I Built Client Side Scanning Using My Local AI - Demonstrating Windows Recall/Apple CSAM Scanning

Подписаться 493 тыс.

Просмотров 15 тыс.

50% 1

Key Point: In the AI Series, this Ep 2 video focuses on the AI Agent aspect of AI. This is where an OS can utilize AI dependent programs to do tasks like surveillance.
Using Local AI Tools and my own AI Chatbot, I will demonstrate how I built the pieces of what would have been the functionality of Windows Recall. And also some of the functionality of Apple Client-Side Scanning (CSAM) scanning. This is a proof of concept video as well as a heads up on the dangers of AI embedded in Operating systems which can do invisible things on your machine.
-----------------------------------
Brax Virtual Phone, De-Googled Phones, BytzVPN, BraxMail, BraxRouters are available on brax.me. Sign in to the platform to see the store. You will not be asked for personal information to sign up.
-----------------------------------
Merch Store
my-store-c37a5...
-----------------------------------
I'm the Internet Privacy Guy. I'm a public interest technologist. I'm here to educate. You are losing your Internet privacy and Internet security every day if you don't fight for it. Your data is collected with endless permanent data mining. Learn about a TOR router, a VPN , antivirus, spyware, firewalls, IP address, wifi triangulation, data privacy regulation, backups and tech tools, and evading mass surveillance from NSA, CIA, FBI. Learn how to be anonymous on the Internet so you are not profiled. Learn to speak freely with pseudo anonymity. Learn more about the dangers of the inernet and the dangers of social media, dangers of email.
I like alternative communication technology like Amateur Radio and data communications using Analog. I'm a licensed HAM operator.
Support this channel on Patreon! www.patreon.co...
Contact Rob on the Brax.Me App (@robbraxman) for encrypted conversations (open source platform)
brax.me/home/rob Store for BytzVPN, BraxRouter, De-googled Privacy AOSP Phones, Linux phones, and merchandise
bytzvpn.com Premium VPN with Pi-Hole, Cloud-Based TOR Routing
whatthezuck.net Cybersecurity Reference
brax.me Privacy Focused Social Media - Open Source
Please follow me on
Odysee
odysee.com/$/i...
Rumble
rumble.com/c/r...

Наука

Опубликовано:

6 окт 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 127

@jondoe9581 11 дней назад

As usual Rob , VERY WELL DONE presentation. I sincerely hope people give this serious deep thought to prepare for counter measures.

@KillTheFear 3 месяца назад

You're my new favorite RU-vidr.

@robbraxmantech 3 месяца назад

Are you a new subscriber? Please let me know what content you are interested in.

@lukasbruderlin2723 3 месяца назад

@@robbraxmantech I also started to follow you more closely. I believe this AI series will be super important. I would appreciate if you continue taking us on a well educational journey on how to use AI in a best way to solve the PRVCY issues. As far as I understand you will teach us on how to use smaller LLM models directly on the local machine and also explain how to start training it according to specific needs. While I think this concept would probably be the best for PRVCY, I think there are also other approaches to at least somehow mitigate the problem, like there is at least one privacy oriented AI Tool that I found on Github. He uses a freemium model and I think that tool is mixing queries from dozens of users, so the AI API will find it hard to analyse the interests or character of one single person, but I'm not sure and haven't tried it out yet. Can't find the name right now - but to make my point, it would be really cool, if you would also find and discuss such AI Tools / APIs and share your thoughts on those with us... even you might no directly profit from such mentioning. Thank you.

@danielberg5908 3 месяца назад

Mine to! Can you do more step by step tutorials about how you set up this tests because i want to learn how to use embedded AI to use for myself. Its really fascinating how you can get also good things from it like a helper that can teach you everything!

@Growler-F18 3 месяца назад

That was a very good tutorial as to what is to come - scary stuff if you just stick your head in the sand!😮

@VertegrezNox 3 месяца назад

W00t! So hyped that you're covering "AI" topics like this!

@nulllogos5846 3 месяца назад

Thank you for all the work you do.

@BurgerMeisterGang 2 дня назад

I hear about companies data being hacked, but if its encrypted well, it helps. IT needs to also block mass data transferse , and only do them manually with authorization signatures, etc.

@Mr3X7R3M3 3 месяца назад

Great video Rob! Very informative as always

@reggiep75 3 месяца назад

I watch way too much RU-vid but I always like the content on this channel.

@dreamphoenix 3 месяца назад

Thank you.

@angelthymera2681 Месяц назад

Funny. This is how I imagined Google and Facebook worked from the start. Thanks for showing how easy it is to create and implement such a tool. This will be a prime example to naysayers.

@cue108 3 месяца назад

awesome work and presentation! Thanx a lot!!

@robbraxmantech 3 месяца назад

Glad you liked it!

