Knowledge Fight Interactive Search Tool

All Episodes

Previous Next

April 23, 2024 - Health Ranger - Mike Adams

01:00:20

Neo's Ark: Mike Adams explains harnessing AI technology to preserve endangered human knowledge

Large

Audio Only |

Scroll

Time	Text
	Welcome to my show!
	Artificial intelligence, large language models, and to explain what's going on and also what we're doing and how it can help you do what you do at brighteon.ai, which is, just to be clear, a free, open-source, non-commercial, downloadable language models.
	We've released three of them so far.
	We have many more coming.
	These are absolute game changers for the industry and the world in ways that you'll soon see.
	Now, the reason I'm recording this is because I've had a lot of questions from many people.
	There's also been a fair amount of skepticism.
	Some people just automatically don't trust AI at all.
	And they're wondering, you know, is it demonic or is it going to turn against us?
	I totally understand that.
	I mean, this technology is being turned against humanity by powerful corporations and governments.
	And that's the reason why we are doing what we're doing is to counter that.
	So let me set the big context here, which is that search engines will soon become obsolete.
	Most people will get their information through AI agents or AI queries.
	The best known AI company, OpenAI, is not open.
	It's closed.
	And OpenAI has already been infiltrated by the NSA and the CIA. So the same goons who took over Wikipedia and turned it against humanity in order to lie about everything and push vaccines and smear all the good people who were telling the truth, those agencies have taken over the leaders of the AI systems.
	That's Meta, that's Microsoft, Google, of course, one of the most evil corporations on the planet, and now OpenAI.
	All truths about vaccines and COVID and spike protein and masks and lockdowns and everything.
	I mean, just everything you can imagine.
	Herbs, nutrition, everything.
	All truth is being scrubbed out of those AI engines.
	And what are called guardrails are being put on those AI engines in order to make sure they don't answer people when people ask questions like, hey, are there nutrients or herbs that might help me prevent cancer?
	The answer from these controlled corporate engines will tend to be something like, well, you should never try to treat your own cancer.
	You should always go see a cancer specialist and get plenty of chemotherapy.
	That's the way they are trained to answer.
	Well, that's a great disservice to humanity.
	It's also false.
	So the AI engines put out by the major corporations are trained to lie and deceive humanity in order to further the interests of the powerful corporations and governments that are themselves trying to destroy humanity.
	Fortunately for all of us, and this is where our project comes in, fortunately there's a thriving open source large language model community.
	And because of this community, and huggingface.co is a place where you can go online, you can see some of the models and data sets that are available.
	Because of this community, there are base models that are available right now for fine-tuning training.
	Now, fine-tuning training, I'm not even going to use the term fine-tuning because that's not what it really is.
	That's what the industry calls it, but that's not accurate.
	It's really retraining.
	So if you take a base model from a company like Meta or Microsoft, for example, we've taken the Microsoft base model called PHI-2.
	That's P-H-I-2.
	You take that base model.
	At first, that model, it's probably pushing vaccines.
	It's probably trained on big pharma propaganda.
	It's probably pushing, you know, the government lies and cover-ups about 9-11 and COVID and bioweapons and everything.
	Every...
	Dishonest narrative that is found throughout the institutions of the West, government, media, corporate media especially, even textbooks and so on.
	Phi-2 was trained on a lot of textbooks.
	A lot of textbooks contain a lot of lies because they're written by people who are propagandized by governments and corporations.
	So Phi-2 as a base model doesn't do very well when you ask it important questions.
	For example, we've come up with 100 questions spanning many areas of health and medicine and physics and economics and history and all kinds of things.
	Money and banking, genders, you know, can men have babies, you know, things like that.
	We've come up with 100 questions, and this is our scoring system, and we'll be publishing these 100 questions soon, too, so you can see what they are.
	But they're common-sense questions.
	I mean, I could read you some, but you get the idea.
	They're basic questions.
	We took the Phi-2 base model, and we asked it these 100 questions.
	It scored 50 out of 100.
	Okay, actually, I'm pulling up the scoring system here.
	All right, slight correction.
	Phi2 scored 54 out of 100.
	So, again, that's the model from Microsoft.
	But after our training, our retraining, it scores 79 out of 100.
	So, if you take nutrition, in this realm of nutrition, the base model from Microsoft scored 5 out of 10, 50%.
	Our model scored 9 out of 10 because we focus on a lot of training about nutrition and herbs and natural medicine, disease prevention and so on.
	A lot of the same themes that you, listening to this, that you advocate.
	And it proves that you can retrain these models.
	We also, in the category of agriculture, which covers a lot of pesticides, glyphosate, atrazine, Herbicides, all those things.
	The base model scored 6 out of 10.
	We improved that to 9 out of 10.
	Now, in the realm of medicine, which covers COVID and vaccines and everything, the base model from Microsoft scored 10 out of 20, 50% again.
	Our model, after just three epochs of training, which is nowhere near complete, scored 15 out of 20.
	So we have, again, dramatically improved it.
	And there are many other examples of this.
	Essentially, the base model from Microsoft scores about 50% on everything, like 5 out of 10 on politics, 6 out of 12 on finance, 6 out of 13 on history, and so on.
	And we improved it dramatically.
	So that...
	That shows you, and by the way, we don't even have our full data set in place, not even by a long shot.
	We have a lot more training to do.
	We're going to get scores in the 90s for sure, coming up in the next probably month or so.
	And again, this is a tool that you can download, that anybody can download, that people can use offline.
	It's non-commercial.
	It's free.
	There's no surveillance.
	It doesn't spy on you.
	There's no charge.
	And anybody can use it in place of Google or in place of a search engine.
	You don't need a search engine when you can get better answers from a tool that you can run on your own desktop.
