really feel criminally underused in much of the US right now, I'm most certainly planning on getting one next time my house needs a replacement air handler. The only question for me is geothermal heat pump or standard one.
And is scarcity of rare earth metals a concern? Just something I've heard about but not in any position to evaluate it. Thanks
Tesla uses nickel, cobalt, aluminum, manganese, and lithium for batteries (a mix and match, depending on which battery).
As @A1sound pointed out, they're called rare earth metals for a reason.
There is some hoopla currently going on regarding a lithium mine on federal lands in ID, where Native Americans still aren't exactly happy (https://www.npr.org/2023/06/28/1184812267/western-tribes-last-ditch-effort-to-stall-a-large-lithium-mine-in-nevada).
There are also concepts for deep-sea mining for some other minerals, which do not really sound great for the earth. (https://www.dw.com/en/underwater-mining-to-extract-lithium-cobalt-threatens-biodiversity/a-65219511. Note this article points out: According to estimates, in just three years the world will need twice as much lithium and 70% more cobalt. And this is despite the slow progress of the energy transition. According to the International Energy Agency, if climate goals were properly pursued through the massive expansion of renewable energy, about five times as much lithium and four times as much cobalt would be needed by 2030. But then also this: https://www.economist.com/science-and-technology/2023/07/02/deep-sea-mining-may-soon-ease-the-worlds-battery-metal-shortage)
Another article on the problem of mining nickel: https://www.bizjournals.com/atlanta/news/2023/06/05/ev-batteries-nickel-mining-environment.html.
So yes, it's not necessarily "better" than what we're doing for pulling out petroleum (oil & gas); and additionally, I've yet to see a cost-and-material-efficient way to recycle these batteries we're creating. And please, feel free to show me if you know of some good proof!
(I originally meant to chime in on where this power will be coming from for these EVs, as I work in the power generation industry, but I find the EV situation more dire at the moment.)
Not an expert, but I can only imagine that they're called 'rare' earth metals for a reason.
You're right, there is a reason, but thankfully, it's not what those unfamiliar with them would expect:
Rare earths: Neither rare, nor earths
For the lazy:
Thank you, that helps me understand better.
Much would need to be done to get (a) commercial real estate and (b) privately owned utility companies to actually "do good" with this, but it's absolutely feasible.
Let me take California as an example. In 2022, California hit a high with 103% of its electricity needs being generated by renewable energy sources. It only lasted for a few hours, but that's still a huge achievement when you consider how many people live in California and what the demands of its combined grid must be. Beyond that, California has frequently hit a large portion of its grid needs from renewables. Their energy commission stated that in 2020, almost 60% of the grid was using renewable and zero-carbon energy sources. They estimate that it should be possible to hit 100% by 2045.
However, there are some big caveats. PG&E, one of the utility companies in California that provides gas and electricity, is responsible for a number of wildfires due to their lack of maintenance on existing infrastructure. So even though it should be possible to hook up renewables to the existing grid, getting energy to people is an entirely different question, and that problem is stepped in greed. We just need to hold utility companies accountable and have them maintain, upgrade, and expand their infrastructure to ensure everyone is benefiting from a modern grid.
Electricity reliability tends to be worse in rural areas, but that's also a good place to install solar panels and batteries to last through outages. Perhaps there won't be that much more demand on the grid in such places? It would increase resilience, too.
It doesn't mean cheap electricity, though. For example, Hawaii has the highest electricity prices in the US, and that's why a lot of people are installing solar panels.
Explain like you're five? Easy:
"Go ask your mother."