

Meta Uses Open Source AI To Design Greener Concrete For Its Next Data Center (nerds.xyz) 51
BrianFagioli writes: Meta has partnered with Amrize and the University of Illinois to develop an "AI-optimized" concrete mix that cuts carbon by 35% for its new data center.
No it's not (Score:2)
Re: (Score:3)
Does it cut carbon by 35% or does the AI say it cuts carbon by 35%? Assuming it does, does the new formula substituting in Elmer's glue actually hold up?
I've noticed a lot of people even really really high level people are trusting what it says waaaaaay too much.
Re: (Score:2)
The AI is still faulty.
Limecrete and Hempcrete exist and are far better than 35 % , they are carbon NEGATIVE!!!!.
Why did the AI not suggest this ?
We continue to stay on the path to extinction.
Re:No it's not (Score:4, Interesting)
The AI is still faulty.
No.
Limecrete and Hempcrete exist and are far better than 35 % , they are carbon NEGATIVE!!!!.
They do, and they are.
Why did the AI not suggest this ?
Because they have terrible compressive strength, and are generally used as insulation, not load bearing, what was not what they were looking for.
We continue to stay on the path to extinction.
The Earth naturally sequesters and emits carbon in the various carbon cycles of varying lengths.
It's not quite so simple to point at any single carbon emissive behavior and say "that there- that's leading us toward extinction!"
The biggest target needs to be fossil fuels. But in the interim, less carbon-intensive concrete will help.
Re: (Score:2)
Does it cut carbon by 35% or does the AI say it cuts carbon by 35%?
They replace about a third of the cement with fly ash.
So, yes, that reduces carbon since it uses less cement.
The fly ash comes from burning pulverized coal, so this technique won't work after coal is phased out.
Re: (Score:2)
The fly ash comes from burning pulverized coal, so this technique won't work after coal is phased out.
Whew! Good thing we're "reinvesting" in coal, then, isn't it? /s
Re: (Score:2)
They replace about a third of the cement with fly ash. So, yes, that reduces carbon since it uses less cement. The fly ash comes from burning pulverized coal, so this technique won't work after coal is phased out.
Pretty much all concrete uses aggregate. Fly ash is just one specific form of aggregate.
Using fly ash in concrete is not new, though. https://www.fhwa.dot.gov/pavem... [dot.gov]
Re: (Score:1)
Re: (Score:2)
What the "AI" did (Beyesian Optimizer, if we're being technically accurate) was help them find proportions that gave them an ideal trade-off between green and required compressive strength.
The "AI" portion, is what is called the surrogate model. It stands in as a replacement for a physical simulation. It's highly accurate, even for unknown configurations, and requires a microscopic fraction of the power to give its answer.
The B
Re: (Score:2)
Fly ash is a supplementary cementitious material (SCM), not an aggregate. .
Correction noted; thanks.
Re: (Score:2)
that's the f'up thing about this whole thing.... the company in charge "estimates" it cuts carbon by 35%.... Why estimated? it is something one can actually measure and calculate... no reason for an estimate
Re: (Score:2)
As for the CO2/kg for this particular concrete- yes that is more or less precisely knowable.
Also, FWIW- on the company's actual website, they don't say estimate- they say "at least 30%"
But again- unless you precisely specify the concrete mix you're comparing it against, you can't really give a % decrease, and giving the precise mix is practically meaningless for anyone b
Re: (Score:2)
AI *is* math.
Mind = Blown.
Re: (Score:2)
No. AI of this variant is statistics with is a small part of Math and one that basically has no connection to concrete mixes.
Re: (Score:2)
It's a predictive model trained to be a surrogate for a physical simulation of compressive strength of concrete over time, based on its mix.
You never get tired of being wrong. It's fucking incredible.
Re: (Score:2)
And you seriously think statistics is no math?
You already had better troll posts.
Re: (Score:3)
Bayesian optimization is good way to find better ways to do things if you have enough data to train a sound surrogate model.
The alternative is a physical simulation which may require supercomputers to do reasonably.
or... (Score:3)
you could not build it for 100% green plus the bonus for not actually burning more energy for AI porn generation.
Re: (Score:2)
also my understanding is this will be biggest center to date... 35 percent less of 200 percent bigger?
Re: (Score:2)
What's better, 0% less of 200% bigger, or 35% less of 200% bigger?
You wouldn't think this would be a difficult math problem.
Re: (Score:2)
i am questioning the purpose of growth, sometimes obesity is just bad
Re: (Score:2)
What is the purpose of growth? Simple- we can't stop fucking. All other growth follows from that.
Re: (Score:2)
Re: (Score:2)
Disclaimer - I'm not a concrete expert. But I do know that there are many varying grades and mixes of concretes depending on application. Sure, this may not be a fix for all concrete applications, but then no concrete formula is. That doesn't mean that lessons from this application can't be applied to others.
Re: (Score:1)
There exists concrete products already that are carbon negative.
Hempcrete and Limecrete.
You and the AI are both wrong.
Re: (Score:2)
*crete is not a set of synonyms made from different things.
They have a set of requirements, neither of which are met by the poor structural strength of limecrete and hempcrete.
