this post was submitted on 22 Jun 2024
54 points (90.9% liked)
Environment
285 readers
1 users here now
Your definitive source for news, information, issues and activism related to the environment.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I run AI locally on my computer.
I live in an area where power comes from hydro-electricity.
That isn't a solution, you're still putting unnecessary load on the grid
Its still better.
No it isn't, data centers and top of the line servers will do the same work more efficiently than a home setup, the issue is AI in general.
Decentralization is never more efficient from an energy usage perspective in the tech sector.
Give me a better solution -_-
Just... Don't use AI?
That's like a crypto miner saying their mining is ok and doesn't waste energy because it comes from hydro electricity, well, no, the issue is using energy for that in the first place.
No matter the source, you're putting a huge load on the grid compared to the output and that energy would be better spent elsewhere. Hydro is green in the sense that there's very little emissions while it's produced, getting to the point where you produce it releases a shit ton of emissions and the more people use it for useless shit, the more dams we need to build, the more land we need to flood, the more trees will decompose under water, the more damage we need to do to ecosystems, the bigger the impact on biodiversity.
Sure but what is a more worthless waste of time, mining crypto or running AI?
It's a hard choice.
In the grand scheme of things? Both current implementations are pretty much equal.
A load on the grid is not a problem. We need loads on the grid, or there is no point in even having or improving a grid.
Solar generation is so plentiful now that the market price for generation frequently goes negative: grid operators are charging generators for putting unneeded power on the grid. (The problem is that generators are limited in how quickly they can ramp up production, so they have to maintain a certain output even during surplus, or they won't be able to meet demand later.)
We don't need less demand. We need more demand, and more flexible demand. We need to shift away from "supply shaping" - matching the power we produce to what the consumer wants. We need to focus on "demand shaping": matching the amount of power we use to our ability to produce it. We need to look for ways to use the power we have, when we have it. We need ways to make that "excess" power profitable to produce.. Economic losses on solar production are beginning to strangle solar expansion. Why build a solar farm when the power it produces can't be profitably sold?
We also need to reduce our reliance on "peaker" plants. To do that, we need to artificially inflate the off-peak "base load" so that a higher percentage of our power comes from efficient baseload plants, and less of our total consumption comes from inefficient "peaker" plants. But, at the same time, we need to shift loads to times of day where solar and wind can meet them. We need to be able to better match demand to available supply.
The transportation industry is far worse for ecology. We would be far better off with transportation operating on the electric grid than the oil grid, but the electric grid cannot currently support the demand from both industries.
That load should come from providing clean energy to places that burn fossil fuel, not from creating artificial demand for useless stuff.
As I said, you're using the exact same arguments as crypto Bros.
I don't think you're actually understanding my argument at all.
Early morning and late evening, or under cloudy skies, a given solar array produces only a fraction of the power that it can at noon on a clear day. You might get 20 times as much power from that array at noon as under those degraded conditions.
So how big of an array do we build? Do we build a large array that can easily meet demand even under degraded conditions? Or do we build a small array that will rarely fully meet demands, but will rarely produce unsaleable excess power?
The supply shaping model we currently have uses that second strategy. In certain areas, we are already producing more solar on ideal days than needed to meet demand. Even with pumped storage facilities to time-shift demand, we occasionally have far "too much" solar. Increasing the size of the arrays further will decrease profitability by increasing the amount of negative-rate power we put on the grid.
The environment is far better served by the first strategy: building out so much solar that it still works in the early morning, late evening, and under overcast skies. And figuring out some way of using excess power when conditions aren't degraded.
The only way we can use that first strategy is if we can come up with a way to shape demand such that we can make use of that excess. So, we will need massive, energy-intensive projects that can be switched on and off depending on weather conditions. Desalination would work. We can use the massive excess power to recharge aquifers from the ocean anytime the skies are clear.
Hydrogen electrolysis would work. We can fill massive storage tanks with hydrogen any time we have excess sunlight, and shut down production when we don't.
Fischer-Tropsch synfuel production would work. Produce synthetic jet fuel from biomass.
Direct Carbon Sequestration would work. An energy intensive process to pull carbon directly out of the atmosphere.
We need beneficial ways of utilizing the excess power we get from solar arrays sized large enough to fully meet normal demand, and we need a grid capable of handling the excess production.
