riskable

joined 2 years ago
[–] riskable@programming.dev 12 points 1 month ago (5 children)

That's not "upscaling". That's having the AI color it in for you. Like a comic artist who has a colorer (person that literally does that).

Upscaling just makes the image bigger (resolution-wise). It uses the same exact technology as regular AI image generation though 🤷

There's degrees to everything. AI haters are at the point where they're arguing with digital artists over what counts as art and it's getting insane.

[–] riskable@programming.dev -3 points 1 month ago (24 children)

Don't say, "stolen". It's the wrong word. "Copied" is closer but really, "trained an AI model with images freely available on the Internet" is more accurate but doesn't sound sinister.

When you steal something, the original owner doesn't have it anymore. AIs aren't stealing anything. They're sort of copying things but again, not really. At the heart of every AI LLM or image model is a random number generator. They aren't really capable of copying things exactly unless the source material somehow gets a ridiculously high "score" when training. Such as a really popular book that gets quoted in a million places on the Internet and in other literature (and news articles, magazines, etc... anything that was used to train the AI).

Someone figured out that there's so much Harry Potter quotes and copies in OpenAI's training set that you could trick it into outputting something like 70% of the first book, one very long and specific prompt at a time (thousand of times). That's because of how the scoring works, not because of any sort of malicious intent to violate copyright on the part of OpenAI.

Nobody's stuff is being stolen.

[–] riskable@programming.dev -1 points 1 month ago

I don't know about the carbon emissions, the water thing in the article is extremely misleading. It claims that AI is using up more water than the entire yearly consumption of bottled water. The water usage estimates include the water used to cool the power plants generating the power (running the data center).

The last study on this said that the actual usage of water in the data centers is 12% of the total water usage estimate. Data centers don't normally use that much water. It would be like Niagra Falls pouring water over every data center.

Simple reality check: If you look at the cooling system outside any given data center—if they're using as much water as d article suggests—they'd be emitting a massive cloud of water, 24/7. It would be so much, they'd need a cooling tower on par with a nuclear power station.

So what's with the statistic? If you look at any given power plant on Google Maps you'll see cooling ponds all around it. That's the water they're talking about. It's part of the build of the power plant. It's not using potable water that would be going into people's houses.

Having said that, 12% of the water usage is potable water—in the worst-case data center/power plant matchup scenario. Where you have an older data center that doesn't use modern closed loop cooling systems that don't lose as much water to evaporation. I don't know what the statistic is, but I can sure you it's a lot better than 12%. A wild guess would be 4-6%.

Background: I was a security consultant for many years and traveled all over the US going into many data centers (sometimes, by breaking in! Hah). Inside, they're loud AF (think: standing next to a jet engine) and outside they'll have some big ass cooling units that are also kinda loud but not as loud as some of these articles make them out to be.

That was about 7 years ago but I doubt much has changed since then. I guarantee that those data centers are still being used and have been renovated to support AI-style hardware. The power from the utility was just increased and more cooling units were added. I seriously doubt they did much more than that.

From what I've read about the new "giga scale" data centers, they're much more efficient (and quieter... Outside). Those are the ones we want. If we replaced all the old stuff with new stuff, the statistics in articles like this would drop by and order of magnitude (just a guess).

[–] riskable@programming.dev 4 points 1 month ago (2 children)

Shitty oil that's very expensive to process (and can realistically only be processed in the US). Venezuelan oil only makes sense when the price of oil is guaranteed to be high.

How do you keep the price of oil high? Decrease gas mileage standards for cars and trucks. Start stupid wars. Un-subsidize electric cars. Make it more profitable to produce gas guzzling trucks than sedans.

This Venezuela thing is just another step in the oil billionaire TODO list.

[–] riskable@programming.dev 18 points 1 month ago (6 children)

He's going to announce war with Venezuela. Because of fentanyl... Which doesn't come from Venezuela.

[–] riskable@programming.dev 5 points 1 month ago

Probably an anti-counterfeit watermark.

[–] riskable@programming.dev 0 points 1 month ago

Data centers typically use closed loop cooling systems but those do still lose a bit of water each day that needs to be replaced. It's not much—compared to the size of the data center—but it's still a non-trivial amount.

A study recently came out (it was talked about extensively on the Science VS podcast) that said that a long conversation with an AI chat bot (e.g. ChatGPT) could use up to half a liter of water—in the worst case scenario.

This statistic has been used in the news quite a lot recently but it's a bad statistic: That water usage counts the water used by the power plant (for its own cooling). That's typically water that would come from ponds and similar that would've been built right alongside the power plant (your classic "cooling pond"). So it's not like the data centers are using 0.5L of fresh water that could be going to people's homes.

For reference, the actual data center water usage is 12% of that 0.5L: 0.06L of water (for a long chat). Also remember: This is the worst-case scenario with a very poorly-engineered data center.

Another stat from the study that's relevant: Generating images uses much less energy/water than chat. However, generating videos uses up an order of magnitude more than both (combined).

So if you want the lowest possible energy usage of modern, generative AI: Use fast (low parameter count), open source models... To generate images 👍

[–] riskable@programming.dev 1 points 1 month ago

You know it was a joke, right? 😀

Pam Bondi couldn't put together a Lego set let alone a cash bounty system.

[–] riskable@programming.dev -3 points 1 month ago (3 children)

My remark about using copyrighted characters was in the context of AI. As in, that's a thing a lot of people seriously hate (with a passion).

Also, if I posted it anywhere but an anime community no one would get the joke! It'd be pointless.

[–] riskable@programming.dev -1 points 1 month ago

The power use from AI is orthogonal to renewable energy. From the news, you'd think that AI data centers have become the number one cause of global warming. Yet, they're not even in the top 100. Even at the current pace of data center buildouts, they won't make the top 100... ever.

AI data center power utilization is a regional problem specific to certain localities. It's a bad idea to build such a data center in certain places but companies do it anyway (for economic reasons that are easy to fix with regulation). It's not a universal problem across the globe.

Aside: I'd like to point out that the fusion reactor designs currently being built and tested were created using AI. Much of the advancements in that area are thanks to "AI data centers". If fusion power becomes a reality in the next 50 years it'll have more than made up for any emissions from data centers. From all of them, ever.

[–] riskable@programming.dev -5 points 1 month ago (5 children)

Without Made in Abyss this comic wouldn't make any sense. Would you have preferred if I used copyrighted characters from some anime?

[–] riskable@programming.dev 2 points 1 month ago

Historians of the future will mark this event as the beginning of the end of civilization.

view more: ‹ prev next ›