asura

The next step is China will inevitably lead with better LLM hardware since the US shot itself to death via boomer - and we will be data center detroit and end up the same way the auto industry did, but doing a large transfer of wealth beforehand.

As the models get bigger we still don't have any meaningful improvements - but we do have a bigger Nvidia bill.

Smaller companies are doing miles with smaller models at a fraction of the requirements (and nvidia bill). Matter of time before specialized AI hardware overtakes nvidia - it's a race

If there is a such thing as AGI it will require a fundamental architecture change and another transformative (no pun intended for transformers) technology to get there.

It's a scam. OpenAI was supposed to be open from the start and claimed AI was "too dangerous" (unless you paid for it)

I'm not sure you understand that I do understand all of that :)

People wanted, demanded something to do with all the GPUs and transformers were the answer. Nvidia is very loudly running the lead on that in spite of balancing on the razor thin advantage they have until actual AI processors catch up

Probably a good time to remind everyone that Bitcoin became unprofitable for the majority of people (including data centers) in 2018
2018 was the same year transformers showed a lot of progress, one year after the Attention [...] paper
2018 was when OpenAI promised *open ai*
2022 was ChatGPT

Quite a lot of this was already in place and the easiest way for smaller data centers to make cash was with crypto rentals - which shifted right over to LLM workloads. Google for quite a while was using 8 bit processors.

The new "AI data centers" are scams. :|

And is taking a loud boost from LLMs

Serious researchers (think like Google and IBM) had quite a lot of this in place already. Very unserious delusional divorced dads on ketamine did not - which is why they went with Methane even though he owns a solar company...

The original use for the NVIDIA GPGPU/CUDA push was literally what we're doing today. It was just not very good. The same thing that makes CUDA good for LLMs is the same thing that makes it great for protein folding and crypto - simultaneous matrix math.

Nvidia quietly took a boost from crypto.

People are being very obvious about using the phrase "feeling unsafe" in place of "being super racist" lately

I mean really at this point we know that's nonsense. The intentionally delivered meme of "AI can do all the work" is just to cover up the crippling of every American system so a wealth transfer can happen.

All of Anthropic can't even port an electron app to a linux desktop. Kind of telling.

tbh when I first read this I interpreted it as Chronic[ally gives stupid] Inflammatory Response[s to everything] Syndrome

I'll give a little respect for the guy if he applies his evolutionary psychology bs to himself, says evolution has predetermined through genetic disorder that men are useless.

join us join us join us

Depending on the provider, those data centers have agreements to reduce power during peak or emergency events (like hot/cold weather) and are closed loop water.

Most things labelled "AI Data center" is a crypto farm trying to break even and destroying one just lets them rebuild with new hardware. 🙃

Wow. So uh the post you quoted was completely insane LARPing. Impressive.

But you are completely correct. As of right now AI workloads require specialized hardware split across clusters across containers, etc and sitting right next to healthcare organizations, pornhub, steam, and bluesky all alike.

Midjourney Accuses Disney of Trying ‘Have It Both Ways’ with AI Midjourney is claiming 'fair use.'

Midjourney responds with training on Disney's materials as fair use.

If this was a couple of years ago, people would be throwing middle fingers at Disney... but we are in a weird spot where they've been trained that AI is stealing their grandma's meatloaf recipe.

Step this up and ask if revenge porn of strangers, faking white genocide talking points, and provocative content of ambiguously aged characters is "patently offensive and lacks serious artistic value"

Then ask why you can use Mastercard with twitter and Microsoft Azure that hosts Grok

$5 says that for "reasons" watching any Ben Shapiro and JRE videos automatically slots you into age exempt status but watching one Guilty Gear video throws you into verification.

"Learning ChatGPT" means literally nothing.
To your point, if you want to learn AI tooling and learn about prompts well... that has nothing to do with ChatGPT. There are a couple dozen better options to start with and dozens that run locally :)

Video shows Department of Justice official urging Jan. 6 rioters to 'kill' cops The Department of Justice hired a former Jan. 6 defendant who was caught on tape urging rioters to "kill" police. The department calls him a "valued member" of the administration.

"Less than five years after urging rioters to "kill" police at the Capitol, a former Jan. 6 defendant is working as a senior adviser for the Department of Justice [...]"

I sure do miss when every day wasn't embarrassing
Why is it that everyone says "meritocracy" looks like this upward failing fool

ink & quill
Don't know why I thought "ink a quill" made sense :P

Analogy: You should use ink a quill because a keyboard is bad for you, ... but then the very next year claiming wireless keyboards are the best. :P

Nothing changed in the presentation between those two technologies. Very silly.

Yeah, as much as I hate it... even individually, I will certainly remember less if I type it out than write it. I don't think I own a pencil.

Actual pedagogical data aside, I have no idea how these boomers say AI will teach the kids with a straight face after claiming ebooks and zoom were unusable.

The people inventing AI hate humans and want us to cheerfully and gratefully enslave ourselves to machines (and become petulant whiny babies when anyone tries to resist, which seems to be the going culture of US fascism)

As an experiment, I replied to someone on LinkedIn who was pitching "You must learn chatgpt ... blah blah blah" with asking why they said chatgpt instead of Allen Institute, Gemma, Distillation, Pydantic AI, ...

They had no idea what those words were.
Moral of Story: Marketing machine hits hard.

This has been bothering me.
To be clear I'm very much a Kindle person, but the same people screaming children can only learn from printed books and that Zoom ruined education....

Are now pitching *checks notes* Linda McMahon's Education with A-One [she meant AI] 🥁

Let this be a kick off to open algorithms and open reporting.
Where's that algorithm, Elon? The one that was going to be open and published right before he cleaned out all of the bots?

@RationalDreamerGeniusTruthLibertyAmerican2024 and the rest of the twitter gang forgot about that.

Unsurprisingly a cobbled together piece of junk that's nothing more than a vanity project meant to fail - like everything Musk

Pretty transparent. Ask Musk to fold proteins. Sorry, they can only ship anime waifus and revenge porn with (allegedly) the "world's most powerful AI"

DeepMind doing fine

Yeah, don't use grok
A depressed divorcee on ketamine that owns a failed solar company but is poisoning people in Memphis with gas generators is not a feature of any specific AI related technology.

That's a feature of a very specific fool.

This is an opportune time for organizations to widely and loudly state that since RFK Jr. says trust is not a feature of democracy or science then nothing he ever says or does is trustworthy, rational, or truthful.

He is no more than a hare brained idiot practicing magical thinking by admission.

From 2020 to 2022 COVID vaccines prevented over 18.5 million hospitalizations, 3.2 million deaths and saved the US $1.15 trillion.

mRNA vaccines and treatments are being developed for diabetes, HIV, cancer, and more. Imagine how many more lives we could save if we don't defund science.