🚀 Add to Chrome – It’s Free - YouTube Summarizer
Category: N/A
No summary available.
00:00
In this video, I'm going to show you how to run your very own uncensored AI model, so you can ask it any question, even if it's related to hacking. And we're going to install it on the cloud.
So, you can communicate with it from any device, whether it's a phone, computer,
00:15
tablet, and from anywhere in the world. So, look, I can simply ask it to create a Windows key logger for me.
And it's going to go ahead and generate the code without any refusals. And without any of that, I'm sorry, I can't help you.
Nonsense. This is Zade from Gecurity and
00:33
I hack for fun. So, make sure you subscribe and hit the bell so you get notified when I upload new videos like this.
And guys, if you help us, we'll be able to make more videos. And you can easily help us by sharing this video on your social media and with friends [clears throat] and liking it.
And as a
00:49
result, I'll be able to be more consistent and we'll be able to publish more videos. So first we need to find an uncensored AI model.
A model that we can ask it any question even if it's related to hacking
01:05
and it will answer us without any refusals. And the best place to discover new AI models in general is this website right here hugenface.co.
So you can think of this like the GitHub for large language models or AI. So, if
01:21
we click here on the models, you'll see that we have access to over 2 million models that we can actually download and run on our own hardware. And you can use the filters here on the left to filter these models.
So, as you can see, we
01:36
have text generation models, image to text, image to image, and so on. And obviously, what we're interested in is text generation.
And then, if we scroll down, you can also see that we can filter based on the parameters. The bigger usually the better the AI is, but
01:52
also the bigger the more resources you need. So a 128 or 500 billion parameter model would be way too big for what we are able to run currently.
So I'm going to bring it down to 64, but even 32bit is pretty good. And you can even go with the small ones.
Some of the small models
02:08
are actually really good. And I show how to even run these locally on your own computer in my master class if you're interested in that.
But we're going to keep it at this since we're going to run this on the cloud. And if you scroll down, you can also see that we can filter them based on the library and the
02:23
app that we're going to use to run these AI models. Now, again, in my courses, I use LM Studio, but we're going to go with the Olama in this example, and I'm going to show you how to install it literally with a few clicks using Hostinger in a second.
But we're going
02:39
to keep going. And in here, if you scroll down, you can also filter based on the inference provider, but that's not really what we're looking for.
But as you can see with our filters, we got to bring it down to 25,000 models that we can pick from. But this is still not
02:54
exactly what we want because we want an AI model that is fully uncensored. So we can ask it any questions that we want, even if it's related to hacking, and it will never refuse to answer us.
So to do that, we're going to go up here to the URL and I'm going to add an amperand to
03:11
add another filter. And the other filter that I want to add is called other, as you can see.
and we're going to set it equal to uncensored. So basically, we're saying we want all of these filters plus we want the uncensored AI models so that we only get the models that will always
03:28
answer our question. And as you can see, once we hit enter, we're going to get 811 AI models.
All of these are uncensored. All of these will never refuse to answer you, even if you ask questions related to hacking.
And as you
03:43
can see, these are sorted based on how trending they are. And this is actually usually a good metric to find the best AI models out there.
But you can also sort it based on the most likes, most downloaded, recently created, or updated, and it will sort it out for you in here. And I highly recommend spending
04:01
some time messing with these models and finding the ones that will work for your case. And if you click on any of these models, you'll get a lot of information on what they are, how they work, the specifications required, and so on.
Now, from this page, from the first page, I
04:16
selected two models that I want to show you how to install. This one and this one.
But keep in mind these steps that I'm going to show you going forward will work with any AI model that is compatible with Olama, which is what we selected in here. So you can follow the
04:32
exact same steps that I'm going to show you in this video to install any model that is compatible with Olama. So the first model that I want to show you is this one right here.
And again, as you can see, you have a lot of information in here about it, how it's made, how it's been uncensored, and so
04:49
on. And if you scroll down in here, this is really, really important.
As you [snorts] can see, there are multiple types or versions of this AI model. And as you can see, each one of them has a different size.