So exactly, it's moronic to use it for AI when we could store it or reduce our usage and have less needs or transport it over longer distances for it to be used where they rely on fossil fuel.
By the way, this conversation started with hydro so challenges are different and worst case just letting water through isn't a bad thing as there's ecological impact to just keeping the water in the reservoirs and letting the level rise while there's no water downstream.
Hydro is irrelevant. It has long since peaked. The water sources that can viably produce power are already utilized. Relatively little additional hydro capacity is feasible. Even if there were, hydro is easy to incorporate into any generation structure. It can be used as baseload, peaker, storage, etc. We don't need any special considerations to utilize hydro. There is little point in discussing hydro in this context.
Solar is underutilized. It will always be underutilized. There is more solar power available than humanity could ever hope to use. It is far more abundant than hydro.
Hydro will remain useful no matter what we do, but if we want to be able to effectively utilize solar (or wind), we need to focus on demand shaping. We need more load on the grid, not less. Without additional load, solar rollout will stagnate long before it successfully replaces traditional generation methods.
Like the other projects I mentioned, AI can create an intermittent demand for excess power produced under ideal conditions. The majority of AI power is consumed while learning, training the model. Responding to queries takes very little energy. AI data centers can burn power when it is readily available, and shut down under adverse conditions, conserving power for more essential services.
Without extra demand during optimal conditions, there is no economic incentive to build enough solar capacity to meet demand during suboptimal conditions. The greater the extra load we can add, the worse conditions have to be before solar cannot meet normal demand. We need much more "extra" demand for intermittent generation to economically justify additional solar and wind.
Find me one major AI or crypto company that will accept to shut down their infrastructure when there's high load from households and offices.
You guys are fucking delusional. There's so much efficiency to be gained by stopping all the energy waste, instead you're pushing for us to use more and more energy so we build more infrastructure as if there was no environmental impact to it. Until you're able to pull the material to produce the the panels from your ass, the most efficient and environmentally friendly thing to do it always to reduce first and foremost.
Offer them ridiculously cheap power under normal conditions, but price them high during shortages. They aren't "shutting down their infrastructure". They will still be able to handle requests. They just won't be incorporating new training data to their models until the sun comes back out.
There is much, much, much, much more energy feasibly available when we focus on demand shaping instead of traditional supply shaping models. Ever hear the phrase "penny wise, pound foolish"?
You are either unaware that we are already regularly experiencing "negative rates" or you are not considering the ramifications. You are not considering how drastically "negative rates" are already stunting solar and wind development.
What I am talking about is boosting intermittent demand so that rates don't go negative. We need ways to soak up every available watt when we have more than we would normally use.
Think of it this way: our current, "supply shaping" model requires extensive use of expensiv , inefficient "peaker" plants. Peakers give us the ability to match an unregulated demand with a variable supply.
A "demand shaping" model, if enacted effectively enough, eliminates the need for inefficient peaker plants, leaving us with a moderately efficient baseload plant for overnight, and extremely efficient solar and wind during the day.
Current peaks are higher than they need to be because people are wasting energy, all you want is to increase demand as if that had no environmental impact. I'm done here.
You do realize we are already using demand shaping, but for the traditional baseload/peaker model, right?
Power companies offer steep discounts to industries like aluminum smelters and iron foundries to move their production to a night shift. Doing this increases the base load, which allows a larger percentage of the total power demand to be met by baseload generators instead of peaker plants.
The problem with this should be obvious: demand shaping on the baseload/peaker model drives demand to hours of the day that solar and wind cannot possibly meet.
Current peaks are not nearly as high as they should be. As much night-time demand as possible should be moved to daytime, where it can be met with solar. We need peak demand to correspond to peak generation. We can't move peak solar production; the trick is to shift our demand curve to match the solar production curve. Both peaks need to occur simultaneously.
You're still not comprehending. I'm meeting a far higher percentage of our current energy consumption with solar instead of dirtier and less efficient coal/oil/nuclear baseload generation, pumped storage, battery storage, etc. The "extra" demand I am asking for is 100% met by the excess capacity of our solar generation that arises under optimal conditions. That excess capacity is baked in. When we have enough solar capacity to meet normal demand with overcast skies, optimal conditions will give us massive surpluses.
Because it is met entirely with excess capacity that would otherwise be wasted, the "extra" demand I am calling for has zero additional environmental impact. It monetizes excess capacity that we wouldn't normally be able to utilize.
