News Power Consumption of AI Workloads Approaches That of Small Country: Report

[Admin]

AI workloads will consume 4.3 GW of power in 2023, set to over quadruple in five years, says Schneider Electric.

Power Consumption of AI Workloads Approaches That of Small Country: Report : Read more

 

[Findecanor]

Now this is one reason why I am against AI.

AI is the new cryptocurrency. Making money at the expense of others and of the environment.
And scams will be prolific.

 

[wbfox]

Yay, stupid search enginges that hallucinate and steal all the intellectual property of humanity to "learn" are going to be adding to the stupid impact of great products like crypto and nft. Water needs to be reserved so some idiot can get poorly written code or be told a summation of human author's work, or be prevented from getting insurance because the AI doesn't think its a good bet or etc, etc, etc, etc. Time to stop the stupid. We can do these things, but why? So google and amazon and 2000 other companies can steal and sell your info to then sell or ban you from things. This is the darkest, dumbest timeline.

 

[TheOtherOne]

But we can always put the blame on third world countries for the major cause of Carbon Pollution and Global Warming!

Even tho WE "consume" way more electricity (sometimes even 5x/10x more) that is mostly generated by fossil fuel sources.

 

[bit_user]

Now this is one reason why I am against AI.

AI is the new cryptocurrency. Making money at the expense of others and of the environment.

Yes, but... one should also consider that if AI is used to replace human workers, what's the carbon footprint of those humans. If we're being honest, we really should look at both sides of it.

Even if it's not replacing humans, the economic benefits of AI imply that it's delivering greater efficiency or else people & businesses wouldn't be willing to pay for it. That typically means less of some sort of resource gets used in the process. So, unless we're talking about AI being used by fossil fuel companies to locate and extract more coal, oil, or gas, there should be some efficiency upside to using AI.

Then, the real question becomes how to make sure that upside is greater than the negative impact it has. And that brings us back to carbon pricing. Yes, I know it'll probably only ever happen long after renewable energy becomes dominant (i.e. due to the influence of fossil fuel lobbyists), but carbon pricing is ultimately the way to help ensure everything in the economy that still uses carbon is delivering more good than bad.

 

[bit_user]

The document specifies that the rack densities in substantial AI clusters vary between 30 kW and 100 kW

Wow... the notion of dissipating 100 kW in a rack is pretty mind-blowing.

One thing that's kind of sad is that AI chips are being run far beyond their window of good efficiency. I know it's not exactly analogous, but this article showed you could get about 77% of the performance from a RTX 4090 at 50% of the power:

Improving Nvidia RTX 4090 Efficiency Through Power Limiting

We tested the 4090 Founders Edition at 50% to 120% power limits

www.tomshardware.com

 

[bit_user]

We can do these things, but why? So google and amazon and 2000 other companies can steal and sell your info to then sell or ban you from things.

There are lots of other things it's good for. Improved medical diagnosis, improving crop yields, and even designing more efficient hardware.

It's just a tool. Whether it's good or bad depends on how we use it.

 

[Findecanor]

Yes, but... one should also consider that if AI is used to replace human workers, what's the carbon footprint of those humans. If we're being honest, we really should look at both sides of it.

Are you seriously suggesting that we should euthanise the unemployed?!

 

[bit_user]

Are you seriously suggesting that we should euthanise the unemployed?!

Not necessarily. There are two ways you could consider the carbon footprint of the human workers. One is to consider their lifetime carbon footprint. The other is merely the difference between their baseline footprint (i.e. just sitting at home and sustaining themselves) and the additional carbon used by the to commute and sit an an office which needs to be climate-controlled and have ample space, lighting, cleaning, plumbing, workstations, etc. for them to do their job.

 

[Co BIY]

Equivalent to the electrical use of Cyprus mean it's essentially a rounding error globally. That's a small Island in a very temperate climate.

May grow to to size of Iceland's use, again a small country with total population the size of a smallish city.

