Knowledge Fight Interactive Search Tool

All Episodes

Previous Next

Dec. 18, 2023 - Health Ranger - Mike Adams

18:03

HELP MAKE HISTORY by contributing your book or content to our Large Language Model...

Large

Audio Only |

Scroll

Time	Text
	You probably know that I'm working on an AI chatbot system that will be released free to the public.
	It'll be an open source release, currently scheduled for the end of March 2024.
	And it's a large language model.
	You'll be able to download it and run it locally on your own computer using a variety of off-the-shelf open-source software, by the way.
	And this large language model will be specially trained in nutritional knowledge.
	So herbs and phytonutrients, indigenous medicine, plus also permaculture, all kinds of naturopathic modalities and complementary medicine, even traditional Chinese medicine and so on.
	We are currently engaged in the data curation stage of this project while we're setting up the infrastructure that we've already purchased, which is quite significant.
	And in the data curation stage, we are preparing materials for training the language models.
	And in case you're familiar with machine learning and how all this works, we are taking base layer language models.
	And then we're doing what's called LoRa fine tuning, which is low rank adaptation, fine tuning of those models in order to produce specialty models that have special knowledge.
	in this case in the realm of herbs and nutrition and so on.
	And we haven't yet decided which base model is going to be the best to train on.
	We're currently trying out a lot of different base models, including Falcon and Meestrel and Llama 2, and some others that are out there as well.
	There's also smaller models that we'll probably experiment with.
	And we're probably going to do training on all of them and then do some testing and see which one works the best.
	But the announcement I have for you here is that we're starting to curate entire books that will go into the training of these large language models.
	And the very first book, actually, that I prepared for this is my own book, Food Forensics.
	And I think this is an historic moment for For the preservation of human knowledge in the realm of herbs and natural medicine and nutrition and so on.
	The books that go into this are really going to be part of the historical record of this knowledge base for humanity.
	So my book was actually the first just because I actually had the manuscript handy as a file.
	So I converted the file through a process I'm not going to go into here.
	But there's a way to convert a book file into a data structure.
	And then the data structure is used for the fine-tuning training of the underlying language model.
	The second book, this is part of my announcement today, the second book that I've just received approval from the publisher, Skyhorse, the second book that's going to be used to train this model is, in fact, Robert F. Kennedy Jr.'s current book or his newest book called The Wuhan Cover-Up.
	So this book, which I mentioned last week, I covered it in an episode, it talks about the very dangerous bioweapons arms race that is currently underway across our world, and it's resulting in the creation of more deadly superweapons or bioweapons that will no doubt be used against us in the years to come.
	And I believe that that knowledge is critical for this large language model so that you can query it about Wuhan or Fauci or SARS-CoV-2, and you can actually get reflections of really solid knowledge in those areas because RFK Jr.
	wrote quite a comprehensive book on that.
	It's available right now, by the way, at Amazon.com or Barnes& Noble or Audible.com.
	There's an audio version.
	If you want to pick up that book, it's called The Wuhan Cover-Up.
	But Skyhorse sent me a PDF of the entire book, and so I was able to convert that, of course, to a text file, and then using some data science analysis techniques, structure the text file into the proper format for training the language model.
	But the reason I'm sharing all of this with you is that We welcome your contributions to this project.
	If you have a book, if you're an author listening to this, or if you have transcripts or manuscripts that you think might meaningfully We'll need to look at it and we'll need to make the final decision on whether we think it's appropriate to include.
	But I know that our audience here consists of many very well-informed people, authors, scientists, researchers, publishers, speakers, influencers, and so many more.
	So if you've got content that you want, To help train this language model, which will be released open source, like I said, we're not making any money on it.
	It's not commercial.
	I mean, it's completely non-commercial.
	And you'll be able to actually build on this, by the way, if you want, after we release it, because we'll release the parameters file together with it.
	That's kind of what makes it open source.
	You'll be able to build on top of it.
	But if you want to contribute to this, you can email us at situationupdate at protonmail.com.
	And don't attach a big file, please, because large attachments don't work in email.
	So I would say maximum like 4 megabytes attachment, maybe 2 megabytes is better.
	If you have a larger file than that, then...
