How AI causes serious environmental problems (but might also provide solutions) | DW Business
ฝัง
- เผยแพร่เมื่อ 30 พ.ค. 2024
- Companies and the academic community are rushing to take advantage of big, recent advances in artificial intelligence. In many cases, they're using AI to minimize humanity's impact on the environment and find solutions for the climate crisis. But AI consumes an enormous amount of natural resources, like water and electricity, in order to function. So, is AI a climate solution or problem?
Chapters:
0:00 ChatGPT, how can I help save the environment?
0:53 AI's positive climate impact
3:37 AI's resource drain
7:55 Balancing the scales
9:11 AI regulations and the environment
12:20 Conclusion
#dwbusiness #ai #energy
Subscribe: th-cam.com/users/deutsche...
For more news go to: www.dw.com/en/
Follow DW on social media:
►Facebook: / deutschewellenews
►Twitter: / dwnews
►Instagram: / dwnews
►Twitch: / dwnews_hangout
Für Videos in deutscher Sprache besuchen Sie: / dwdeutsch
ai #environment #naturalresource
AI can find solutions to the climate crisis that we won’t use because other options are better for short term profits.
We have already had the solutions for the past 60 years. This is just BS so people associate AI with fixing climate change. That's a good thing right?
We don't need AI to find solutions; we know the actions required to minimise the impacts. Climate change is one of the symptoms of ecological overshoot, and AI can't help us with that, especially since introducing new technology usually leads to more resource and energy use, as shown in this video.
Lots of pipedreams start with the words "Our AI"
And with greed .. AI = money:
Thanks for covering this. This needs more attention. Shipping, aviation, sure. But most forget the enormous and fast-growing datacenter impact.
A medium-sized data center needs about 200 liters per second. This means it could process 200 * 30 (= 6000) questions per second. This corresponds to around 260 million questions per day. It should be noted that the data center also consumes electricity when it is not processing requests.
How is fixing an error faster going to gain you a 40% energy reduction? Let's be honest here, this is not about energy efficiency, it's all about personal efficiency. If your technician can fix an error twice as fast, you'll need only one on a shift not two. And that's why the company that runs the building will pay for the AI services. The eventual minimal energy efficiency improvement will be a bonus, not the main goal.
“An engineer in a bottle” can possibly save energy.
If you would have the amount of intelligence from a human instead of an AI system:
-You would need a lot of people, which have a high salary and spend their income on products producing carbon emissions. A holiday vacation. Food. Marriage…
So in that sense it is reasonable to assume, we can get very good intelligence for lower cost.
And that intelligence can work on climate solutions, monitoring…
Sounds realistic to me, that energy can be saved, new solutions can be implemented.
Even if energy usage will spike a lot.
Fun fact it could make our planet 100% renewable by doing all the jobs in order to create renewable energy but for free
Maybe they are running the building all windows open and the AI suggested to maybe close them...only way I can see you could gain 40%
Efficiency is code for profitable.
U should always have 2 techs anyway in case of backup when one is off. Crunching tons of data with the AI is great but it will never replace effiency of a real human being. The efficiency of a well trained, real human if utilized well will forever stay on the top and unsurpassed😾
'Forestry Waste' Now you really have to swallow hard for this one
Greenwashing at its finest.
These AI *solutions* are not enough balance the costs. Not to mention the untold amount of human slave like labor extracted to create it. (One google search is enough to prove)
To create what? The AI models or labelling the datasets?
@@aurethebest98I assume they means the mineral mining etc. required to make the hardware.
tbh there is no way to harness AI and use it for human advancement without incuring the cost,, the goal is just to minimize said cost
I knew that it spends a lot of electricity now the water is an issue as well.
They don't drink water. It is only used for cooling so after it is used by data centers and it cools down it still can be used by people for other things.
@@michaldopieralano it can't. It becomes a waste water, the same as water that we used for washing hands, dishes, flush toilets...
"Yes AI is terrible for the sustainability which include AI" says AI
Why don't they just build the centres somewhere cold ?