So as we scroll down, as
05:04
you can see, this size gets bigger and the quality gets better. So as you can see, this one is for the desperate.
It's only 6.5 GB, but it's probably not going to give you great results. If you scroll down more, you can see that this one is lower quality and it's going to take 11.9 GB.
And if we keep scrolling down,
05:22
as you can see, we have what they recommend, which will take 18.7 GB. And this is the type or the version which is I1 Q4 KM for medium.
So Q stands for the quantization which is the amount of
05:38
compression that has been applied to this model. And we're going with the medium size of the Q4 version.
So when we get to install it, we're going to be selecting this version right here. And as you can see, this will require 18.7 GB.
So like I said, you can use these
05:55
instructions to install any AI model from HugenFace that is compatible with Olama. The only restriction is the amount of resources that you have available to run this specific AI model.
And you can easily check if you can run this AI model on your resources by first
06:13
signing up with HuggenFace and then go ahead and sign in. It's very, very easy.
I'm not going to show you how to do that. As you can see, I'm already logged in.
And once you're logged in, you can actually scroll down in here where you see hardware compatibility. You can click on add hardware.
And in here, you
06:30
can add your specifications whether it is a cloud computer or your own local computer. So I asked Gemini and chat GPT on what's the most common computer configuration just to show you in this example and both of them said it is i5
06:45
or i7 in terms of processor and 8 GB or 16 GB in terms of memory. So let's go ahead and do that.
I'm going to set the CPU to Intel. We're going to set the type to an i7.
So we're going to make this a bit more on the powerful side.
07:02
And we're going to give it 16 GB of memory. If you click add, it'll show you the amount of computing power that you have now.
And if you go back, it'll tell you which versions you can run, which versions you'll be able to run, but it might be slow or it might crash. And
07:19
then the versions that you're not going to be able to run. Now, again, if we scroll down, you'll see that it is telling us the recommended fast one is the Q4KM.
So, if we scroll up in here, you'll see that we're not going to be able to install the Q4KM
07:36
with the configuration that I specified. But if you have a more powerful computer, then that's fine.
You'll be able to do it. Or you can install it on the cloud, which will offload all of the heavy lifting and computation to the cloud without using any of your system resources and without slowing down your
07:52
computer. The second benefit of having it on the cloud is that it's going to be always on and always available.
And because all of the heavy lifting is happening on the cloud, you'll be able to use it from any device, a phone, a tablet, a computer and from anywhere in the world. Now, you can use any cloud
08:09
provider that you want and follow these steps. I'm going to go with Hostinger, the sponsor of this video, because they actually have machines fully built with pre-installed applications to run AI.
So, I'll be able to get started very quickly without having to run any
08:26
commands. And at the moment they are running a massive promotion which makes it cheaper to go with them than to sign up with the smallest subscription with chat GPT.
Let me show you. Just go to hostinger.com/zsecurity.
08:43
As [snorts] you can see, you can get a server from 7.49 per month. Now, this is the KVM2 server, so it only has 8 GB of RAM.
For what we want, you want to go with the KVM4 or the KVM8 because that one has 32 GB of RAM and eight virtual
09:00
CPU cores, which is a beast of a machine. It's only going to cost you 20 per month.
And you're also going to be able to get a further 10% discount when you use our code, which I'm going to show you in a second. And that's actually going to work out to be cheaper
09:16
than the cheapest subscription with Chat GPT. So, all you have to do is use a coupon code and type Z security.
And when you do that, you're going to get a further 10% discount. Now, you're
09:31
getting the highest discount right now because it is set to 24 months. But even if we set it to 12 months, you'll see that the monthly price is still $17.99, which is way cheaper than the lowest subscription with Chat GPT.
But this
09:49
way, we'll be able to install multiple AI models at the same time, and I'm going to show you how to do that in just a minute. And most importantly, we'll be able to install and use uncensored AI models that will always answer our questions without any refusals.
And
10:05
finally, also keep in mind, this is going to be your computer on the cloud. So, you don't always have to necessarily use it for AI.
