freagle

What a fucking moronic oped. Open-weight models developed by China do not need to be run on Chinese infrastructure. You can run open weight models anywhere you want on any infrastructure you want. Contrast this with ChatGPT, Grok, and Claude: you can’t run them locally because they hold the models to be proprietary and you have to run it on their corporate infrastructure.

The whole essay reeks of orientalism, so it’s no wonder this knuckle dragger doesn’t know what the fuck he’s talking about.

Actually I am wrong. You can skip season 1. Just seasons two and three to start with

Yeah, I think it’s worth listening to. They’re short and easy to consume episodes. Very This American Life style production. Very accessible. Give it a whirl.

You don’t bring up race issues because you’re white and you say “I identify as Anglo Saxon”? I would offer to you that you have some work to do in deepening your understanding of race, racism, and whiteness. I would recommend to you a great podcast: Scene on Radio, seasons 1 through 3 seasons 2 and 3

Is everyone as unconsciously racist as they are in my chapter full of white 20-something liberals with rich parents?

Because NVidia made a whole bunch of GPUs that are specifically for parallel processing massive amounts of information and floating point math, which is what is needed for machine learning, model training, LLMs, and everything else we call “a.i.”

They took on a huge amount of investment in 2024, raising $25B in debt to fund more production. In the last 3 years they have sold a ton of hardware. That hardware was bought by people with a ton of cash or a ton of credit.

How much did they sell? In the last 3 years, something like the equivalent of 12GW of GPUs have been sold by NVidia alone. In the US, we have about 30GW of data centers, almost all of them were already full. So in 3 years NVidia sold about a third of the total capacity of the country.

Those cards have nowhere to go to be powered on. They are sitting in warehouses right now. People have spent hundreds of millions on GPUs from NVidia and they are sitting in boxes doing nothing. Because there’s not 12GW of available data center space for them to go.

It took the US something like 30 years to get to 30GW of data center capacity. Even we assume doubling every 10 years, it’s going to take another 7 years to get the next 30GW.

And that means building a shit ton of data centers.

You can’t disagree or agree with something you don’t understand. You’re just saying words that make you feel safe and refuse to engage in the conversation

You are speaking from ignorance. It has nothing to do with what it is called and everything to do with what it is. You don’t know what it is, therefore you can’t form an analysis about what is going on nor can you even begin to understand what people are talking about. You don’t even understand what the uncanny valley is nor are you aware that the uncanny valley is an incredibly frought concept with many many studies completely unable to replicate the original research’s findings and significant debate in the scientific community about what it is and what causes it. There’s honestly no possible way for the problem to be a problem of the uncanny valley because the problem of the uncanny valley is far too problematic to be explanatory for the behavior that you are observing but can’t even understand.

So this is the other reason people hate a.i… It’s because people like you who are totally ignorant of how the technology works and how the economics work, when presented with facts, hand wave your own ignorance away as though it’s irrelevant and then snap back to your original position which is that the critique of a.i. comes from some emotional or mental deficiency on behalf of the criticizers when you can’t even understand their criticism because of your ignorance.

The antibiotic you’re talking about had absolutely nothing to do with LLMs. You don’t understand what a.i. is and you’re lumping everything together because of your ignorance.

LLMs are only used for natural language processing. That has a lot of potential, but none of the things you mentioned are natural language processing (except processing linguistic surveillance and gathering linguistic data). Everything else you listed has literally nothing to do with Claude, ChatGPT, Gemini, etc. LLMs are not the only a.i. in the world and just because something uses a.i. doesn’t mean it’s using an LLM

I’m sorry. Are you trying to say that disagreeing that a.i. is conscious is the equivalent of hating a.i.? a.i. isn’t conscious. It’s a static set of numbers and an algorithm to do math against those numbers.

Ecofascist identified

You’re being way too broad with your use of the word. a.i. means a lot of things that are very clearly not dead ends. Text generating LLM technology is a dead end, logically speaking, because it consumes the same stuff it produces. That’s fundamentally the definition of a spiral, and the last couple years of major improvements to LLM services have come from adding “skills” and “reasoning”, not from model evolution.

a.i. being publicly owned makes a ton of sense, but it just doesn’t financially work. The industry is completely and deeply in the red. There are no paths to fixing this that have been identified.

Before web browsers we basically had 2 other renderings: terminal text and email readers. Prior to the WWW, most content you could consume on internet was text only email and text only forums. These formats did allow for attachments, though, so you could get images, audio, and video that way.

Yeah, you’re just wrong. Lemmy hates the a.i. business, not convolutional neural network transformer technologies.

Most of lemmy is fully aware the a.i. does incredible work already in protein discovery, in cancer research, in climatology, etc. But LLMs, the stochastic parrots they are, have shown that their cost/benefit ratio is ridiculous.

As Ed Zitron said: if your business model is selling forty dollars for 1 dollar, you don’t actually have a business model. That’s the current state of LLM services today. Everyone that is willing to pay for LLM service today is currently paying a dollar and getting back about $10 - $100 of compute in return. That’s not economically sustainable, there are no theories for, nor indications of, costs coming down for providers, and consumers are already saying it costs too much even at these prices.

And this problem has gotten worse, not better, in the last 5 years. Costs are going up by multiples for the providers and in multiple areas. To make LLM responses more useful they now have to make the systems run dozens of inferences in parallel and select among them. The cost of fiber optics, of GPUs, of RAM, of storage, of electricity, of generators, of data center leases, of data center construction, and many other inputs has gone through the roof with some inputs jumping 300% and other inputs locking in price increases for a decade-long contract.

There are lots of problems with LLMs, but Lemmy has a much more mixed response with open-weight LLMs or ethically sourced LLMs than it does with the business of LLMs. Because the business of LLMs is very clearly a failed business currently sucking up our entire economy.

Western imperialists and their vassals held a fake trial 10 years ago in attempt to say “nuh uh! You have to let our military create national security threats for your country because it’s not fair! If you stop us, how are we going to win?!”

:nelson-laugh:

They left it out most likely because it is irrelevant and puts the victim at greater risk of victim blaming

I didn’t know Platner was a cardinal

The American model of free speech is anything but. It’s a narrative cover over reality.

The US government has always and will always choose what speech is unacceptable and repress it. They have been doing so for centuries. They have killed so many people for speaking. They have entire departments dedicated to media manipulation and censorship. They have vast financial incentives organized by the government to limit speech. There have been at least 2 open purges of communists and a third purge is burgeoning as we speak.

The American model is not actually that the government should not have the power to limit speech but rather that the government will limit violent speech if it goes against the government’s goals and will protect violent speech if it supports the government’s goals. That’s why white supremacist hate speech is protected in the US - not because America is a shining beacon of free speech but because America is shining beacon of white supremacy.