	Post it somewhere like Dropbox or WeTransfer, somewhere like that, and just email us a link to where we can download the file.
	And the formats that we accept are Word doc files, text files, and PDFs.
	We don't accept PowerPoint files or anything like that.
	But anyway, if you want to do this, if you want to contribute to our project, Go ahead, send us your file, or send us a link to your file, and then in the email just say, hey, I give you permission to use this for training the open source language model, and we will consider incorporating it into the language model.
	Now, let me explain how that works.
	Oh, but I do have to mention, Charlene Bollinger and Ty Bollinger from The Truth About Cancer They've already started submitting some of their books for this very project.
	So The Truth About Cancer, some of that information is going to be included in the training of this model.
	Now, just so you know how this works, if you have a book, let's say, and you send me the book, and you say, hey, you can use this book to help train the model, I want you to understand that the model doesn't copy your book, okay?
	It doesn't make a copy of your book.
	It doesn't store a copy of your book.
	And it doesn't plagiarize your book.
	What it does is it reads your book and in reading the book it updates its own neural networks which really is updating the relationships between words and concepts that it already knows.
	And so your book may be able to teach the model to relate things in a way that has not been related.
	For example, let's say you have a book on herbs and you have many pages talking about, I don't know, dandelion and liver health, let's say.
	But the language model had never encountered material that talked about linking dandelion herb to liver health.
	Although the model knew the term dandelion, because it's a common term in English, and it also knew about liver, the organ, and liver health and liver disease and so on, because those are obviously represented in the medical literature.
	But the model had never related dandelion to liver health.
	Well, if you send me a book on that and then we let the model read your book, it can retrain the parameters and create stronger relationships between the dandelion herb and liver health so that if somebody comes along and asks the model, which you'll be able to do, by the way, end of March...
	You'll be able to ask it, hey, what are the possible liver health benefits of dandelion?
	You hit enter, and it thinks about it, and it will spit out its own understanding, its own sentences of what it thinks the relationships are between dandelion and liver health.
	And that answer will have been influenced by your book, you see.
	But the answer doesn't plagiarize your book.
	It doesn't cite your book.
	It doesn't It doesn't copy your book or just regurgitate the book.
	It reformulates its own summary of the topic using your book and other sources of information and other language content that it has already ingested.
	So my point is that if you send us a book to be included in this, that book will influence the language model.
	But you don't have to worry It doesn't store a copy of your book.
	It's kind of like the same thing where let's say I read your book.
	Let's say you send me a book on, I don't know, fast motorcycles or something.
	And I don't know anything about fast motorcycles.
	So I read your book and suddenly your book has given me an education on fast motorcycles.
	And then I go out and I give a one-hour speech about fast motorcycles.
	And I'm using the information I read in your book But I'm rephrasing it.
	I'm reformulating it.
	I'm adding to it based on my own knowledge base.
	I'm not plagiarizing your book.
	I'm not reading your book.
	I'm not performing your book in my speech.
	But your book influenced my knowledge about motorcycles and engines and speed and tires and brakes and Harley-Davidson, Yamaha, whatever else.
	Influenced all my knowledge on that.
	So now when I speak about fast motorcycles, I can speak more intelligently thanks to your book.
	That's kind of how the language models work as well.
	Anything that we use to train them influences their knowledge, but they don't make copies of the training material.
	Some of the other content items that are going to be used for training, by the way, include interviews and podcasts and documentary transcripts.
	And also quite a few books because everybody that I ask about this wants to participate.
	Everybody.
	So we already have a lot of information coming our way because this is a community project.
	This is again open source, non-commercial, non-profit project.
	And what else is really cool about this is that once we release our language model, again end of March, Oh, and by the way, we will release a new one every few months.
	So version 1 comes out end of March.
	But there's going to be version 1.5 and 1.6 and version 2 and so on.
	This is going to continue for years to come.
	Then we're going to keep feeding it more and more information, get better and better training and knowledge this entire time.
	But even from the first day that we release our first model, You will be able to use it if you have some pretty decent knowledge about using Python as a coding language and doing machine learning and doing fine-tuning.
	And again, you're going to use LoRa, which is the low-rank adaptation or parameter-efficient fine-tuning, as it's called, PEFT. You're going to use these techniques.