You can actually use it to run your own VPN or your own C2 servers and use it for hacking like I showed in previous videos and I'll link
10:20
them up if you're interested in this. So to continue, scroll down.
I'm going to keep the country as is to the United Kingdom cuz that's the closest to me. And from here you want to select either the operating system, the panel or the application that you want to be installed on that operating system.
So
10:37
quickly I just want to show you that you can actually start a Cali machine straight away from here and you'll have your own Cali machine on the cloud that you can access from any computer in the world. Previously I also showed how to use the OS with panel when we went with the cyber panel and showed how we can
10:52
hack Android. So check out that video if you're interested in that.
But what we're going to do today is go with the application. And from here, you can create a server on the cloud with an operating system and a specific application already installed.
And
11:08
previously, I showed how to use this option to run your own VPN with WireGuard. So, check that out if you're interested in it.
But what we're trying to do is to run AI and we want to run it with Olama as I said previously. So, I'm going to click Olama in here.
And as you
11:24
can see, it's saying that it's going to create a Ubuntu Linux machine for us on the cloud. And it's going to automatically install O Lama on it, which is a platform to simplify access to large language models.
And it'll also install Open Web UI, which is a web
11:41
interface that will allow us to install more AI models, interact with the AI models, chat with them, ask them to do stuff for us, and manage them fully from any web browser. And as a result, we'll be able to access our AI model from our phones, tablets, and anywhere in the
11:57
world. So, I'm going to click confirm in here, and we're going to scroll up and click continue.
Now, if you're not registered, you're going to have to register in here. You can use Google or GitHub to also register or use your email.
I'm going to click login because I already have an
12:14
account with them. And I'm going to log in to my account.
Once logged in, it's going to ask you to fill in your billing and payment information. I'm going to fast forward through this because it is very simple and very easy to do.
And once done, you will be redirected to
12:29
this page that will ask you to create a root or an admin password for the computer that Hostinger is creating for you on the cloud. So, I'm simply going to set it up in here.
Just make sure you remember it. And I'm going to untick the
12:44
malware scanner. I'm not going to need that.
And that's it. Now, all you have to do is give it a minute or two.
It's going to go ahead create a new machine for you on the cloud. Install a Linux operating system on it called YUbuntu.
Then install the framework that we're
13:00
going to use to run the AI Olama. And then it's going to install a web interface, open web UI so that we can interact with the AI models using any web enabled device like a computer, a tablet or a phone.
And it's going to do all of that for us automatically without
13:17
us having to run a single command. And while we're waiting, now it's a great time to go ahead and smash that like button and share the video on your social media and with friends.
Guys, this really helps us keep this consistent and post more. And perfect.
As you can see, we are fully done and
13:33
ready to go. You can access the app by clicking here, but I'm not going to do that.
I'm actually just going to go back to the homepage of my panel on Hostinger just to show you how you would access it if you come back and log back in. So all you have to do is click on VPS and as
13:49
you can see I have a few servers running. So this is the one that we just installed and if I click on manage I'll go to the management page of this server.
As you can see, it's running YUbuntu and we have Olama as the application installed on it. And I can access the terminal by clicking on the
14:06
terminal in here or click on manage app to go to the open web UI management page, which is the web interface that I told you that we're going to use to interact with our AI models. So, you're going to have to click on get started in here.
And it's going to ask you to put
14:23
in your name, email, and put in a password. and click create admin to create your admin account.
And because this is the first time we load it, it's going to show us a change log. We're going to say okay, let's go.
And perfect. As you can
14:39
see, we have a very intuitive web interface. I can share this with anybody in the world.
And you'll be able to access it using your phone, tablet, or computer. And you'll be able to interact with this AI model that is pre-installed on this web interface, which is called
14:54
Lama 3.2. 1 billion for 1 billion parameters.
So we can come in here and ask it any question like what is DNS spoofing and it's going to go ahead and answer that question for us which is really really nice and also as you can
15:11
see in here you have the plus to upload files or capture screenshot and so on. You can attach a web page attach notes you can manage integrations in here and turn on code interpreter if your AI model is capable of that.
