Only if you really think AI is going to solve global warming does any of this make sense at all.
Maybe only allow them to run where they can generate the CLEAN power thy need.
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.
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?