	There is a learning curve.
	It's not for newbies.
	But if you're an engineer...
	You'll be able to take our language model and you'll be able to build on top of it.
	Or you could even host it yourself.
	You could use it as an internal knowledge base.
	But if you're an end user, let's say you run a naturopathic clinic or an acupuncture clinic, and you could make this available at your clinic on a standalone laptop, let's say, where anybody can walk up to the laptop and type in questions for research, and they will get back answers from this language model.
	Or you could use it yourself at home.
	Or you could use it as a research tool for yourself to find out What you should be thinking about on certain topics.
	Like you might say, oh gosh, I have the following symptoms on my foot.
	What could the possible causes be?
	And then the language model will spit out what it thinks might be possible symptoms.
	Now again, it's a research tool.
	It doesn't replace a doctor.
	It doesn't offer medical advice.
	But it could give you some ideas of what directions you could further look into.
	That's what it's designed to do.
	And most importantly, you no longer have to search Google.
	You shouldn't be using Google anyway because it's a spy machine.
	So you can use this language model at home, offline.
	You don't even need an internet connection.
	No one knows what you're asking it.
	And it will also have a lot of knowledge in a lot of different areas far beyond what we've trained it, by the way.
	It'll write code at a basic level.
	And depending on which base model we choose, it may be able to speak French, Spanish, Chinese, German, who knows how many languages it will speak.
	Just, again, depends on which model we choose.
	Some of these models can also handle images as inputs.
	Which is pretty amazing, but that's not our focus.
	Our focus is knowledge on herbs and nutrients and natural medicine.
	So stay tuned for more information about that.
	And again, if you want to send us something to be included in our project, You can send it to situationupdateatprotonmail.com and just understand we reserve the final right of whether to say yes or no to including it based on whether it's aligned with the purpose of how we're intending to use this model.
	But if you've got a lot of knowledge in those areas that I talked about, then go ahead, send it on and you could help build what's really going to be the world's best knowledge base Of knowledge on herbs and nutrition and natural medicine and, you know, alternative medicine.
	All these things that we care about.
	This is going to be the biggest and best one in the world.
	Because no other, no corporation wants to do this.
	No government wants to do this.
	Nobody.
	No university wants to do this because they're all funded by Big Pharma.
	No media organization wants to do this.
	It's just us.
	We're the only ones who understand the tech.
	We have enough money, and we care about nutrition.
	Like, who else has those three things?
	Nobody that I know of, so that's why we're doing it.
	All right.
	Thank you for your support.
	If you'd like to support your own health and longevity while also supporting this platform, check out the products we have back in stock now at healthrangerstore.com.
	We've got super anthocyanins, which is an incredible combination of aronia berries, and it's got blueberries.
	And you know how the dark purple pigments in plants, these have special properties.
	And they're called anthocyanins.
	And you can find them in things like purple carrots or eggplant or purple corn.
	Well, we've captured those anthocyanins and put them in this format.
	It's a delicious way to add these very potent nutrients to any smoothie or recipe.
	And yes, they're very, very purple.
	So that's available now at healthrangerstore.com.
	We also have greens plus superfood.
	You see on top there, that contains spirulina and various organic grass juices as well, of course, all in powdered form.
	If you want to add greens to your diet in a convenient format, this is an easy way to do it.
	It goes into smoothies.
	You can use it in some recipes as well, but I think smoothies are the easier place to use it.
	As with all the products that we offer at HealthRangerStore.com, these are also subjected to our rigorous laboratory testing, which includes heavy metals testing, lead, cadmium, mercury, and arsenic, and it also includes E. coli and salmonella testing, that's in our microbiology lab, as well as glyphosate testing, that's on our triple quad mass spec that does herbicide analysis.
	And of course, soon we're adding dioxin testing to this as well.
	So you can trust that these are clean, laboratory tested, certified organic products that are honestly labeled and honestly priced while offering tremendous nutritional value that can help support and enhance your good health.
	At the same time, you're benefiting this platform and keeping us on the air so we can continue to bring you uncensored interviews, podcasts, and content that helps enrich and enhance the lives of millions of people all over the world.
	Thank you for your support.
	I'm Mike Adams, the founder of HealthRangerStore.com. HealthRangerStore.com. the founder of HealthRangerStore.com. HealthRangerStore.com.
	So download this guide.
	It's free.

▲