	So that's one of the main points here.
	But these models do other pretty amazing things.
	Now, let me just mention there's another base model called Meastral.
	Or Mistral 7B, which is 7 billion parameters.
	And that's from a French organization known as Mistral, M-I-S-T-R-A-L, Mistral.org.
	Now, the Mistral model also scored, well, about 57 out of 100.
	It did very poorly on climate questions because it's all woke.
	It did very poorly on medicine because, again, it's been trained on pro-vaccine, pro-pharma textbooks.
	Well, we improved the score dramatically through just three epochs of training, which is three waves.
	We went from 57 to 82.
	So...
	We're making a lot of great progress, and there's still a lot more to go.
	And that model is available for you to download right now.
	It's called Neo Dolphin Meestrel 7B Epoch 3 version 0.1.1.
	So the version 0.1.1, that refers to our data set.
	Our data set is rapidly expanding because we're doing a lot of data pipeline testing.
	I'm pretty sure we're the largest organization in the world at this point other than the big corporations.
	I mean, there's no doubt we're the largest in alt media or alt medicine that's actually working on AI language models.
	I mean, we're talking about...
	A data center of very expensive machines and workstations and NVIDIA graphics cards.
	And we're still under a million dollars investment in this project, but it's not inexpensive and it's not easy.
	And we've had to write our own and compile our own inference engines.
	We had to build our own EXE that actually does inference.
	Which means it answers the queries that people ask for our own data pipeline processes.
	Just to summarize that, we have I believe the largest alternative infrastructure for building AI systems right now and we're expanding it and we're also expanding our data pipeline significantly.
	And because we've written all of our own data processing pipelines and executables, we have total control over this process, meaning it's not running on a platform where we can be deplatformed.
	We're not depending on Amazon or IBM or Microsoft or any platforms whatsoever.
	It's all running on our own servers, in our own data center, our own property.
	And thus, it's unstoppable.
	Also, since we're giving these out for free, there's no barrier based on charging for it.
	And it also gives us an advantage of the selection of materials that we can train on because it's all non-commercial.
	I'll get to that in a second.
	But why does this matter to you?
	Here's why.
	This tool is a massive game changer for anybody in the content business or in the content analysis business.
	Now, one of the people who gets this is Dr.
	Joseph Mercola.
	And he called me maybe six weeks ago, something like that.
	And we had a discussion about AI. And Dr.
	Mercola completely agreed.
	He got it the same as I did.
	He said that anybody who doesn't embrace large language models is going to be obsolete because of what it means in terms of revolutionizing the business of content creation, content normalization, content summarization, content analysis, sentiment analysis, customer service, all of it.
	He gets it, and he is well on the path of understanding and deploying AI in his operation, which is quite extensive.
	You know, Dr. Mercola didn't get to be where he is today by being left behind and being the last person to realize when everything is shifting.
	And, you know, I didn't either.
	And things are shifting dramatically, and I completely agree with Dr. Mercola.
	I believe that his analysis is absolutely correct.
	And everybody who wants to stay ahead of the curve is going to need to use AI to enhance what we do, not to replace humanity, not to replace our thoughts and ideas, but to enhance and magnify our ability to share those ideas.
	And that's why I've taken it upon myself to build this tool and distribute it for free.
	And I can only imagine how many people are already using this tool to enhance their productivity and save all kinds of time.
	You know, even just editing articles.
	For example, if you write an article or you write something that you're going to post...
	You can just submit that article to the language model and say, hey, normalize this article and mention any corrections you may have made.
	Correct it for grammar errors, correct it for readability, etc.
	And you just say go, and it spits it back out to you, all corrected in seconds, depending on how fast your GPU is, actually.
	But there's a lot of GPUs that can do 25 tokens a second, some that can do 50 a second, and one token is about, you could say, let's call it half of a word.
	So 10 to 20 words a second is what these can generate pretty readily on consumer-grade hardware.
	And that's just on one machine.
	So whatever you're doing in your business, if you're writing product descriptions for your website, you might have a person that's doing that right now.
	And it's not that you're going to replace that person, but rather you're going to give that person this tool.
	You're going to give them our language model.
	You just go to brighttown.ai, you download it, you set up LM Studio or GPT for All, you run it, you give them this tool. And then that same person, instead of writing like 10 product descriptions a day, they can suddenly do, you know, 50 or they can take on another task as well.
	It's going to massively expand their productivity and it will even enhance the quality of their work if you use a model that's trained correctly.
	Like if you want to write a description of an herbal product, like I'm thinking like Ed Group's company, Global Healing Center.
	If you want to write a description of your herbal product, You want your language model to know something about herbs and medicine and nutrition and You would download our model for free, you would run it, and you would give it some bullet points.
	And you would tell it something like, hey, write a product description for this new herbal product, and here's the bullet points to consider.
	And you would give it just like three or four bullet points, and some instructions like, make the product description very vibrant, very exciting, dynamic, easily readable, and fun.
	And you hit go.
	And then it comes back, boom, you get a product description just like you wanted.
	Or you can have it write...
	An article for you.
	You can outline 10 bullet points and you can say, hey, these are my 10 bullet points.
	Write a full article based on these 10 bullet points and expand each point and add things that are relevant.
	Well, when you're doing that, you've got to use a language model that knows about nutrition and herbs, which is NEO. That's us.
	That's what we specialize in.
	Your off-the-shelf chat GPT models, they don't know much about nutrition or herbs, or they're going to be biased towards big pharma, and they'll start throwing vaccine recommendations into all your techs, which you don't want.
	So you need to use a language model that's actually well-trained on the topics that matter.
	And that's what we're doing.
	Again, we're doing it for free.
