These kinds of articles really need to answer "compared to what?" at which point they immediately cash out into nothing.

So, 185,000 gallons of water were used to train GPT-3 which is apparently enough water to grow just 100 lbs of beef or 210 gallons of milk.

(420,000,000 lbs of beef are discarded in the US every year just due to discoloration)

That seems like a trivially good deal.

I saw similar faux outrage over carbon emissions from training large models. For example, this paper[1] involves hand wringing over the fact that training an original BERT model could have similar emissions to a trans-Atlantic flight. That despite the fact that thousands of such flights occur daily and BERT is (was?) used in a majority of all Google searches.

It actually seems like they mean a single flight as in, a single person's share of CO2 emissions from a flight, not like, the whole flight itself. Their reference for the emissions is this paper[2] which says BERT took 80 hours on 64 v100's and estimates that as 1.4k pounds of CO2 compared to Googling which suggests a single person's share of a flight from the US to the UK is around 1k kilograms of emissions (and about 10-20 dollars of carbon emissions offsets).

Some people just want to find things to complain about.

1 - https://dl.acm.org/doi/10.1145/3442188.3445922

2 - https://aclanthology.org/P19-1355.pdf

if there aren't water shortages where the data centers are it's not really a problem is it

Does ChatGPT consume fresh water or it is OK with salt one?

What happens with the water next?

And how 185K gallons of water may be sufficient to produce 370 BMW cars?

Crypto, particularly the Ethereum ecosystem and many other newer blockchains out there were able to reduce their carbon footprint significantly and ditch adopting proof of work, showing it is possible to switch from PoW to PoS.

Deep learning on the other hand still has not found any better efficient alternatives or solutions to training, inferencing and fine-tuning their AI models like ChatGPT, or GPT-4 and caused excessive and unecessary emissions for decades. Now it is water that they are wasting.

Absolutely wasteful.

You are joking right?

Ml has real benefit crypto doesn't.

'Bitcoin alone is estimated to consume 127 terawatt-hours (TWh) a year — more than many countries, including Norway. In the United States, cryptocurrency activity is estimated to emit from 25 to 50 million tons of CO2 each year, '

We already know Proof-of-Work (PoW) is wasteful and that is outside of my point. My point is, crypto already has greener alternatives to PoW available today and a former PoW blockchain was able to switch to a greener alternative, proof-of-stake, making that a known possibility.

Deep Learning on the other hand has NO known efficient alternative methods in training / fine-tuning and inference of their AI models today. All these models are complete black-boxes that require costly fine-tuning, retraining and can hallucinate on the slightest bad input. Once that happens, go through the same wasteful steps again.

> Ml has real benefit crypto doesn't.

AI bullshit generators, disinformation accelerants and the proliferation of deep fakes sure is a better use case than transparent transactions being viewed by anyone on a public chain making fraud harder to commit and easier for the authorities to trace up. /s

AI grifters will do anything to drive and burn the whole world to the ground for their so-called utopian AGI end-game nonsense to pretend about having 'real benefit' to everyone.

There is nothing easier than making ml training green.

There is NOTHING connecting those two things.

Ml/agi is revolutionizing our society.

Crypto made a lot of people poorer...

There you have it. Absolutely ZERO examples of existing technologies to solve the eternal emissions problem that Deep Learning still has for training, inference and fine-tuning at scale in both cost and waste created.

All hype and hot air from the AI bros.

> Ml/agi is revolutionizing our society.

A 'revolution' in unchecked nonsense.

This sort of euphoria without any consideration of the downsides and regulations to even begin reasoning about this AGI fantasy that won't help anyone other than accelerating more bullshit, disinformation and deep-fakes.

Green Datacenters are easy because we only have a few big ones in comparison to millions of small flexible Bitcoin miners.

Google is already working on green DC 24/7. Ms and was too.

Those companies have a lot of money and in comparison to crypto shit a real incentive to go green.

What is "consuming"? Does the water vanish when they're done with it?