It cost more to build somewhere really cold, likewise, it costs a lot to establish the infrastructure to transfer anything including data between population centers and datacenters. It is less of an issue in Europe, where many of the economic/population centers have cold seasons and datacenters can be incorporated into local heating solutions (or not, in the of a place like the UK). However, a disportionate amount of large US cities and economic centers are located in what is basically deserts.
Like... where? You will find that not too many people are living in places that are cold all year round, and as nobody is living there, infrastructure of all kinds are absent from these places. Take Antarctica for example. But cooling is only one issue. As mentioned in the video, the need for cooling comes from the high electricity usage in the first place. So operators will aim at locations where they can get free electricity. And then there are further criteria like finding qualified operations and support staff, etc.
a company tax on electricity, relieved by lowered employer tax, helps two problems at once as AI consumes power and jobs.
And the company will hire a thousand folks from Nepal to pedal in 8 hours shifts 24-7 to produce electricity.
Remember the bitcoin mining craze? The multi-billion farms were hooked to residential grids at allegedly residential rates, but more often they just bypassed the meters... as the professor said in the movie, "nature will find its way"
Cloud computing is underrated technology, mainly because of the connectivity issue
Very important to highlight this!!!!
1:29 Control + F also works pretty well and requires almost no computer power.
Great pronunciation of Uruguay! No one ever gets that right! Kudos!
No mention about how AI could advance research into carbon-neutral energy production, battery technology, etc.
If you as a neighbour hear the electric humm of a nearby data center, there are a few revolutionary novel concepts like acoustic insulation according to building standards that could be adjusted by regulators for such data centers.
I think in order to cool Ai servers, the piping should be made of aluminum so that salt water can be used to cool servers and not put pressure on fresh water. From that angle the alloy of aluminum can aid solve water needs for Ai servers, as for electricity that is encrypted.
i can`t understand how come the cooling system is not a closed system so it uses mostly the initial amount of coolant and adding more water just to top up every time or so..
because a closed cooling system cost, well... money. it just cheaper to use drinking water than a huge filtration system for the local pond.
I actually laughed out loud when they played the hum and said this is what people here 24 x 7. Literally could barely hear that sound.
So building AI costs alot of resources(emissions), water and energy to run(also emissions), and then we can use that AI to make some efficient changes, and also to make advertisement more efficient, sales, videos and effects that make us glue to the same products that AI is build from. Sounds like we are taken over by machines, and are powerless to change.
Sounds like AI will replace alot of jobs, and at the same time degrade environment to some extend. I would want to see some analysis on the overall positive and negative impact and what is the ballance.
I recently heared from several youtubers about whistle blowers from US tech companies alarming the public about how these tech companies have misslead the public to believe that their hiring stop or rather the firing of employeers would be still due to the pandemic hirings. In contrast to that these HR whistle blowers claim that it is due the implementation and use of more and more AI within these companies. Why have 5 employees if one can do the job of them all with the help of AI? I doubt that any company who has their ears on the pulse of their sector will miss the fact how some companies become leaner, having less cost, being more productive while pressuring the pricing of their products or filling their war chest to buy up other companies. The race is on and the employeers will be the cattle brought to the slaughter.
Companies are downplaying that this would be happening not for a long time, while they already adepting it to processes and workflows.
Instead of wasting water to cool the data center, why not use the heat to heat up the water heater?
I have to say that if das goog can function at least in one Svensk facility in an eco neutral manner, so any such facility can find a way to do so.
The sole pursuit of researching and learning about AI and machine learning spend a lot of energy compared to other disciplines
As technological 'solutions' always did more damage than good, AI will be another coffin nail for mankind.
Build the servers underground~ It'll reduce noise, insulate them with cool stone which will reduce cooling needs, and leave the surface space for communities to use differently~ You could also use the heat generated by the servers to heat water to run generators to make the electricity for the servers' AC. I feel there must be a way to create a closed loop where the water is recycled through the system... that way there wouldn't be a need to draw any freshwater from the surrounding area... Or maybe they could create a closed loop coolant system that uses super cooling fluids instead of water? There are options~ They shouldn't need to rely on using people's drinking water.