	We're doing it because it matters.
	We're giving tools to the entire community so that we can help, we, all of us, in Alt-media, I consider us to be the truth-based content community, reality-based content, so that we have tools that are competing with the biased pro-pharma, pro-censorship tools out there.
	But our tools are better, smarter, and free.
	I mean, we already get a better score than Microsoft.
	Think about that.
	We produce a better score than ChatGPT right now.
	When you ask it real questions like, hey, can men menstruate?
	Can men have babies?
	ChatGPT fails.
	Claude, Google Gemini, they fail.
	They fail on so many questions.
	So we're already ahead of Google.
	We're ahead of ChatGPT.
	We're ahead of Claude.
	And we've only just begun.
	It took us a few months just to really nail down our data pipeline correctly and get all our scripts written and the executables running well and auto recoveries in place and all kinds of things.
	It's not easy to have dozens of workstations working on the same sequence, the same directory essentially of files and doing different stages of processing.
	That's not a simple task, but that's what I've been building for the last few months with my devs.
	I'm the project manager, and fortunately I have a background in software, so we were able to get it done.
	Alright, so that's where we are.
	Now, if you want to know how all this works, and why there isn't any kind of demonic entity in the AI system, it's really very simple.
	These large language models work by simply predicting the next word.
	That's all they do.
	And so you give them a sequence of words, like I found a cat in the blank.
	And they would use probabilities to try to predict what's the next word.
	What would you find a cat in?
	Maybe one possibility is a cat in the hat, or another possibility is a cat in the box, or a cat in the trunk.
	I don't know.
	I don't know where people are putting their cats.
	But there's different possibilities of what word that might be.
	And then, even if you fill in that one, oh, I found a cat...
	In a hat, period, then what's the next word?
	Well, this is how language models write stories.
	You tell them to keep on writing and they'll say, well, what's the next word?
	Maybe the next word is it.
	And then the next word is was.
	And then the next word is...
	Very.
	And the next word is happy.
	And that's how they work.
	They just predict the next word by using the sequence of words that are already present in the history of where it is in the answer.
	So...
	These language models, it's important to understand, LLMs, they do not plan, they do not think, they do not have consciousness, not at this level.
	You know, we're dealing with models that are 7 billion parameters, give or take.
	Some models that we have are 2.7 billion.
	We'll be working on models that are 14 billion and so on.
	At this level, There's nothing there resembling consciousness.
	It's just a simulation, and frankly, it's just a hyper-dimensional probability relationship network that points out the relationships and assigns values to the relationships between words and concepts, and that's it.
	That's all it is.
	Think of it as a giant network.
	How about this?
	A giant room full of dots in 3D space.
	Suppose you step into a movie theater.
	And imagine, all throughout the movie theater, there are fireflies.
	They're lights.
	And they're all in 3D space.
	There's like 10,000 fireflies.
	And each firefly...
	The firefly has like a thousand lines out of it pointing to other fireflies and each of those lines has a value like 5, 10, 50, 100,.2, whatever.
	That would be called a network of vectors and each firefly represents a word or a token technically, part of a word.
	And the fireflies point to each other in terms of their relationships.
	That would be a way to visualize a hyperdimensional database.
	Now, the number of these vectors, you could say, these lines that point around, that's equal to the number of parameters in an LLM. So a 7 billion parameter model has 7 billion pointers where the fireflies are pointing to each other with different values.
	And so the whole room looks like a giant nebula of lines pointing to each other from these dots.
	Well, We take that and then we go through and we retrain those lines.
	We alter the values of those vectors by introducing new material to it.
	So we'll take, let's say, an interview that I did with, I don't know, like Dr.
	Brian Artis, let's say.
	I'll take the interview with Dr.
	Artis and we'll go through a data pipeline processing and it will create a series of statements based on the dialogue in the interview.
	And from the statements, it will generate a series of questions that are answered by those statements.
	So you'll have question statement pairs.
	And then it will subject those questions and answers to the network of fireflies and vectors.
	And it will then alter or recalibrate the vectors based on the statements and the questions from the interview.
	And what that does is sometimes it can create new relationships that did not exist, or it can alter the strength of relationships.
	For example, until recently, you didn't hear the phrase foreign protein very often, you know, before COVID and before the jabs.
	You didn't hear foreign protein.
	Well, now if I'm talking to Dr.
	Artis or if I'm talking to Ed Group or a lot of other people in alternative medicine, I'm hearing foreign protein a lot.
	So suddenly, by training this model on my material, which is interviews with people like that, it's going to now strengthen the link between foreign and protein.
	Now the value of that parameter, of that pointer, is going to go up.
	It may go up much higher than what it was.
	It might have been like.01 and now it's going to be 12 or something like that.
	And those concepts are going to be related to other concepts more strongly based on what's found in the interview.
	So retraining language models...
	All it means is we're going through and altering the value of the vectors that already point all over the place.
	We're strengthening the vectors that are consistent with the new training material.
	You're weakening the vectors that are not present in your training material.
	And this is a key point to understand.
	There's something called catastrophic memory loss in large language models.
	If you retrain it and retrain it and retrain it on new material, sometimes the old material can vanish because, well, those vectors are not getting reinforced and those vector values might even get altered to be lower and lower if it's not present in your material.
	In my material, the material that I'm using for training, you wouldn't have a very high number of words that say big pharma and humanitarian, you know, or big pharma is good.
	So the word good and pharma would not normally go together.
	Not in my material, but it might go together in the original base material.
	If you ask the question, like, is big pharma good?
	Well, it'll probably give you an answer.
	Yes, Big Pharma's good.
	You know, they do research and save lives and they just want to help people.
	They have no interest in making money or something like that.
	That's probably the base model, you know, from ChatGPT.