@idcrafter-cgi 3 месяца назад

The Recall is a interesting idea and would be a great tool if it would have been focused more on security and safety like making it harder to get personal data by discarding it which a llm should be able to do.

@statinskill 3 месяца назад

AI is not your friend. AI is their friend.

@jtjames79 3 месяца назад

I was thinking similarly. I would want to break this out into a server. For the exact same reason you don't want to stream and game on the same machine. I'm building a server to make an agentic rag Lydia game agnostic minion AI, to carry my burdens.

@DaveS-hd9qu 3 месяца назад

I found it interesting the bear was identified but not the specifics of the blue flowers. There must be some mechanism to preclude too much detail during analysis.

@robbraxmantech 3 месяца назад

It depends on the run. You could ask it specifics about the flowers. In one run it said blue wildflowers. You could ask it multiple times without reuploading the image

@TheBigRat770 3 месяца назад

Just switched to Linux mint on my laptop. Its from like 2014 and windows was slow. Now its fast again

@johnyepthomi892 3 месяца назад

There’s lighter distro if you have older systems. The trick is to find a distro that doesn’t need a lot of ram for older systems. Anyways, great decision. Edited a lot of autocorrect typos

@Mr3X7R3M3 3 месяца назад

Mint isnt bad. Like bigrat said, there are even faster distros available if you only care about speed. That is, of course, at the detriment of “creature comforts”. I personally run Arch, and game on it, but I also recommend Pop! for new users as well

@frogmanjack 3 месяца назад

Mint crew reporting in 🫡. Been using Linux full-time since 2018. I try all distros at some point. But always find myself back home with Mint

@JimShingler 3 месяца назад

Wow, .... Thanks for making us aware

@huddunlap3999 3 месяца назад

Interesting content.

@miketurgeon5506 3 месяца назад

Thanks Rob....Very scary......You're fantastic

@turkeydoctor5546 3 месяца назад

the use of spam was a nice touch 😅

@robbraxmantech 3 месяца назад

Actually, my original intent was to see if the AI could detect C-SPAM (like CSAM. LOL)

@turkeydoctor5546 3 месяца назад

@@robbraxmantech we have to be mindful of what we teach it, or don't teach it.

@purdysanchez 2 месяца назад

I can tell you are an experienced software architect based on the fact the you considered minimizing the snapshot sample rate as one of the first concerns.

@beyondpc9421 3 месяца назад

If an AI is to be trained it must have a huge collection of quality data to train on. If an AI is trained to detect CSAM then the companies MUST have a huge library of CSAM. How are we not busting these companies?

@goodfortune6399 3 месяца назад

" agents" this really is the matrix

@BobFromPineCreek 3 месяца назад

That's scary. Why did people invent this stuff? I guess people want to live in that dystopian sci fi world..

@diotitus 3 месяца назад

It's all part of a grander plan to create a one world government that sees everything you do and hears everything that you say. They are gradually inventing this technology because they want to slowly condition society to get used to this kind of thing.

@mibo747 3 месяца назад

Excellent example how microsoft will get most bitcoin wallets and passwords....

@edmondhung181 3 месяца назад

So when you take a photo the AI agent in your phone can comment/opinion on the photos and send to big brother if the GPS location and/or content raise their suspicion. 😢

@Notarrusianbot 2 месяца назад

Exactly. I'm right now having this problem.

@davj1481 3 месяца назад

They should run AI on Windows only. That way if it gets out of control you can just force it into the blue screen of death. And it can't harm anybody.

@VertegrezNox 3 месяца назад

A lot of development for 'AI' tend to occur in containerized environments.. However it is always safe to consider such, if "AI" was misused like a badusb it could execute code far faster than a human's ability to read it, &/or likely do much to stop it (if it happened).

@phoneywheeze 3 месяца назад

what does that mean? How will AI program harm anyone?

@ericsbraun 3 месяца назад

Problem; AI will fix WinDoze errors, if MSlough lets it. They won't. It would ruin their business model.

@Chris_Fer 3 месяца назад

Now just gotta learn how to set up Private AI, on my device and how to use GPU acceleration, how to download dependencies like pytorch, Troubleshoot the godamn CUDA (Nividia I am looking at you), and find some safe to use scripts, which will make my life more difficult then it should be. Yep, kinda frustating, anyone else in this phase? Especially after how OpenAI, might be using all the data to train their own AI? Once the setup is done then one should be golden.

@phoneywheeze 3 месяца назад

I think for local recall-like feature, the standard would be just running a background script that captures SS and maybe appends OCR every few seconds. The multimodal AI should only be used for querying the result whenever needed so it wouldn't be running all the time

@Oldgreycowboy 3 месяца назад

I’ll break out my window 98 and see how that works.