U sshould always have 2 techs anyway in case of backup when one is off. Crunching tons of data with the AI is great but it will never replace effiency of a real human being. The efficiency of a well trained, real human if utilized well will forever stay on the top and unsurpassed.
THIS is a major reason I don't worry about an AI taking my job. It costs a company less to hire a 3rd world or otherwise desperate person that they didn't have to feed or train than the compute power required for an AI.
Nature has been fine tuning the process of making an energy efficient brain/body combo for some time now and making people is relatively easy.
Not mention the fact that cheap labor is still overwhelmingly used today despite advancing manufacturing automation technology is telling about the overall nature of human economy.
lol how far into the future do you think that’ll last? It’s getting cheaper and cheaper
This is temporary. You can now run very capable models on your computer.
I love how the water usage is always quoted as being so massive.... but they never say what happens to the water after use? it doesn't just go poof! It either gets evaporated (creating rain which is much needed!) or getting cleaned and made ready for re-use.
The water is never going to be hot enough to evaporate. It will typically get discharged back into the lake/river/stream from where it came - which is a problem for marine life.
In my humble opinion the most effective way for companies as of yet to actualy strive to be carbon neutral, I think would be to invest in optimized planting as of both economicaly and ecologicaly, which would create jobs and food, and to solve and spread the energy creation problem the sooner the better.
I believe in efficiency, a path where preservation of energy starts having higher value than squandering it.
As for big tech, plundering billions in government funds to build renewables and still ending up requiring mostly fossil to operate, exactly as predicted. There are many conclusions that one can draw from this, none are favourable for any party involved to admit and the charade continues.
Companies should invest more money on research and development on building these data centres under water using the wave energy.
Intriguing reflection on evolutionary discourse about the costs of evolving a large brain.
i think the first example is key.. yes - using AI takes energy, but is it more effecient to use AI than to do it manually? Would looking up error 15 and getting th manual etc consume more than asking AI to help? The manual way is harder to calculate - human time, cost of extra effort - food eaten, etc.. its much more complex than just thinking of the electric\water bill of the AI.
If you look it up error 15 on a computer.
You may google a bunch of times or open a document and use the device for a couple of minutes.
The advantages of fixing 'error 15' are exaggerated. AI might provide an answer quickly but how long then does it take to apply the fix? If error 15 was a computer fixable problem, it would not be flagged as error 15(think how scandisk automates fixings disk errrs); so error 15 is likely to be a problem that the fix cannot be automated. ie it's a manual fix job, which would likely take a lot more percentage of the time to fix than looking up the problem. Additionally problem and their fixes can be indexed for quick access in the future. If only comparing the use of the brute force approach then ai is faster; but why would humans artificially constrain themselves to only using the brute force approach?
Hmmmm. I'm thinking of a northern country with so much cold fresh water flowing that we observe that it is harming a critical ecosystem. All it would need is a new, preferably green power plant. Greenland ought to capture some of the glacier melt that threatens the AMOC, sell some of that to countries already desperate for fresh water, and permit new data centers. Yes, the water intake would have to be pumped higher and require frequent and dangerous maintenance, but at least LESS of it would go to waste and we might figure out how to slow glaciers that are floating on this runoff.
Car gasses, are the least people should worry...
How many currently empty office buildings in the cities of the USA could be converted to AI data centers? Many of those empty buidings are currently lit up and kept cool in the summer and warm in the winter using energy without benefiting anyone since the building owners are getting no rent money from tenants.... Just wondering here in Portland OR USA.
It's irrelevant to the topic, but I am curious where Sebastian Lehuede 5:34 is from because his accent is cute...
Also training AI models is a one-time process. Meaning once it's trined it can be use by everyone. It is true training will be done monthly or even weekly, but this is nothing compared to oil company's.
Nice
It's sad to think that as we're watching this, it's using resources....it ain't free....one of the reasons I rarely use large language models, except Google of course. IOW I use massive quantities, like everyone else.
real artificial intelligence would be able to figure out how to power itself without money
It is still physics. There is no harry potter magic to solve energy issues.