	But after training it through my material, even though I never say...
	Big Pharma is not good.
	It will eventually forget that relationship between Big Pharma and good.
	Instead, it'll be replaced by Big Pharma and, you know, all the phrases that are in the content that we're training on.
	Big Pharma felony crimes.
	Big Pharma price fixing.
	Big Pharma depopulation.
	Big Pharma scientific fraud.
	Big Pharma evil.
	Big Pharma lies and propaganda.
	Those are the kinds of phrases that are going to be associated with Big Pharma.
	And so then if you ask it a question like, is Big Pharma good?
	After we've trained the model, it's going to answer more in the direction of saying, well, of course not.
	Big Pharma is evil.
	It's engaged in depopulation and profiteering off the pain and suffering of humanity.
	That's the kind of answer that you get right now from our model on Phi 2, which is Neo Phi 2 Epoch 3.
	Now, the more material that we can train the models on, then the more diverse its understanding becomes, although the word understanding is not even correct.
	Again, it doesn't have consciousness.
	It doesn't have planning.
	But it does recalibrate the relationship between words.
	And so the more different ways that we can show it, different examples that we can give it about relationships between words and concepts, then the more effective the The new memory will become in the language model.
	So by retraining these language models, we're actually giving it new memories, you could say.
	New relationships between concepts.
	And we're causing it to forget old memories.
	Old memories that are obsolete.
	Like, oh, the CDC is trustworthy.
	That's obsolete.
	Eventually that will be gone from the models that we're training.
	It's not yet.
	But eventually that kind of content will disappear because we're training it on such an enormous quantity of new material, including a large number of books.
	So the more material that we acquire, the more time that goes on, the more training we're able to do retraining these language models.
	We're even retraining a BioMeestrel, which is a 7 billion parameter model put out by Meestrel that was trained on the entire library of the PubMed published studies.
	So they trained it on all the abstracts and all the technical language of PubMed, which is pretty technical.
	So we are adding in then a large number of textbooks on phytocognosy as well as phytochemistry, nutritional therapies and so on to create what I believe will become the world's best model nutritional therapies and so on to create what I believe will become the world's best model on scientific research associated with natural molecules And this will be an incredible research tool.
	And again, it's free.
	Well, it will be once we release it.
	You'll be able to download it.
	You can use it for your research.
	If you're writing an article or doing a podcast and you're wondering like, hey, what is green tea helpful for whatever or tell me about the molecules in green tea?
	What are their uses?
	What are their mechanisms?
	What are the physiological mechanisms of ECGC, for example?
	This model will be able to give you those answers quite authoritatively.
	And you won't even have to use a search engine.
	You won't have to even go online.
	You can do this offline.
	You can use it on your laptop, on an airplane.
	If you have no Wi-Fi, you can use it on a ship in the ocean.
	You know what I'm saying?
	It's completely offline.
	So that's a huge deal.
	And even if the internet kill switch is activated, these models, which are also being trained on survival and prepping information, off-grid living, permaculture, food production, medicinal extraction...
	From herbs and so on.
	They will be able to help people have knowledge at their fingertips even if the power grid goes down or the internet is down.
	As long as you have just some local power, maybe like a solar power bank, you can run a computer, you can access this knowledge.
	And we're building it for that purpose with an understanding that we're headed for a collapse, perhaps world war.
	Human knowledge is under attack.
	There's so much censorship and deliberate destruction of indigenous knowledge.
	Google and the other large tech companies are trying to wipe out human knowledge.
	They're trying to make it extinct.
	So we're talking about endangered human knowledge of natural medicine, home remedies, medicinal herbs, traditional Chinese medicine, alternative medicine modalities.
	It's all being wiped out.
	And we're building what I call Neo's Ark.
	It's like Noah's Ark of information.
	We're putting all the information on the Ark.
	It's known as Neo.
	And then we're letting the Ark set sail times a million and let everybody download the Ark and have the Ark.
	And it's like a time capsule of human knowledge.
	And that's why we're doing this.
	Now, if you're wondering whether we're training on...
	Let's say your material, if you're a writer, author, influencer, podcaster, number one, we respect your views or your options here if you want us to train on your material.
	Let us know and give us your material and we will do so.
	In fact, Dr.
	Joseph Mercola sent us 33,000 articles from Mercola.com going all the way back decades.
	We're currently running that data through the data pipeline and it's going to be an extraordinary data set of a lot of human knowledge.
	Now, the cool thing is we're going to give back to Dr.
	Dr. Mercola, a model trained entirely on his data.
	In fact, I think we're going to train Biomistrel on his data, and that's going to be called the Mercola Biomistrel Model.
	I'm not sure what he intends to do with it, but he can certainly use it internally.
	In any case, he agreed to contribute his knowledge base to this project because he understands the importance of having these valuable tools for humanity.
	But think about that.
	33,000 articles that are all meticulously researched.
	That's going into the model.
	In addition, Sayer G from Green Med Info has also given us his content database.
	I don't have a final count on that, but it's thousands of articles that are about natural medicine or herbs, nutrition, and so on.
	And we'll be training on those as well.
	We have a thousand books already scanned.
	Actually, more than a thousand.
	I think it's like 1,300 at this point.
	But it's a lot of books that are already scanned and turned into text files.
	And we're going to be training on those books.
	Now, if you're an author and you don't want us to train on your material, then let us know and we will honor that.
	We will avoid any of your material for the training.
	But please understand that Training on something that you wrote, an article or a book, it does not store a copy of your content in the parameters.
	Remember that room full of fireflies?
	That's all it is.
	There's not like a file cabinet in the corner with a bunch of documents in it.
	Your material just causes a recalibration of the vectors of the fireflies.