Re: (Score:2)
Don't knock AI porn too soon... it eliminates human/child/sex trafficking needs and a whole shit ton of evil if porn was strictly/wholly AI generated...
people will do what people will do... AI porn generation would reduce victims to a degree from the loop...
Re: (Score:2)
I used AI the other day (Score:2)
I used AI the other day to invent anti-gravity. A few weeks ago it helped me solve world hunger.
AI used more than the savings (Score:3)
Too bad using the AI was a non-green solution to a green problem. How many times over will the AI in that datacenter burn though the savings that the concrete created, in a day?
What is the win from ditching the AI completely? You won't even need that datacenter, talk about going green.
Re: (Score:1)
Worse that its solution is wrong, and therefore a total waste of resources. ... ie waaaay better than 35% and these products are already commercially available.
Hempcrete and Limecrete are carbon negative
We are "Led by donkeys"
Re: (Score:3)
Re: (Score:2)
It was a surrogate model for a physical simulation- far cheaper in energy terms than a traditional physical simulation.
Whether or not that was all cheaper than just using traditional concrete is a quite simple "Yes."
Cooking limestone into calcium oxide emits 1 CO2 molecule per calcium oxide molecule. It's a fuck-ton.
The Iciocracy ... No need to do this. (Score:3)
Hempcrete is carbon negative.
If the AI doesnt reccommend this , then the AI is defective.
Once again more bullshit and #greenwash from the worlds billionaire cretins.
Heres a few citations in case you think im a loon.
( Im not ! thats the idiocracy )
https://www.researchgate.net/p... [researchgate.net]
https://www.researchgate.net/p... [researchgate.net]
https://en.wikipedia.org/wiki/... [wikipedia.org]
https://www.anthropocenemagazi... [anthropocenemagazine.org]
Re: (Score:1)
You've commented multiple times about hempcrete in this story, while completely disregarding that (according to your own Wiki link) it is not fit for purpose, as it's not load-bearing. The majority of concrete in any build is in the slab, which is load-bearing by definition.
Re: (Score:2)
The AI didn't invent something ...
What I'd like is some clever additive... such as carbonated water makes it better and greener concrete... and how it could be used for structural footings then a way to fill the rest with limecrete and foam blocks.
Re: (Score:3)
Since we seem to be calling this particular Bayesian Optimizer "AI" (fine by me), the AI did, indeed, invent it.
Re: (Score:3)
During WW2, the Japanese mixed coconut fiber into concrete to make bunkers.
The fibers reduced the amount of cement needed, and made the concrete more flexible so it was resilient to shelling.
Re: (Score:2)
Beautiful news .. (Score:1)
Re: (Score:1)
Unfortunately not
Hempcrete exists and is carbon NEGATIVE.
That the AI didnt suggest this proves your assertion that AI is going to save us is wholly incorrect
In all honesty though that the AI didnt suggest either Hempcrete or Limecrete which are commercially available and CO2 negative tells us either that the AI is unfit for purpose or for some reason these products were deliberately excluded. It tells us that those who enlisted the services of the AI have been promoted to a high level of incompetence.
There
Re: (Score:2)
Meta AI (Score:2)
How is most of that 'savings' created? (Score:2)
If they are increasing aggregate size to reduce the refined parts of the mix, that's also cause for concern.
And if this new mix is kept as a trade secret, why are they hoarding information that could help save the world?
And how much carbon did we burn to do this greenwa (Score:2)
Basalt based cement is carbon negative (Score:2)
I've heard of carbon neutral cement before. I've heard of carbon negative cement before. There's been papers by Prof. Darryl Siemer on this, here's one I could ind with a quick web search: https://www.ecolo.org/document... [ecolo.org]
That's not the best paper to describe his logic and goals, in part because it meanders about a bit and in part as it is more about CO2 sequestration by other means than by cement production, but it hopefully makes sense if you can tolerate reading it to the end.
I remember him because of
Re: (Score:2)
You could use crushed basalt as an additive (it does, after all have a lot of stuff in it that are already added to cement), but you cannot avoid the hydraulic lime, which you cannot get hydraulic lime without calcining some kind of limestone analogue.
That doesn't mean you can't have low or negative carbon cement- after all, hydraulic lime must re-absorb the CO2 lost in the calcining, and by adding additional CO2 absorbers, you can actually push it negative.
Basalt has bee
Re: (Score:2)
You don't get lime from basalt.
A professor from Idaho State University says otherwise. The link I gave above points out that "average basalt" consists of 12.4% CaO. This "average basalt" also consists of 7.9% MgO that will form a carbonate like lime when exposed to air and water.
You could use crushed basalt as an additive (it does, after all have a lot of stuff in it that are already added to cement), but you cannot avoid the hydraulic lime, which you cannot get hydraulic lime without calcining some kind of limestone analogue.
I can be convinced to agree, and I stated something similar in my earlier comment. I'll consider the chemical composition of cement I found here: https://civiltoday.com/civil-e... [civiltoday.com] (I chose this source as it uses the chemical notation I'm familiar with than th
Bad news is another new data center !!!!! (Score:2)