The cooling and energy requirements in the actual building housing the racks are an interesting challenge that I'm sure Schneider Electric can solve for a tidy but not unreasonable sum.

What is the ideal amount of CO2 in the atmosphere ? Not sure we know. It isn't zero.

 

[Co BIY]

One thing that's kind of sad is that AI chips are being run far beyond their window of good efficiency. I know it's not exactly analogous, but this article showed you could get about 77% of the performance from a RTX 4090 at 50% of the power:

It's the high price of the processors that result in them being run beyond their electrical efficiency sweet-spots. Perhaps when the producers are not as capacity constrained they can market a bigger chip but run it slower and cooler for the same performance.

When the processor/accelerator costs $40,000 then the electrical expense is negligible and will be disregarded.

And ... This is how it should be - We have many more resources to produce electricity than we have to produce semi-conductors.

And ... they are much more widely (and equitably) distributed. If we required increased electrical efficiency out of the processors rather than allow users to run them at the max it would increase the power of the Fab giants at the expense of many others.

 

[bit_user]

What is the ideal amount of CO2 in the atmosphere ? Not sure we know. It isn't zero.

A lot lower than currently! The goal should be to get back towards historical norm, which seems around 275-285 PPM.

​

Don't worry - natural processes will ensure it never gets to 0! In fact, if human output went to net zero, then it's unlikely it would even get back to that historical norm any time soon, since the atmospheric CO2 sticks around for quite a while.

 

[bit_user]

It's the high price of the processors that result in them being run beyond their electrical efficiency sweet-spots. Perhaps when the producers are not as capacity constrained they can market a bigger chip but run it slower and cooler for the same performance.

That's definitely a major factor. Limited real estate & management overheads are further reasons they'd prefer to run fewer servers faster. And we mustn't forget time-to-market, for companies training models in a competitive business environment (or even academic researchers rushing to beat paper submission deadlines).

So, I don't have much hope of frequencies scaling back much, any time soon. However, higher energy prices could certainly apply some additional downward pressure on energy consumption.

If we required increased electrical efficiency out of the processors rather than allow users to run them at the max it would increase the power of the Fab giants at the expense of many others.

Your homework: determine the approximate energy inputs per mm^2 of leading-edge silicon.

I don't believe it'll be anywhere close to the lifetime utilization of these chips, during operation. That said, I believe modern EUV lithography is indeed very energy-intensive. Still, it's not like our fabs each need to be paired with a nuclear reactor, nor have I heard of fabs requiring exotic cooling systems. I know there are further upstream energy inputs, but probably the bulk of it happens during the lithography & post-processing stages.

 

[Co BIY]

A lot lower than currently! The goal should be to get back towards historical norm, which seems around 275-285 PPM.

​

​

Don't worry - natural processes will ensure it never gets to 0! In fact, if human output went to net zero, then it's unlikely it would even get back to that historical norm any time soon, since the atmospheric CO2 sticks around for quite a while.

My local climate is great so perhaps the current is ideal. In fact, climate related deaths are at an all-time world-wide low. I consider that not a bad place to be.

Historical warm periods were times of great advancement and human flourishing. Why should now be different ?

Non-zero based graph is very dramatic ! (If that was in a GPU review the fanboys of one team or another would certainly notice.)

 

[bit_user]

My local climate is great so perhaps the current is ideal. In fact, climate related deaths are at an all-time world-wide low. I consider that not a bad place to be.

Flooding, storms, wildfires, heatwaves, loss of ocean-front property... are not only all very costly, but fueling mass migration which is not only a tragic story for each person uprooted, but also leads to political turmoil and societal stress on the destination countries and states. Yes, even wars can/have been triggered by climatic factors or events. Then, we could talk about the impact on food production, because droughts, flooding, and heatwaves aren't good for agricultural output.

Even if you're not personally impacted, price increases on food and insurance are likely to filter through society and increase costs on many goods and services we all use.