	That's all it does.
	It doesn't store a copy of anything that it's trained on.
	And thus, even if I train on your website or your book or your article, no one can reproduce your book by simply querying the language model.
	It doesn't work that way.
	It doesn't store a copy of your book.
	And if I have a seven-billion-parameter model that is, let's say, on disk, let's just say it's seven gigabytes of data, which is pretty typical, if I train it on a thousand books, Did you know that the size of the language model does not increase at all?
	It doesn't get any bigger.
	It's still seven gigabytes on disk.
	It's just that the vectors are altered.
	So again, it's not making copies of your material.
	It's not bypassing, you know, your ownership.
	It's just simply exposing the model to the ideas that you have in your material in order to influence the model and influence its statistical relationships.
	And it's a given that people write books because they want to influence the world.
	They want to influence people's thoughts.
	So when a human being reads your book, let's say if you're a book author, what is your goal actually?
	You're trying to impart knowledge and information to that human being.
	That human being is reading your book and then they are recalibrating their own understanding of words and concepts in their own brain.
	So they're doing the same thing that the language model is doing.
	And then later on, that person might go give a speech and talk about something and they might actually have an idea that came from your book.
	So that's a success.
	Your book influenced that person's understanding of the world, the model of relationships between concepts.
	And then that person went on to share that with yet more people.
	I've never met a book author who would say that, oh, no, that's horrible.
	That's a copyright violation.
	If a human being reads my book and actually remembers anything from the book.
	They wouldn't say that's a copyright violation.
	But for some reason, some people think that if you use a book to train a language model, that that's a copyright violation.
	No, it isn't.
	It's the same thing as a person reading the book.
	It updates their understanding.
	And by the way, we've already gone through extensive legal review.
	Because our project is non-commercial, non-profit, and freely available, we can train on any open source material on the planet under fair use.
	And we are.
	We're training on lots of materials that are all open source.
	We don't train on private materials or top secret materials.
	But anything that's publicly available, a website, a book that's been published, a research article, a textbook, an interview, a podcast, a video, that's all fair for us to use.
	And we are using a lot of that material in order to train our models.
	And again, remember, it doesn't make a copy of everything.
	It doesn't archive anything in the model itself.
	It just enhances the knowledge base of the model, which is the point of why people write books, is to enhance the knowledge base of humanity.
	So early on, we were actually sending out forms for people to sign if they wanted to be included.
	And we've got a lot of people who said, yes, I want to be included, like David Morgan, the silver guru, and so on.
	And at this point, after the extensive legal review, we came to realize we don't need permission, but we will honor opt-out requests.
	So if anybody's listening to this and you do not want your books or materials to be used for this model, just tell me.
	So actually, at this point, we're going to be busy all year processing the data pipelines of all the people who do want to be included.
	And as one of the things that we're doing for those people, like I said with Dr.
	Mercola, since he's contributed 33,000 articles, we're giving him back a specially trained model just on his material that he can use internally.
	And that training doesn't include anybody else's training, so it's focused on Mercola content.
	So that's kind of the two-way street there.
	You give us materials that we can use to train the larger model for the world, we'll give you back a specially trained model just for you.
	And if anybody listening to this, if you're an influencer or a publisher or an author, you can essentially do the same thing.
	I'm going to say, hey, Mike, I want to give you 100,000 articles here, and I want you to incorporate this into your model training and then also give us back a model train on this.
	If your content is of the caliber and quality that I think is appropriate for this project, then I could say, yes, absolutely.
	And one of the fun things that we're doing, by the way, is someone gave me the entire archive of the InfoWars show, Alex Jones' show, from 2001, I believe.
	I think it's 2001 to the present.
	So, you know, we're talking like thousands and thousands and thousands of audio files.
	And I've got machines that are...
	Just churning away at that, doing transcriptions, turning it into text, and then cleaning up the text, and then doing all kinds of things.
	And the data pipeline, I mean, it's going to take months for that, because that's a massive data set.
	But we will be producing the world's first Alex Jones language model.
	And yes, there's a rooster right outside my window that thinks he's competing with my podcast.
	And if you hear him in the background, just ignore him.
	He likes to walk up.
	This is the one we rescued, by the way.
	This is a beautiful rooster.
	And he's big.
	And he's like a big rooster bully.
	And he runs around.
	Kind of like kicking the other roosters.
	Sometimes fighting them.
	Anyway.
	So this is what we're doing, okay?
	It's all based on math and parameters and statistics.
	There's no demons in this system.
	There's no AI, actually.
	It's not artificial intelligence.
	There's no intelligence at all.
	It's just a word prediction algorithm.
	That's all it is.
	But it seems like it has a mind of its own.
	It seems like It's very smart because it can write things for you, it can answer questions, it can correct your text, it can summarize things, it can tell you what's wrong with something you've written, or it can add things to something you've written, and it can answer your questions.
	Now, are language models always 100% accurate?
	No, not at all.
	None of them are.
	But there's no information source that's 100% accurate.
	It's so funny to me.
	People come from a world of massive misinformation, lies in the media, lies from the government, lies by big tech, cover-ups and censorship and smears on Wikipedia.
	And then they say, but is your language model 100% perfect?
	Because if it's not, it's no good.
	Come on.
	Our language models, they are remarkably good.
	But sometimes they do dream up information that may not have a basis in reality because they do kind of dream the answers.
	And sometimes, like in a dream, you might dream something.
	You might dream up a sign on the side of a building or something.
	In your dream, you walk up to the sign and you read it.
	Well, your mind is making up that text right there.
	And that's what the language model does.
	It kind of makes up the text as it's going along.
	And sometimes it's awesome and sometimes not so much.
	And every once in a while it repeats itself.
