Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The librarian immediately attempts to sell you a vuvuzela

kaveland.no

Post reply on HN
Print view View original

281–290 of 346 posts

Re: The librarian immediately attempts to sell you a vuvuzela

#281

Post by root_axis »

autoexec wrote: post #267
baggy_trough wrote: post #259

Earlier quoted context omitted.

I agree that there is some incremental electricity usage. I do not think it can be characterized fairly as "massive environmental harm".

As an example, Ren and his colleagues calculated the emissions from training a large language model, or LLM, at the scale of Meta’s Llama-3.1, an advanced open-weight LLM released by the owner of Facebook in July to compete with leading proprietary models like OpenAI's GPT-4. The study found that producing the electricity to train this model produced an air pollution equivalent of more than 10,000 round trips by car…

> 10,000 round trips by car between Los Angeles and New York City.

That seems like very low impact, especially considering training only happens once. I have to imagine that the ongoing cost of inference is the real energy sink.

Re: The librarian immediately attempts to sell you a vuvuzela

#282

Post by ahepp »

autoexec wrote: post #267
baggy_trough wrote: post #259

Earlier quoted context omitted.

I agree that there is some incremental electricity usage. I do not think it can be characterized fairly as "massive environmental harm".

As an example, Ren and his colleagues calculated the emissions from training a large language model, or LLM, at the scale of Meta’s Llama-3.1, an advanced open-weight LLM released by the owner of Facebook in July to compete with leading proprietary models like OpenAI's GPT-4. The study found that producing the electricity to train this model produced an air pollution equivalent of more than 10,000 round trips by car…

> The study found that producing the electricity to train this model produced an air pollution equivalent of more than 10,000 round trips by car between Los Angeles and New York City.

I am totally on board with making sure data center energy usage is rational and aligned with climate policy, but "10k trips between LA and NY" doesn't seem like something that is just on its face outrageous to me.

Isn't the goal that these LLMs provide so much utility they're worth the cost? I think it's pretty plausible that efficiency gains from LLMs could add up to 10k cross USA trips worth of air pollution.

Of course this excludes the cost of actually running the model, which I suspect could be far higher

Re: The librarian immediately attempts to sell you a vuvuzela

#283

Post by astrange »

autoexec wrote: post #277
umvi wrote: post #270

Earlier quoted context omitted.

> the people whose work has been stolen to create it "Stolen" is kind of a loaded word. It implies the content was for sale and was taken without payment. I don't think anyone would accuse a person of stealing if they purchased GRRM's books, studied the prose and then used the knowledge they gained from studying to write a fanfic in the style of GRRM (or better yet, the final 2 books). What was stolen? "the prose sty…

AI has been trained on pirated material and that would be very different from someone buying books and reading them and learning from them. Right now it's still up to the courts what counts as infringing but at this point even Disney is accusing AI of violating their copyrights https://www.nytimes.com/2025/06/11/business/media/disney-uni... AI outputs copyrighted material: https://www.nytimes.com/interactive/2024/01/…

> at this point even Disney is accusing AI of violating their copyrights

"even" is odd there, of course Disney is accusing them of violating copyright, that's what Disney does.

> AI is getting better at data laundering and hiding evidence of infringement, but ultimately it's collecting and regurgitating copyrighted content.

That's not the standard for copyright infringement; AI is a transformative use.

Similarly, if you read a book and learn English or facts about the world by doing that, the author of the book doesn't own what you just learned.

Re: The librarian immediately attempts to sell you a vuvuzela

#284

Post by astrange »

gknoy wrote: post #250
baggy_trough wrote: post #241

Earlier quoted context omitted.

What is this massive environmental harm? That sounds like hyperbole.

Training AI models uses a large amount of energy (according to what I've read / headlines I've seen /etc), and increases water usage. [0] I don't have a lot to offer as proof, merely that this is an idea that I have encountered enough that I was suprised you hadn't heard of it. I did a very cursory bit of googling, so the quality + dodginess distribution is a bit wild, but there appear to be indiustry reports [2, pag…

There's an "AI is using all the water" meme online currently (especially on Bluesky, home of anti-AI scolds), which turns out to come from a study that counted hydroelectric power as using water.

Re: The librarian immediately attempts to sell you a vuvuzela

#285

Post by autoexec »

root_axis wrote: post #281
autoexec wrote: post #267

Earlier quoted context omitted.

