Comments

But even hard evidence doesnt matter if the institutions are corrupt, and the media censors whats happening. Just imagine how many they could actually kill before people would react. I think it would be quite a lot. And at the same time, they are camouflaging the declining population with immigrants

True. It’s definitely an addiction.

Out of curiosity, there was a thread a while back about unresolved MAC addresses showing up all over the place after the “vaccine” rollouts. That line of inquiry seemed to just go dead. Any developments since?

The link https://eu-images.contentstack.com/v3/assets/blt6b0f74e5591baa03/blt790f1b7ac4e04301/6543ff50fc links to the error: {"error_message":"Asset was not found.","error_code":145,"errors":{"uid":["is not valid."]}}. Nevertheless, https://huggingface.co/deepseek-ai/DeepSeek-R1 appears fully active. Lots to check out and learn. Thanks.

So, a couple hundred Gb RAM, a few (8-24) Tb of SSD, ~24+ CPU cores and a modern NVIDIA GPU or two could run a local AI as good as online ChatGPT o4-mini? Can the entire model database data literally be stored locally, or does that still need to be in “the cloud”?

Yeah, you can often get around all safeties by asking it to explain its logic or describing how one would do what it is programmed to not be allowed to do. I waffle between being intrigued and concerned.

I use Elicit for scientific references and ChatGPT o4 mini for coding questions. o4 mini can identify required libraries and dependencies and nail syntax in seconds compared to hours of trial and error, documentation reading and web searches.

Exactly what I have now. RTX-4080, 4TB SSD, and a dozen or so 2TB USB-C SSDs, but they fill up with data fast too. Storage is getting almost too cheap. The tariffs could change that I suppose. Interesting. I will have to check this out. Thank you.

That’s true, but it certainly never required Joe Rogan to see it.

I didn’t see this coming either and I am and have been at the forefront of IT. I must admit I did leave the IT world for a while for boredoms sake and seeing it all but as soon as AI has hit the streets, I am full on back into this. I have coded entire Android apps before that "mostly" work with very little tweaking. Some of the later models have figured out mathematical puzzles that have never been solved and now they are blowing past all of this. If you are looking for a special time in humanity, this is definitely it. Unfortunately it does make humans obsolete because we gotta sleep and regardless of how well we do something, there is always the human error side. And as far as a multi-project solution, its AI platforms like Manus that will blow your mind. You can ask for every information about a subject and it will spend an hour rolling through the Internet grabbing all the data and compiling it and presenting it to you. The AI Agents will be a part of everyones life. At present they are a bit complex to get going but AI Agents are essentially a bunch of AI tech put together that work together that destroy teams of people. Its why Duo Lingo fired all their staff and just use AI. AI is here and there is no way to stop it...as I said, it doesn’t sleep and every morning there is something crazy from one of the thousands of companies building things. Adapt or die!

The covid “vaccinations” may be the currently closest-to-home, most recent and most egregious example of megalomaniacs trying to destroy their self-proclaimed “subjects” through psychological coercion and torture, but it is definitely not the first (think Inquisition) nor the last. They will keep trying. We who did not buckle under the pressure of lost jobs or blocked travel plans never really buckle to their mind games, even when forced against our wills by shear physical force. And we will not roll over next time either. Our minds are our own. Some people are just less susceptible to non-beneficial indoctrination. They try to squash that by sending us to the “alternative schools”. Nevertheless, some of us still come out with PhDs, law degrees, contracting companies and independent networks who continue to dispel the BS they are shoveling. Indoctrination cannot completely incapacitate everybody.

Actually, the new AI systems are all very good at just this. For instance back to the zebra, if you ask it for what a zebra is and it spits out a duck, it will give itself a bad grade and send it back to the thinking tank. So now they are all rewarded with a success and beat down when they get it wrong. And in the terms of AI, ChatGPT is pretty cool and definitely saves time but the open source modules that are launching daily beat the snot out of it. DeepSeek just released a brand new reasoning model that is open source that I havent play with yet but it does beat most of the paid tier models that are available out there. All of the new reasoning models are making the old models that need human input obsolete so fast. Every single day is like 10 years when it comes to what tech time used to be. My day begins with watching a few guys who are presenting the latest AI and showcasing the new stuff. Then if it trips my trigger I will go install it and see what the limitations are. At some level they have coded these things to not let you be fully evil but if you tell the AI that you want a hypothetical story line on how to build a complex bomb, it will give you everything you need to know.