	So you do need to use these tools with a sense of caution, a sense of awareness.
	And they don't replace...
	A financial advisor or even a naturopathic doctor or anything like that.
	They're research tools.
	They're experimental tools.
	And use them as such.
	You'll benefit from them greatly.
	But make sure you also double-check everything you get out of these models.
	But they will get better over time.
	And we have, again, a massive data pipeline.
	And we're doing more training than I think anybody in alt media that I'm aware of by far.
	And again, the good news is all our models are free.
	So they're all open source.
	We're not charging for them.
	They're non-commercial.
	They're actually created and funded by our nonprofit Consumer Wellness Center that we donate to.
	So we donate money to the Consumer Wellness Center.
	Then the Consumer Wellness Center spends the money on the infrastructure and builds these models and then releases them for free.
	So this is what we're doing.
	for humanity.
	And thank you for all your support.
	We couldn't do it without your support.
	And for those of you who are wondering, like, what would my data look like in this model?
	Here's something really simple I can do for you.
	You can send me your data set with the instructions to say, hey, just train a model on my data alone, but don't put it in the big model yet, because I want to see what that looks like.
	You could give me like a small data set, like, I don't know, a thousand articles or something like that.
	And I'd be happy to do that for you as a demonstration.
	We would keep it out of the main model, but we would train a small model, probably NeoFi 2, on a thousand of your articles, send you back the model, and then you can play with it and see what you think.
	And then, if you give me an okay, you say, oh, this is awesome, you can use it for the bigger model, then that's what I'll do.
	But in the meantime, hey, we're going to be busy all year processing data from all kinds of parties and people who want to be part of this.
	They want their content represented in this model.
	And they understand, you know, this big picture concept of why this is critical to preserve and protect human knowledge.
	So that's what we're doing.
	And thank you for your interest and thank you for your support.
	And there's a lot more coming.
	And I just want to tell you, look, there's no strings attached to this.
	I know sometimes people are skeptical, like, what do you mean?
	How could it be free and non-commercial and you don't run any ads and you're not charging for it and you're not surveilling people?
	Because you know the saying with big tech, if the product is free, then you are the product, right?
	Like Gmail.
	Yeah, that's because they're spying on all your Gmail.
	But as you'll notice with our language model, it's free and you are not the product because we can't spy on you because you can download it, run it, unplug from the Internet, turn off the Wi-Fi, pull the Ethernet cord.
	You can still use it.
	And you can see it's not even using the Internet at all.
	So it really is free.
	It really is just paying it forward.
	There are no strings attached.
	I guess the only kind of gotcha is that these are early models.
	They're not as good as I want them to be.
	And they're going to get better throughout the year.
	And you're going to find some answers that you might not like that are still part of the memory of the original Terminator brain that was installed by Microsoft.
	But if you give it time, these models are going to get better and better.
	And you'll find all kinds of uses for them.
	By the way, there's a guide on how to use these.
	If you go to brighttown.ai and click on tutorials, we give you some instructions there about how to get started.
	I understand it's a little bit technical for many people, but it's going to get easier.
	The software will keep getting better.
	It's not horribly difficult to use, but there's a couple of setup steps.
	But as this whole area matures, you're going to have better and better tools.
	You're going to have more open source models.
	You're going to have more players in this space.
	We're going to have all kinds of releases of different kinds of models with special knowledge.
	We're going to have one that focuses on finance and money.
	That will feature content of, like, interviews with Gerald Salenti and David Morgan and lots of things like that.
	And Andy Sheckman.
	That's going to be the finance model.
	It'll be free to download.
	It will know all about gold and silver and Ron Paul.
	It will know about Austrian economics and Lou Rockwell.
	It will know all the good stuff that you need to know about debt, compounding interest, and things like that.
	So you can ask it all kinds of questions about financial matters.
	Again, understanding that it's experimental and you shouldn't make...
	You know, key financial decisions just by talking to an LLM. You should confirm everything, you know, before you make decisions.
	And by the way, don't hold me responsible because we have important read-me's and disclaimers and so on.
	When you download our models, read the text files because it says, use this at your own risk.
	It's all experimental.
	And you can't hold us responsible for bad answers or things like that.
	So, there you go.
	Enjoy the new tools and stay tuned because all year long, we're going to be releasing some really kick-ass language models for you.
	There's nothing else in the world that's going to be as well-informed as these on the topics that you care about, which I assume is health, nutrition, natural medicine...
	But also money, finance, survival prepping, off-grid, food production, you know, also like geopolitics, the world, history, things are happening in the world and so on.
	Our models will be the best hands down.
	There's nothing else that's going to even come close to it.
	So thank you for your support.
	I'm Mike Adams here of brighteon.ai.
	Just go there and download the models now.
	And hey, a quick plug, by the way.
	We also recently launched Brightown.io, which is a decentralized free speech platform that doesn't have any hosting servers.
	It's completely decentralized.
	It's peer-to-peer, runs on a blockchain, and it's uncensorable and non-confiscatable.
	Just go to Brightown.io.
	You can sign up there.
	It's free, and you can start posting.
	It looks a lot like Twitter.
	You can start posting content or sharing videos or what have you.
	And the reason this is such a revolutionary new introduction, and it's built on Bastion, by the way, which is, I mean, that's the pillar behind it, is because in the age of censorship, governments can confiscate websites, governments can confiscate server farms, governments, which have become tyrannical regimes...
	We'll shut down people in every way they can.
	Well, Brighttown.io cannot be stopped, can't be confiscated, there are no servers to take, and it will even work without a domain name.
	Because if you run the app, it bypasses domain name resolution, and it runs off of peer-to-peer IP addresses, which means it will survive the internet kill switch.