As an example, Ren and his colleagues calculated the emissions from training a large language model, or LLM, at the scale of Meta’s Llama-3.1, an advanced open-weight LLM released by the owner of Facebook in July to compete with leading proprietary models like OpenAI's GPT-4. The study found that producing the electricity to train this model produced an air pollution equivalent of more than 10,000 round trips by car…

> 10,000 round trips by car between Los Angeles and New York City. That seems like very low impact, especially considering training only happens once. I have to imagine that the ongoing cost of inference is the real energy sink.

It doesn't happen only once. It happened once, for one version of one model, but every model (and there are others much larger) has its own cost and that cost is repeated with each version as models are continuously being retrained

Re: The librarian immediately attempts to sell you a vuvuzela

#286

Post by GuinansEyebrows »

umvi wrote: post #270
autoexec wrote: post #239

Earlier quoted context omitted.

> As a user it's a great free ride though. Maybe there IS such a thing as a free lunch after all! if you consider the massive environmental harm AI has and continues to cause, the people whose work has been stolen to create it, the impacts on workers and salaries, and the abuses AI enables that free lunch starts looking more expensive.

> the people whose work has been stolen to create it "Stolen" is kind of a loaded word. It implies the content was for sale and was taken without payment. I don't think anyone would accuse a person of stealing if they purchased GRRM's books, studied the prose and then used the knowledge they gained from studying to write a fanfic in the style of GRRM (or better yet, the final 2 books). What was stolen? "the prose sty…

> It implies the content was for sale and was taken without payment

that's literally what happened in innumerable individual cases, though.

Re: The librarian immediately attempts to sell you a vuvuzela

#287

Post by thih9 »

karel-3d wrote: post #78

The money situation is what gives me pause with LLMs. The amount of money it's burned on this is giant; those companies will need to make so much money to have any possibility of return. The idea is that we will all spend more money on AI that we spend on phones, and we will spend it on those companies only... I don't know, it just doesn't add up. As a user it's a great free ride though. Maybe there IS such a thing a…

Speculation:

AI companies want the cultural shift, i.e. get everyone used to having their data, art, work, etc, turned into models. Plus they want the PR, i.e. AI agents to be seen as helpful, friendly and genuinely useful. They want this to happen fast and before legislators react too. Releasing for free seems safe and efficient.

Re: The librarian immediately attempts to sell you a vuvuzela

#288

Post by GuinansEyebrows »

hug wrote: post #15

The author would do well to check out Kagi -- the search results for all of the suggested queries are, to my eye, better than the results they found. Given that Kagi's higher tier plans come with search-enabled LLM chat interfaces, and those searches use Kagi's results (which, again, appear to be superior) it seems to me that you get the best of both worlds: Better search, and better search results to feed into your…

i would probably pay for kagi if there was a premium tier that excluded LLM features.

Re: The librarian immediately attempts to sell you a vuvuzela

#289

Post by ASalazarMX »

thih9 wrote: post #287
karel-3d wrote: post #78

The money situation is what gives me pause with LLMs. The amount of money it's burned on this is giant; those companies will need to make so much money to have any possibility of return. The idea is that we will all spend more money on AI that we spend on phones, and we will spend it on those companies only... I don't know, it just doesn't add up. As a user it's a great free ride though. Maybe there IS such a thing a…

Speculation: AI companies want the cultural shift, i.e. get everyone used to having their data, art, work, etc, turned into models. Plus they want the PR, i.e. AI agents to be seen as helpful, friendly and genuinely useful. They want this to happen fast and before legislators react too. Releasing for free seems safe and efficient.

And this is already happening. I have a few acquaintances that will ask ChatGPT thinks they could readily find on Wikipedia, or use it as a shortcut for mental effort. They've become dependent, even addicted, to this patient butler that can work and think for them; who cares if it makes some mistakes? They'll just ask it more.

Once enough people become like this, they will gladly pay to keep it, they'll consider it a basic necessity.

Re: The librarian immediately attempts to sell you a vuvuzela

#290

Post by coryrc »

svnt wrote: post #274
coryrc wrote: post #269

Earlier quoted context omitted.

trump stooge activities cannot be blamed on data centers. The relevant technical authorities did not want this.

What are you talking about? This is literally the only mechanism to allow coal-fired plants to avoid sunsetting on schedule. Who are the “relevant technical authorities”?

"DOE issued the emergency order without a request from the plant owner, transmission provider or grid operator"
Post reply on HN