Historical warm periods were times of great advancement and human flourishing. Why should now be different ?

I'm certain all of these questions have been explored in great depth. There's been a great deal of modeling to investigate the impacts of climate change. It's a story where the bad very much outweighs the good.

However, if we try to look for an upside to warming, I guess there are some sub-arctic regions in Canada, Alaska, Siberia, Norway, and Greenland that would be more habitable. Plus, I heard on the news that people are starting to plant vineyards in southern Britain. Hardly sounds like a net win, to me. Especially when you consider the flip side: like failing olive crops in southern Europe that have even triggered a wave of olive oil theft, in Greece!

Non-zero based graph is very dramatic !

Fair point, but look at the numbers! 420/275 -> 70.9% increase. That's a lot! Obviously, you need to run that through a model to understand the impact it has, but just looking at such a substantial human impact on the environment should motivate one to understand the types and magnitude of impact it can have.

Just this year, there was about 1M sq. miles less sea ice in Antarctica than normal. That ice helps slow the movement of land-based ice sheets, and reflects solar radiation much better than open water.

 

[Co BIY]

Flooding, storms, wildfires, heatwaves, loss of ocean-front property... are not only all very costly, but fueling mass migration which is not only a tragic story for each person uprooted, but also leads to political turmoil and societal stress on the destination countries and states. Yes, even wars can/have been triggered by climatic factors or events.

Global weather related losses are declining as a percentage of global GDP.

This probably indicates that increasingly expensive storms (in Dollar amounts) are more expensive because we have so much more stuff to get damaged rather than that storms are becoming worse. I suspect that rates of insurance and damage reporting is much more comprehensive as well.

The evidence that there has been an increase in extreme weather events is not strong. Article after article on storms you'll hear the scientists say "There has been no increase but so far, but in the future we are sure there will be. "

The evidence that their are many many more expensive beach houses in Florida than ever before is pretty compelling.

As for Climate refugees and Climate wars - They are like Bigfoot and Alien landings. Some people seem to want them to exist but never seem to actually find them.


I do think tracking sea ice is important and interesting. Although we don't have many years of comprehensive Satellite data for long term studies.

 

[bit_user]

Global weather related losses are declining as a percentage of global GDP.

It's cute how that stops at 1H 2019, don't you think? Also, there are some awfully big spikes being "averaged away" by that trend line. I don't see where they say how the trend line was computed, in fact. Perhaps they trimmed some outliers, first?

I think it won't be long before you can't get insurance in large parts of Florida - or it'll be too expensive, if you can - and then its real estate market just implodes, because no one will issue a mortgage on an uninsured home.

This probably indicates that increasingly expensive storms (in Dollar amounts) are more expensive because we have so much more stuff to get damaged rather than that storms are becoming worse. I suspect that rates of insurance and damage reporting is much more comprehensive as well.

I think it was 2022 when we had so many named storms they ran out of letters in the alphabet and had to start using Greek letters. There are a couple other worrying trends we're seeing: rapid intensification and slower-moving storms. Acapulco, Mexico was just slammed with a Cat 5 hurricane which bizarrely intensified from a tropical storm in a matter of hours, in one of the most extreme such examples.

https://www.cnn.com/2023/10/27/weather/hurricane-otis-acapulco-mexico-images-climate/index.html

As for slower-moving storms, the amount of major flooding is increasing due to a tendency of major hurricanes stalling upon making landfall. That means they just sit in one spot, dumping horrendous amounts of rain.

The evidence that there has been an increase in extreme weather events is not strong. Article after article on storms you'll hear the scientists say "There has been no increase but so far, but in the future we are sure there will be. "

I think this tells a different story:

Source: https://www.gfdl.noaa.gov/historical-atlantic-hurricane-and-tropical-storm-records/
​

As for Climate refugees and Climate wars - They are like Bigfoot and Alien landings. Some people seem to want them to exist but never seem to actually find them.