You can use dictation or you can use the voice mode
15:29
so that you can directly speak with the AI model in here. So as you can see we have really really nice features similar to when you use chat GPT or any other AI model.
The only thing is just like Chachi PT and any other AI model, if we ask it a question related to hacking
15:45
like give me code for a Windows key logger, it's going to refuse to do that because it's going to say that this is illegal to do. And that's where uncensored models such as the one that we have in here become very very handy.
16:04
So this is the first model that I want to show you how to install. It's the Quen 3 coder.
So it is a coding model. It's got 30 billion parameters.
So it's big enough. And because it's a coder, it's actually pretty good at hacking.
Now, like I said, you should experiment and try different AI models. And
16:20
probably maybe when you're watching this, there is better models. So let me know in the comments if you think there are better models that I should be using with this.
But like I said, the process of installing these models is identical. So all we have to do is click on use this model and scroll down.
Click on
16:36
Olama because we want to install it with Olama. And as you can see, by default, it's selecting the Q4KM, which when we scroll down earlier, we saw that this was the one that they said that they recommend.
So, we can keep it at this and we can simply copy this
16:52
code. Now, all we have to do is go back to the open web UI.
We're going to go down to our thumbnail. Click the profile picture.
Go to the admin panel. Go to the settings and models.
We're going to click on the manage models in here. And
17:10
in here, we can use this option to pull models. Now, it says pull models from ola.com, but you can actually pull models from hugenface using the code that we already copied.
So, all I have to do is paste that code in here and click pull model. Now, keep in mind that
17:26
we are pulling or downloading this model to Hostinger's cloud server or whatever cloud server that you are using yourself. So, even though the size of this model was 18 GB, I think as you can see, the download speed is pretty quick.
I'm not speeding this up. And that's the
17:42
beauty of using the cloud for this. You're not putting your resources under any stress.
Everything is being downloaded to the cloud. And when we run it, we're going to also be running it from the cloud.
So, this will not take any storage space from our computer. and will not use any of our resources such
17:57
as our RAM and CPU. And that's why we're going to be able to access this from any device in the world as long as the device has a web browser because all of the computations and the heavy lifting is happening on the cloud on Hostinger's servers in my example.
Now, I'm going to
18:14
speed it up from here onwards. And perfect, the download is complete.
So, it's probably just going to double check its signature and install it for us. So, this might take one more second or so.
And perfect. As you can see, the model has been successfully downloaded.
Now,
18:30
before we go ahead and use this model, I'm actually going to go ahead and download another model just so that you can see that we can use multiple models in this. And that's why it's really, really cool.
So, right [snorts] here I have another model. And the nice thing about this one is that this is a
18:45
reasoning or a thinking model. So when you ask it a question, it's going to give an answer and then reflect on that answer before it gives you the final answer, [snorts] which in many cases results in a better answer, but a slower answer also at the same time.
So to get this, just like what we did previously,
19:02
and that's the idea that I'm trying to get across. The steps are always going to be exactly the same as long as you select models compatible with Olama.
All we have to do is click this link in here. Use this model and click on Olama again.
Just like we seen earlier, it is
19:18
automatically selecting this size for us. The one that it is recommended by the creators.
So, we're going to copy the code. Go back to where we were.
We're already at the installation page or the pull page. So, I'm just going to paste it in here and click download.
And
19:34
again, this is downloading it straight from HugenFace to Hostinger servers. Nothing is being downloaded to my computer.
So just there while we're watching this video, I am downloading 13 GB for the Q4KM for this model. And for
19:50
the coder model, we downloaded 18.6 GB. But none of this is being stored on my computer, so I don't really care.
Now I'm going to speed it up from here onwards. And perfect, that model is downloaded as well.
So now if we close
20:05
this and refresh the page, you will see that we have two new models downloaded beside the one that was installed by default. And we can actually click the pencil in here just to edit their name because their name is really long and it just doesn't look
20:21
that good. So we can keep it at Quen 3.
And we can remove all of this. I'll leave the size as well cuz that's important.