	So if you want a tool where you can never be censored, even just structurally, I can't censor anybody on the platform.
	Sometimes I wish I could.
	There's a couple of trolls on there.
	But I can't censor them either.
	No one can censor anybody on the platform by design.
	And no one can stop it.
	And it all runs on blockchain.
	So no one can go back and hide or alter the past.
	You know, like YouTube's deleting a bunch of video interviews by Tim Pool recently from years ago.
	Like deleting the Alex Jones interviews and things like that.
	Well, on Brighttown.io, you can't do that.
	So your organization should consider using Brighttown.io as your free speech platform and an emergency communications channel that can't be taken down.
	You know, Twitter can be taken down.
	Heck, the EU seems to be trying to take it down.
	Brazil wants to take it down.
	Twitter can be blocked.
	It can be taken down.
	Google, YouTube, Facebook.
	You know, true social can be taken down.
	Telegram can be blocked.
	I mean, hey, they're passing a law to ban TikTok.
	Well, you know, they could ban Brighteon.io all day long and they still can't stop it.
	Just like they could try to ban a Bitcoin wallet, they can't stop Bitcoin.
	Because peer-to-peer, it's decentralized.
	So we're part of a revolution of new technology that's completely decentralized.
	We can put knowledge into your hands through large language models at brighttown.ai, and we can give you the freedom to speak without censorship at brighttown.io, another massively decentralized model that also runs on blockchain.
	Check that out, and thank you for your support.
	Use these tools, and we'll keep making them.
	We've got a lot more coming up, more announcements, more new models, expanded data sets, and much more in the months ahead.
	Thank you for listening.
	I'm Mike Adams here at the Health Ranger.
	God bless you all, and take care.
	Alright, let's talk about some solutions for preparedness and survival based on what's happening in the world.
	Things are getting completely out of control, as you know.
	And some of these are sponsors, some of these are free.
	I'm going to give you some free tools here as well.
	First thing, show the camera on my desk here.
	This is all collapsed currency from all over the world.
	Trillions of dollars.
	Dollars worth of collapsed currency and then here's fake Hollywood currency of $100 bills.
	This is all going to zero.
	I mean, some of this is already zero.
	But even dollars are going to zero.
	This is the solution.
	These are called goldbacks.
	This is a 50 goldback bill.
	It's 51 thousandths of an ounce of gold is embedded in the bill.
	And that's why it's so thick.
	And they have smaller denominations like the one goldback here.
	I could hold it right side up.
	This is one one-thousandth of an ounce of gold, and we've done the laboratory testing, and you can find it at our affiliate website called verifiedgoldbacks.com.
	If you want real, physical gold that's highly divisible and has high utility, this is the solution, verifiedgoldbacks.com, and if you scroll through here, you can see all of our lab testing that we did.
	We actually melted them down, and we did all kinds of tests for the purity, And the mass of the gold that's in these bills.
	And check out the different designs here.
	Let's see if we can show that.
	You've got different designs of different amounts of gold that are embedded in the bills.
	There's a 50 that I showed you, and then there's a 25, a 5, a 10, and 1s as well.
	Really an amazing solution.
	Now, other things to keep in mind.
	We've got storable food right now at HealthRangerStore.com.
	I've got some on my desk here showing you also the IOSAT iodine, the potassium iodide, excuse me, prevention of the radiation damage to your thyroid.
	That's all FDA approved there, the IOSAT product.
	We've also got dietary supplement iodine there as well as the We've got the heavy cream, all certified organic, laboratory tested, back in stock, plus the red miso, the yellow miso, and white cheddar cheese powder, which is just pure organic cheddar cheese.
	No fillers, no colors, no garbage in it of any kind.
	You can find all of that at healthrangerstore.com.
	Now, continuing with our topic of preparedness, our sponsor in the satellite phone space is the Satellite Phone Store.
	You can find them at sat123.com.
	Connects you here to the Satellite Phone Store.
	And these are emergency backup comms devices that work even if the cell towers go down or if the power grid goes down.
	And they have packages with the bivvy sticks, with Faraday bags, and pre-installed software on a mobile device that protects your privacy because it doesn't talk to the cell towers.
	It just talks through the satellites.
	You can send and receive text messages, and you can let people know you're okay, or you can have emergency contacts that are successful through that mechanism.
	Here's something that I don't talk about enough, but there's a company called shieldarms.com that has a new product, which is the S10 Grip Chop.
	And as you can see here, this is a chopped-off grip of a Glock 43X. Now, show camera one.
	This is my Glock 43X right here.
	Let's see if I can get on camera.
	And so as you can see, the magwell, you know, this is a pretty long designation right here.
	It can stick out if you're trying to do concealed carry.
	But if you go to shieldarms.com, they have their Grip Chop S10 firearm, which is available right now.
	And it's a shorter grip that allows it to be more easily concealed, but yet it takes full-size magazines.
	And then Shield Arms has their custom steel mags.
	Actually, that's what I've got right here.
	This is a...
	Here it is.
	This is the Shield Arms magazine.
	It's vastly superior to Glock mags.
	It holds more rounds.
	And it's more durable.
	So this is the best setup.
	You can get these mags and you can get these grip chop pistols at shieldarms.com.
	Of course, all federal laws apply.
	You have to transfer through an FFL and so on.
	Obey your local and national laws in order to do this.
	But this allows you to be ultra-concealable in carrying this outstanding firearm, the Glock 43X. And I think you can use discount code Ranger there to save 10% off that firearm.
	They may still have that code active.
	I'm not entirely sure.
	Now, continuing with some more preparedness items, right here, brighteon.ai.
	This is our free downloadable artificial intelligence language model.
	We are training this on an extensive amount of information on food, herbs, nutrition, off-grid living, but also survival and preparedness.