Darfur was said to be possibly the first war sparked by climate change, due to the way it brought herders and farmers in conflict with each other, which then escalated and inflamed other tensions. I think that's how it happens - one tribe migrates after their way of life becomes untenable, upsetting historical balances.

Climate migrants are definitely a thing, but they're easily confused with economic migrants. I wish I could remember which one, but I remember hearing that devastating and historic flooding in a central American country recently sent a wave of migrants northward, after their homes and livelihoods were wrecked.

I do think tracking sea ice is important and interesting. Although we don't have many years of comprehensive Satellite data for long term studies.

Here's 173 years' worth of data from 9 sources. Even if the older data isn't very precise, perhaps it's enough to spot trends. For the past 44 years, there have been satellite observations.

Sea Ice Extent (Arctic only) - 1850 - Present - Science On a Sphere

sos.noaa.gov

 

[Co BIY]

I think this tells a different story:

​

​

​

​

Source: https://www.gfdl.noaa.gov/historical-atlantic-hurricane-and-tropical-storm-records/​

This is the label on the first graph that you just presented from the NOAA site listed above:

Figure 2: Atlantic tropical storm counts adjusted for likely missing storms. Once an estimate for likely missing storms is accounted for the increase in tropical storms in the Atlantic since the late-19th Century is not distinguishable from no change.

This is the label for the second graph you just presented from the NOAA site :

Figure 3: Normalized Atlantic Indices. Since the late-19th Century global (green) and tropical Atlantic (blue) temperatures have risen – an increase that was partly driven by increased greenhouse gases. If one does not account for possible missed storms (first red line) Atlantic tropical storms appear to have increased with temperature; however, once one accounts for possible missed storms (second and third red lines) basinwide storms have not exhibited a significant increase. When one focuses only on landfalling storms (yellow lines) the nominal trend has been for a decrease.

This is the summary key findings paragraph for the NOAA article you cited:

Recent papers (Vecchi and Knutson 2008; Landsea et al 2010; Vecchi and Knutson 2011.; Villarini et al. 2011) suggest that, based on careful examination of the Atlantic tropical storm database (HURDAT) and on estimates of how many storms were likely missed in the past, it is likely that the increase in Atlantic tropical storm and hurricane frequency in HURDAT since the late-1800s is primarily due to improved monitoring.

 

[bit_user]

This is the Label on the above graph at the NOAA site listed above"

Figure 2: Atlantic tropical storm counts adjusted for likely missing storms. Once an estimate for likely missing storms is accounted for the increase in tropical storms in the Atlantic since the late-19th Century is not distinguishable from no change.

You missed the part where they said "since the late 19th Century". However, the estimated data stops by 1970. You can use your eyeballs and see the change since then!

Figure 3: Normalized Atlantic Indices. Since the late-19th Century global (green) and tropical Atlantic (blue) temperatures have risen – an increase that was partly driven by increased greenhouse gases. If one does not account for possible missed storms (first red line) Atlantic tropical storms appear to have increased with temperature; however, once one accounts for possible missed storms (second and third red lines) basinwide storms have not exhibited a significant increase. When one focuses only on landfalling storms (yellow lines) the nominal trend has been for a decrease.

Still, both of the adjusted tropical storms plots show an increase. It's only the landfalling ones which show a decreasing trend.

Anyway, two things:

1. The data only goes up to about 2006. So, that's my bad, for not getting more recent data.
2. They're looking at trends over 135 years. Maybe that adds too much delay, before a real uptick gets reflected in the trendline.

If you just eyeball the data, you can clearly see:

• a big lobe on the right, showing a clear uptick in tropical storms.
• a clear correlation between tropical storms and global mean temperature (top graph).

 

[Co BIY]

Still, both of the adjusted tropical storms plots show an increase. It's only the landfalling ones which show a decreasing trend.