@@antman7673 While you are correct that AI needs to abide by the laws of physics, there supposidly could still be laws of physics that we are either unaware of yet or have not complettly understood. That gives room for inovation and as A.C. Clerk said:"“Any sufficiently advanced technology is indistinguishable from magic"
Powering without money isn't the issue anyways, it is powering without us humans. Sofar that is just not possible ... yet.
just watching this video alone equal of one gallons of clean water and energy equal to fully charge my phone.
Good
I think it will be extremely beneficial in the long run it will be able to revolutionize and make the world carbon negative revolutionizing renewable energy sources
every AI datacenter is power limited. By that we mean the data center can't add any more GPU due to power and water. The newest Nvidia GPU uses any where from 350-800 watts of power and a AI datacenter can have 10-24K GPU. do the math and you'll see it's kilowatts of power every hour just to answer chatGPT questions. That means every month, a busy datacenter can consume megawatts of power.
They're paying they use... but one big question is, why is the water even lost after usage.
It's just running through the data center to cool it.
Building in water stressed areas is not even necessary, so that's just a big question mark.
lot of computing power is wasted in unecesary things
Just played HORIZON
Hmm.... recycle the cooling water...
They must build solar farms alongside their data canters, they can't be pushing the cost onto the local community.
Worry not, they will push it onto the local community.
AI can be helpful in the sense that if it goes rogue and destroys all humans (depending on the method), the world would be in a better state and may begin to heal... unless it subsequently turns the planet into a giant machine, of course.
Might,,,, maybe,,,, perhaps,,,we believe,,,we think that,,,,as far as we know,,,, it's possible that,,,etc,,etc,,,etc,,,
So why are we running data centers on Earth, rather than in space, where solar energy is abundant and cooling is easy?
Because cooling in space is NOT EASY AT ALL. In fact it is incredibly hard. Space is "cold" because there is a vacuum which doesn't ransfer energy, so there is nothing that sunlight could heat up. But that means that a hot body in a vacuum also can not transfer energy. So although space is technically much cooler than a high-performance computer, there is no way for the heat in the computer to go out into space, except for radiation and that is incredibly inefficient. Those huge panels space installations have? That's not just for collecting sunlight, that is also for creating the largest possible surface to radiate excess heat.
A IA irá consumir o planeta......
Let’s be honest. These companies are doing a nice “greenwashing” game. Of course, as always, the local populace gets the short end of the stick.
Curb fossil extraction 2% of today's level per month down to zero by 2030. AI will figure out how to power itself fossil free.
You have to have zero knowledge about previous implementation of energy optimization solutions to believe that AI building monitoring/control can lead to energy savings. Unless you ensure that there is 0.000% percentage ability for humans to affect control.
I won't call it a solution if it solves the problem that it caused.
AI will provide many solutions. One of them will eventually be to eradicate humans as we will present an eventual obstacle to AI's growth and development.
how much carbon does a youtube video emit? or 5 million subscribers watching videos? Is DW responsible for carbon emissions emitted indirectly via hosting, streaming, mobile battery consumption etc. If you think about it, digital media is built on carbon emissions.
No DW is not. TH-cam is. I don't get your point.
AI is so powerful and sophiscate, there will not be any workers left in this world. we will all have AI machine does work and make money for us. and I guess that is the future of AI. priinting your own money, making your house, helping you you to do anything with a touch screen. so much hype and potentials for doing everything, including saving the earth and stopping politicians to take control of our life.
keep dreaming
Tell me you don't understand how the world works without telling me 😂😂😂😂 Print money, he says. Man, you made my day hahahahaha
Ask chatgpt
☺️☺️☺️☺️☺️☺️
stop buying luxury item, watch mvie, don't use AC ....etc before questioning AI electric uses.
well people been doing the crypto bs and scams.. so AI is at least actually helping except for job loss though
Nuclear energy. Power problem solved
Good luck with that. No one is willing to invest into nuclear… and we need that energy to maintain the current level without fossil fuels, adding AI to the mix just creates more energy consumption.