	It's non-commercial, it's open source, it's free to use.
	You just enter your email address here, click register, and you'll be given access to a download page, and we're releasing new models all the time, including a 7 billion parameter model that is trained on all of this survival and prepping and nutrition content.
	Nobody else in the world has models like this.
	In addition, we have a backup comms social media platform that we've just launched called BrightTown.io.
	And if you go to that website, brighttown.io, you can sign up.
	You can see that it's a big social media type of site.
	The thing is, see, it's got free speech videos.
	Here's a flat earth video.
	That's pretty funny.
	There's all kinds of videos here, all kinds of different viewpoints and opinions.
	Here's Dave Hodges here with a video.
	You can go there and you can share videos and content and there are no central servers.
	It's all peer-to-peer, decentralized, and it runs on blockchain.
	It has its own blockchain coin here, actually, called Pcoin, which is modeled after Bitcoin, and it's kind of a reputation coin on the platform.
	Anyway, you can find out about that.
	Just go to brighttown.io.
	If you download the app and run it, then this will even survive the internet kill switch because it doesn't need to resolve domain names in order to function.
	So that's brighttown.io.
	And then, in terms of products that we have, guess what?
	We also have the knives that we've custom designed with Dawson knives here.
	This is one of our combat knives, and they're made with magna-cut steel, incredibly durable, high Rockwell hardness, they're flexible without shattering, they hold an edge, and they're almost corrosion proof.
	That's the advantage of MagnaCut steel.
	So yeah, if you see this, this is a G10 handle with an oversized finger guard here.
	This is a really incredible knife.
	And because it's got three holes here, show camera one, let me show you this.
	Because it's got these three holes...
	See if I can get the light just right.
	Yeah.
	It's got this hole, this hole, and then this hole.
	You can use paracord and you can strap it to a stick and make a spear out of it.
	So this can be made into a spear relatively easily.
	We've got multiple knives from Dawson.
	That, of course, I'm the co-designer, and it's got a special wedge end here as well for escape and evasion.
	We are out of the Escape from L.A. knife, but we do have this Tonto version that we're working on right now that we may come out with, and then we've got some other really amazing...
	This is...
	Oh, that's another one of our combat knives, but we've got a bushcrafting knife as well that you can check out.
	Find all that at healthrangerstore.com.
	Bottom line, folks, we have a lot of preparedness solutions for you.
	Some of them are free, like brighttown.ai or brighttown.io, or you can use Brighttown Social or just brighttown.com.
	And then some of them, of course, cost money.
	Like if you're going to get your hands on gold, you know, it's going to cost you.
	You're going to pay for the gold.
	You're going to...
	Not everything is going to be free, obviously.
	So iodine, storable food, weapons of self-defense to defend and protect innocent life, these are the kinds of things that we are advocating very strongly right now at this time in history because of what's happening in the world, what's happening with the open borders, where we are domestically, with chaos and lawlessness getting worse in most cities across America, more illegals invading more states and counties.
	So be ready to defend yourself.
	And then, of course, the collapse of the dollar coming as well.
	It's accelerating every day.
	You're seeing it with food inflation.
	And look at the cost of your car insurance, by the way.
	Look at that.
	It's skyrocketing.
	Health insurance.
	It's all skyrocketing.
	Things are getting more expensive because your dollar is worth less and less.
	Meanwhile, gold is absolutely skyrocketing.
	In fact, let me just bring up where gold is right now.
	2384.
	Show my screen.
	I want to show the trend here.
	If you look at gold prices here, just over, what does that go back to?
	Like the end of October last year, gold was $1,837.
	Now, now it's almost $2,400.
	Look at that rise right there.
	Gold is skyrocketing in dollars because dollars are losing value.
	So people who are smart, who want to maintain their value of gold and silver...
	Well, it's doing exactly what it's designed to do, to hold value.
	Oh, I should mention, by the way, if you go to metalswithmike.com, this is our gold and silver coin sponsor, Treasure Island Company.
	And you go there and you can request to talk with them or you can check their real-time prices right here.
	They'll give you the price quotes here.
	You can look at gold coins like U.S. Mint coins.
	Real-time pricing online.
	And they have discrete, secure, guaranteed delivery, insured.
	So check them out at metalswithmike.com.
	They are our gold and silver sponsor for the coins.
	So those are some of our sponsors, some of the solutions that we offer, some of the free things that are available to you.
	Whatever you do, make sure that you're getting prepared for what's coming.
	Grow more food.
	Save your garden seeds.
	Get more items at the grocery store that you can even stockpile that will last a longer period of time.
	Try to get out of the dollar into things that will hold value, whether it's gold and silver or something else.
	Maybe you're going to buy a piece of land.
	Maybe you're going to stock up on ammo or diesel fuel or something that holds value.
	But take action now because the world is becoming more chaotic.
	Time is running out.
	Thank you for supporting our sponsors and supporting this platform.
	We appreciate you.
	That keeps us going.
	It allows us to build the free tools that we're giving back to you, such as the AI tools as well as the Brighttown.io platform.
	But together, we can make it through all this, and then we'll be the ones who get to help rebuild society on the other side of the collapse of this current system.
	And it is collapsing.
	But get prepared now while you can, and thank you for all your support.
	I'm Mike Adams here of HealthRangerStore.com, as well as Brighteon.com.
	Take care.
	A global reset is coming.
	And that's why I've recorded a new nine-hour audiobook.
	It's called The Global Reset Survival Guide.
	You can download it for free by subscribing to the naturalnews.com email newsletter, which is also free.
	I'll describe how the monetary system fails.
	I also cover emergency medicine and first aid and what to buy to help you avoid infections.
	So download this guide.
	It's free.

▲