The scientists conducting the analysis and creating the graphs say the results are statistically the same as no change. (Except for the landfall decrease in both tropical storms and hurricanes interestingly.)

You missed the part where they said "since the late 19th Century". However, the estimated data stops by 1970. You can use your eyeballs and see the change since then!

My eyeballs see two spikes of almost equal height to the current one in the 1890s and 1930s. In the primary graph with a labeled data axis.

1. They're looking at trends over 135 years. Maybe that adds too much delay, before a real uptick gets reflected in the trendline.

A longer baseline should allow you to see a recent uptrend more clearly. But here it just shows that the numbers are pretty variable and not that consistent year to year.

If you just eyeball the data, you can clearly see:

• a big lobe on the right, showing a clear uptick in tropical storms.
• a clear correlation between tropical storms and global mean temperature (top graph).

Why would tropical storms show an uptick but landfall tropical storms and landfall hurricanes (For which we have the best historical data) show a decrease ?

By the "Clear Correlation" between tropical storms and mean global temperature I assume you mean that both trend lines on those graphs are inclined by about the same amount and in the same direction.

I note that the two graphs must be measuring very different variables on their unlabeled vertical axes.

Does that mean that the higher global temps get the fewer landfall hurricanes we will experience ? They have a very negative eyeball correlation in this graph series. (Maybe we should put a torch to the Alberta Tar Sands immediately and get Germany to replace their clean no emission nuclear plants with their dirtiest coal!)

I notice the article does not even mention the very strong negative Landfall Hurricane data, probably because that was already very well established and likely what this research was hoping to expand on.

 

[bit_user]

By the "Clear Correlation" between tropical storms and mean global temperature I assume you mean that both trend lines on those graphs are inclined by about the same amount and in the same direction.

Not the trend lines, but the actual shapes of the graphs. Where there are substantial upticks or dips in the global mean temperature corresponds pretty well to when there are increases or decreases in the number of storms.

The scientists conducting the analysis and creating the graphs say the results are statistically the same as no change. (Except for the landfall decrease in both tropical storms and hurricanes interestingly.)

Again, that they:

1. show a clear increase in storm frequency
2. show a correlation with global temperature
3. in spite of lacking data from the past 17 years
4. in spite of computing the trend over such a long time range

...are things that certainly give me pause.

Storm frequency is just one concern. As for rapid storm intensification, the main mode of that is warm ocean surface temperatures:

Scientists find two ways that hurricanes rapidly intensify

Hurricanes that rapidly intensify for mysterious reasons pose a particularly frightening threat to those in harm's way. Forecasters have struggled for many years to understand why a seemingly commonplace tropical depression or tropical storm sometimes blows up into a major hurricane, packing...

phys.org

Florida under water: Why was Hurricane Idalia so destructive?

One of the hallmarks of the destructiveness of a hurricane is how much water it dumps on ordinarily dry land. As Hurricane Idalia intensified, resulting in major flooding even before it made landfall Wednesday, University of Virginia hydrologist Venkataraman Lakshmi was tracking it.

phys.org

There's clear data on that increasing:

Pacific Ocean global sea surface temperatures reach record levels in 2023

In March and April 2023, some Earth scientists began to point out that average sea surface temperatures had surpassed the highest levels seen in a key data record maintained by NOAA. Months later, they remain at record levels, with global sea surface temperatures 0.99°C (1.78°F) above average in...

phys.org

Storm frequency is only one concern. Wildfires are another major area of global concern. This shows annual land area burned in the Western US:

• A 25-day increase in the annual number of days with synchronous fire danger was observed during 1979–2020
• A doubling in the number of synchronous fire danger days is projected by 2051–2080

https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020GL091377

This paper finds a higher risk of temperature-related death, if global warming exceeds 2 C:

Higher risk of temperature-related death if global warming exceeds 2 C

The death rate linked to extreme temperatures will increase significantly under global warming of 2°C, finds a report by researchers from UCL and the University of Reading.

phys.org