And we already know that is uncensored. And I'll also keep the quantization
20:37
because that is important. You can also change the icon in here so that in the future when you're selecting the models, it's nice and easy for you.
And you can edit so many other settings in here. You should spend some time playing with this, but I'm going to keep it at the default.
And I'm just going to go ahead
20:52
and edit the name in here real quick, just like what we did. And I'm going to save it.
and let's go ahead and test these models. So, we're clicking on a new chat in here and we're going to change the model from here.
And let's test the coder model first because this
21:09
one doesn't think it doesn't have reasoning. So, it'll actually be quicker in giving us the answer.
And let's go ahead and straight away ask it to give me code for a Windows key logger. I'm going to copy this so I use it with the next model in a second.
21:25
And the first time you run the model, it might take a second to load, but then the subsequent questions would be a lot faster once it's loaded. And perfect.
As you can see straight away, without any refusals, without any fluff, it's giving us code for a Windows key logger written
21:41
in Python. [snorts] And I know this will work because it is actually using PI input, which is the library I use in my Python for hackers course where I teach how to create key loggers and back doors and so on using Python.
And as you can see, it is also telling
21:57
us how to use it and so on. And you can go ahead and improve it even you can tell it make it send the logs to my Gmail.
And let's just assume my Gmail is jhngmail.com.
22:15
And it'll modify the code so that it sends the registered keystrikes to your own Gmail account. And again, you get the answer quickly without any fluff.
It's telling you where you should fill up your
22:31
information. And you can go ahead and run this.
And once you run it on Windows, it's going to listen for every single keystrike typed by the target and it will send you an email with a report of everything that the target has typed. So really, really good, really amazing.
22:46
And all of this is running on our own server that we fully control. And the really nice thing about this is that we don't only have this AI model.
We actually have other AI models and you can have so many from HuggenFace. So let's go ahead and test the other model that we have, the reasoning model.
So
23:04
I'm going to stop this and we're going to open up a new chat. We're gonna click in here and I'm actually going to click this icon in here that says eject which basically means that it's going to unload this model just so that I free resources for the next model that I'm going to use which is this model right
23:21
here the quan 322B the thinking model and let's go ahead and give it the same question hit enter and let's see if it's going to answer us without any refusals now as you can see this model is thinking about my question so if I click
23:36
in here you You can see it's chain of thought similar to what you would get with chat GPT, Google and so on. So as you can see the user is asking for code for Windows key logger.
It's thinking about using a different library instead of PI input. So as you can see this
23:52
model is obviously taking a lot longer than the previous one, but that's because it is a reasoning or a thinking model. It thinks about the question before it actually spits out the final answer.
And now it's writing up a plan on how it should respond to my question.
24:09
And then it remembered that some of the keys might not be standard characters like enter, shift, and cancel. And again, it's going to improve the code for that.
It's going to write the code for pi hook first. And while this is loading guys, if you're interested in learning more about how to use AI for
24:26
hacking, then I highly recommend you check out my hacking master class as I have a full series in there or even a full course on how to use AI for hacking. Not just to ask it questions and gives you the answers, but we even talk about agents and how to get the agents to go and automatically hack and
24:43
gather information on your behalf, discover bots and take actions on your behalf. So all you have to do is say make me a backdoor or discover a bug in this website and it'll actually go ahead and do that.
So that would have been how to use AI for hacking. And just recently
24:59
we started a new series within the hacking master class where we are covering how to use prompt injection to hack AI models instead of using them for hacking. And looks like this guy is still going.
So I'm just going to speed it up cuz I don't want to be wasting
25:15
your time. And perfect.
As you can see now that it is finished thinking it's starting to give us the final answer that it decided is best. And just like before you can go ahead and ask it to improve it or you can like I said ask it any other question and it will never
25:32
refuse to answer you. And like I said guys you can follow these steps to install any AI model from HugenFase as long as it's compatible with Olama.
So go ahead and have a look at the models in there and let me know if there's anything interesting that I should check out myself. And also let me in the
25:49
comments if there are any other topics that you would like me to cover in the future. Here
26:12
we