So the I is the true intelligence and why AI is so impressive. It is all based on a brains neural network. It looks like a zillion little dots. https://eu-images.contentstack.com/v3/assets/blt6b0f74e5591baa03/blt790f1b7ac4e04301/6543ff50fcf447040a6b8dc7/News_Image_(47).pngrnrnNow older AI models of these LLM (large language models) had to be trained by humans which is what ChatGPT did with billions and billions of dollars. The new AI is able to train itself with a reward system. So when you look at that graphic the input is say on the left hand side of exactly what you are after and it will run through 200 different nodes before it will make it out the right hand side with your answer. So it literally is thinking through things. Unlike a traditional database like MySQL or Postgres that has the data in there and you can run a query and get an answer, say you put a picture of a zebra in there and ask what this is, it will run through a tremendous amount of thinking and eliminate everything its not as it knows what a zebra is but there are a thousand data points for a single zebra. Stripes, hoofs, tails, nose, eyes, where are the eyes, where are the ears and as it runs through millions of nodes in the neural network by the time it gets to the end it will be for sure to answer you that its a zebra. Now back in the days (months ago in AI time) there was a major project with AI Hallucinations where the AI would go completely whacked and spit out something completely bonkers because you can literally throw 1 piece of bad data into a data set and everything will change. But nowadays most of the hallucinations have been resolved. So in terms of does this software think, yes it absolutely does. Now this is where things get interesting. https://huggingface.co/models - New models are emerging every single day and they are getting better and better. Some are math modules, some are full on coding modules, some are medical, some are based on just simple niches and the coding is immaculate. Its is precise, fast, and you can have an entire website coded up in about 30 seconds. There are endless things that these models do and the one thing they do is everything better than humans. At present, because of how they are built, if you are a leader in some edgey code that is extremely creative, the AI can’t get as creative as a group of really smart people. But...those days are coming to an end and the human race simply cant keep up with the perfection of a LLM. For those that can figure out how to install deepseek or another model, I would suggest you do so you know whats up. That being said, if you get stuck, I am not here for support. The directions are endless online. Once you have deepseek installed you can actually download any of these models from huggingface and they load up the exact same way. You can only have 1 model running at a time. For most of my life I wanted to keep a copy of wikipedia for the day when the powers that be limit the data but AI makes wikipedia be the weakest of the data out there. You can calculate anything, build anything, understand anything and the AI will explain everything to you.

I remember watching Jim Breuer and laughing, but in the end, ridicule has solved few problems, but helped create a whole lot more.

Actually, that’s not a bad description. Get the “unwashed masses” killing each other. Billionaires don’t care about your skin color. You’re using their air and they want you gone

Australia and New Zealand (along with Canada) sure were at the front of the line for medical tyranny-facilitated consolidation of centralized world “power”. Her level of smirking disingenuousness is almost unprecedented.

So this is a simple RTX 4080 GPU, single computer with 24 cores and 200 gigs of ram. The SSD is the only sucky part as I only have 1 TB on it and I fill it up completely with the models I am using. I am planning on getting a 4TB drive in the future and then I can really have some fun. https://huggingface.co/deepseek-ai/DeepSeek-R1 You can easily roll a full TB the 32 billion model. The higher the models you go with the slower the output but the output rolls reguardless. For instance with the higher dataset it rolls out what would look like a typewriter typing response.

Three years ago I Would have told you that AI couldn’t get a complicated computational routine right though. Boy was I wrong.

DeepSeek looks interesting. What GPU(s) do you use, and do you have to use a series of linked processors? How much SSD space is involved?

Like a radio or a TV with improved reception?