Give me the benefit of a doubt for a moment, and actually consider what I am saying. Yes, it sounds paradoxical at first glance, but it will make a lot more sense when you realize I'm talking about how to move the overwhelming majority of our electrical production to solar/wind and virtually eliminating peaker plants.
Hey buddy, no one serious thinks the way you do, the industry is using more fossil fuel to meet the increased demand even if there's subsidies for renewable. You're just recycling crypto bros arguments to justify wasting energy.
I can't believe I'm still answering that bullshit.
The only people serious about widespread implementation of solar are, indeed, thinking the way I am. The general concept is commonly referred to as "demand shaping" in the industry. Anyone still focused on supply shaping in 2024 is supporting coal, gas, and nuclear infrastructure. The supply shaping model attempts to resolve the differences in the supply and demand curves through grid level storage: attempting to broadly time-shift generation.
"Demand shaping" understands that storing power is inherently inefficient, and attempts to solve the differences by moving the time of consumption to the time of production.
The industry already has the solar capacity to meet the kind of demand I am talking about. They already have excess solar production that they can't effectively utilize, and we know that they can't effectively utilize it because it is regularly driving generation rates negative.
We are already producing (or capable of producing) the solar energy in question; we are wasting it due to a lack of demand. We are shutting down solar panels in the middle of the day due to a lack of demand. Solar rollout is stalling due to lack of demand for the specific power that solar is capable of producing.
When we create a demand specifically for solar energy, we increase the profitability of our existing solar infrastructure. We make it feasible and profitable to expand that infrastructure, which makes it pick up a bigger share of our normal load as well.
With that argument, we should all stop using microwaves, computers, ovens, dryers, electric cars, etc.
Everything related to technology or that uses power.
Are you comparing the use of things that helps take care of essential needs to the use of AI?
As for cars, the environmental impact is lower compared to gas cars so the comparison doesn't make sense.
"AI" is for the most part a waste of energy to regurgitate information that is very likely incorrect and that people could easily find by using much less energy by using the tools that already exist. They're just language models trained on so much information that it makes them unreliable.
Their usefulness compared to ovens... Well, there's just no comparison really.
AI proponents are the same as crypto fanatics during the last boom, people who are so hyped that they can't think about it critically.
Just because you don't use AI or can't understand its usefulness doesn't mean that it's useless.
I've been using AI every day for about 2 years now for various purposes.
If we're going to use and need electricity for everything including cars, computers, AI, ovens, microwaves, etc. and cut everything related to gas,
Maybe the solution is actually getting and using more electricity more efficiently.
And using it for AI when we have basic needs that still rely fossil fuel isn't using it efficiently.
Mind blown, right?
Evolution isn't going to wait that everyone is using electricity instead of fossil fuel.
If basic needs are still using fossil fuel, get up to date.
Would help if people stopped using renewable for useless stuff so we could use it for essentials instead of having to build new renewable infrastructure in order to provide energy for both.
You must also be a crypto mining apologist, right?
No, I'm not.
That being said, I'm quite fortunate. I live in Quebec, Canada where most power comes from hydro and is quite cheap.
I use AI on my own computer and everything that I own uses electricity, except my car. I personally pay for every bit of power that I use.
I also care for environment. Yes, I do have a gasoline car but I prioritize walking and my ebike whenever that I can and within the reasonable.
See, you're exactly the kind of person I'm talking about. Hydro Quebec will have to invest a fortune to create more infrastructure because of people like you. Our population is wasting about 25TWh at the moment and instead of pushing people to be fucking responsible with their energy usage, HQ will have to reduce their royalties to the government (that are used to pay for services to the population) in order to invest in the creation of wind turbine parks and guess what, that means we'll have to pay more taxes or everyone's hydro bill will increase to compensate for the little shits like you and there will be a huge environmental impact because contrary to what you morons believe we don't just pull the materials necessary to build wind turbines out of our ass.
So you wasting energy on that crap will have a direct impact on my own wallet so fuck you. You don't give a fucking crap about the environment, stop pretending you do.
First, I use Hydro Sherbrooke.
Second, I spend a total of around 30$ each month in electricity. Probably not as bad as you think.
Third, I work in tech and AI is an obvious part of my work.
Fourth, if my energy consumption triggers you that much, get a life bro.
Hydro Sherbrooke can only sell at the same rate as Hydro Quebec, doesn't make a difference what provider you use.