The statement that if you request twice as many queries it accounts to twice as much compute cost is erroneous in every facet of the statement
Human blindness 🤦🏼♂️
AI can help. Heat usage just needs to fall and they need to stop wasting computer power on that cryptocurrency nonsense.
I have calculated that moving to renewables is much cheaper to 100% finance families acquiring zero cost renewable energy(sun,air, battery), than the Government giving subsidies, tax breaks , costs of improvement of the electric grid and spending on country security.
This finding changes the present approach to funding transfer to renewable energy.
I can provide you with receipts for 10kw solar, 20.2 kw Inverter, 14.5 kWh battery and a small air turbine with a cost less than 5000 USD.
This 100 % financing of renewable energy gives the families zero cost energy and will bring them close to abundance and freedom. No UBI will be needed.
Furthermore, we see Ukraine is able to destroy energy sites thousands of miles inside Russia. Countries without energy have lost all kinds of war.
It will be impossible for someone to destroy millions of home distributed energy sites.
Surely this is more useful and more productive than mining Bitcoins 😂
15 questions to ChapGTP cost half a liter of water? I'm skeptical.
A medium-sized data center needs about 200 liters per second. This means it could process 200 * 30 (= 6000) questions per second. This corresponds to around 260 million questions per day. It should be noted that the data center also consumes electricity when it is not processing requests.
Thata's why there is a need 4 fight😂 4 ai
Oh so you mean it would mean that someone would have a job and know the building, what a waist of time money and resources your bs ai......
data center r local drinking water supply? no way! i have already reined in my use. its overrated.
2x as much lol who is this guy. the best models were trained on the same chips as consumer RTX cards with some more processing and ram packed into a single card. not supercomputers.
And now let's talk about crypto...
We can mention heart diseases, autoimmune sicknesses, cancer and diabetes too!
AI Should be used on sicknesses like these, first!
If AI fixes problems like osteoarthritis and other chronic sicknesses than it is a definite YES!
As we are gonna get older almost everyone will be affected by it.
As of now already, osteoarthritis is one of the biggest burden on the healthcare industry financially, worldwide!
This was a surprisingly uninformative and unhelpful video (although most things about AI are quite superficial, so it fits the pattern nicely). One could summarize the content of this 13 minutes in a single sentence: AI in general actually uses a lot of power, you know, and companies will have to do something with this.
Want to save the planet? Turn off your computer!
Use AI to solve this problem XD
I don't believe one chatgpt question costs the the amount of energy to evaporate 0.5 liters of water. That's an absurd claim.
Not with your boiler. It is just an energy comparison.
I checked up on it and boiling 500g of water (half a liter) is equivalent to:
-a 100W solar panel running for 27min 52s.
Depends whether it is GPT 4 or GPT 3.5.
Maybe GPT 4 requires that much energy.
These models are run on many High End GPUs.
It is not an impossible number.
Got to the point in the video:
It is 15 GPT questions.
A medium-sized data center needs about 200 liters per second. This means it could process 200 * 30 (= 6000) questions per second. This corresponds to around 260 million questions per day. It should be noted that the data center also consumes electricity when it is not processing requests.
AI = MORDOR ECONOMY
The dominant business practice of our time is to convince customers to buy something new they don't need, which doesn't even work, but produces higher profit for the company and/or promises market penetration or mass adoption to impress shareholders -- environmental factors only matter to the degree that they cause legal troubles or expose the company to public backlash. And AI, the way currently is, follows precisely this pattern: it's simply an intentionally overhyped, half-baked product companies try to upsell customers into. In most cases it works less well than existing solutions, and it costs more overall, some of which cost will be apparent to the customer only after companies managed to sufficiently chain/addict users to it.
You know what a waste of energy is. This video
Old and outdated news. Just look at Nividia's latest announcement (H200). It is 20x cheaper, that does not mean it going to ask less energy, it is actually going to ask even more becouse we want more!
But there is a lot of research into the efficiency of the chips, and that will benefit energy usage the most.
LOL More reasons why AI should be banned.
That was very poor reporting. Telling us something without telling us something
AI is too important. There are other less important things that can be acted on.