@scotttovey 3 месяца назад

I don't think people see the whole picture with this tech. A person could be have spent months writing a novel, only to have a third party register that novel with the copyright office before they are able to do so. The end result is that the author that actually wrote the novel, looses the ability to earn money off that novel. The individual or corporation that used an embedded AI to track the author's progress and register the work before the Author could, does. Unless the author took the time to print out progress copies of the work, it is unlikely that the author would be able to prove he owns the work. This goes with that nifty new invention you just designed on your computer as well. Big tech is in fact using AI to commit corporate espionage, and claiming to be innocent because a human being is not watching the computer. This is actually a false argument as it took a human being to create the AI program to begin with. The AI is therefore an agent of whoever is using it, and if the user is working for the government, the the AI is by extension, an agent of the government.

@robbraxmantech 3 месяца назад

There are already cases where foreign powers have stolen AI intellectual property directly from companies like OpenAI

@robbraxmantech 3 месяца назад

The solution btw is your own Local AI and I'm really getting a lot of benefit from this kind of use. Will describe my setup and use in the next video.

@scotttovey 3 месяца назад

@@robbraxmantech It's sad that a person has to be able to track themselves on their own computer in order to protect against thieves. Years ago I heard about a company that sold a 3D printer. The machine would not run unless it could contact the company servers. I think that was a lease situation, but even still, I will not knowingly purchase or lease any equipment that needed to access the internet. My personal solution is a bit harder to do. Pass an amendment that declares an individuals right to privacy, that they own all the data regarding them, and that corporations are bound by the various pertinent amendments and may not share that data even when the individual agrees to them doing so.

@phoneywheeze 3 месяца назад

this is problem with internet and not AI. AI can be run locally without the internet, and isn't inherently bad. The problem is big tech wanting your data.

@Corteum 3 месяца назад

I dont think anyone wants a Recall thb, for the exact same reasons they dont want a Windows Recall ---- There's simply no way to guarantee that all those screenshots wont get hacked.

@anandrew6641 3 месяца назад

Unfortunately MS and the large portion of people will go with you and will use it, knowingly or not, and it'll be used against most users, to profile for Ads or other gov reasons

@Corteum 3 месяца назад

@@anandrew6641 That's the whole point of it. They made it too obvious. Pushing it hard to get more people on to win 11. updating ppls win 10 to win 11 without their consent. enabling recall by default, then only making it opt in at the last minute (even though it'll still be on your system). Making it hard for regular folks to disable. They do all the tactics, And it points to one thing - they intend to use it on people without people knowing that it's running in the bg.

@Notarrusianbot 2 месяца назад

I don't have any knowledge about programming but... is it possible to develop a new IOS through AI or to modify windows with AI ? Maybe we could gain privacy that wpay specially if we don't know how to program.

@TimMcMullen-bt6ck 6 дней назад

can these AI tools be used on my laptop without my knowing? do a video on this, Please!!!

@ScorpionCar 3 месяца назад

What a gem of a video! 👍 Btw what song is this called at 5:33

@3aeren 2 дня назад

darude - sandstorm

@ScorpionCar День назад

@@3aeren Nope

@AnthonyPrince01 3 месяца назад

Mabuhay

@boxcarhobo8315 2 месяца назад

This wasn't in my notifications. To answer your most recent video regarding direction, i wonder if you got hit by the "algorithm"

@disdroid 3 месяца назад

Thanks! I'm going to throw my non-AI scanning bots in the bin now!

@cooljets 3 месяца назад

Several years ago, during a routine scan of my ex-son-in-law's computer, while looking for pirated software, the MS bot found an image that should not have been there. MS reported it to the feds and he ended up copping a plea for two years in federal prison. If you need privacy, for whatever reasom, MS is probably not the OS to use.

@ludwigvan8600 3 месяца назад

That's pretty disturbing. At least don't use Windows 11 in its current form, Windows 10 is still ok if Cortana is disabled and especially slimmed down versions made by third parties e.g. Windows X Lite or by using certain apps yourself but time consuming.

@cooljets 3 месяца назад

@@ludwigvan8600 I have not used MS in any form for over 20 years. They cannot be trusted in any form. Open source isn't perfect, but at least independent eyes can check the code.

@benisrood 3 месяца назад

"Image that should not have been there"? Do you mean CP? Because as far as I know that is the only case where this would occur.

@l0I0I0I0 3 месяца назад

Why not use AI to stop AI?

@robbraxmantech 3 месяца назад

In general, you cannot control something that's embedded in the OS. Just stop using that OS.

@LeeGee 3 месяца назад

@@robbraxmantech My problem is that industry-standard music composition/performance software does not exist to any useful degree on Linux.

@zhivkotodorov4323 3 месяца назад

@@LeeGee Just us a VM on Linux to run your software. You might have to upgrade your hardware, depending on what your current set-up is

@l0I0I0I0 3 месяца назад

@@robbraxmantech Ty! Do you have any vids on how to install win 11 without spyware or bloatware? I'm open to other OS but concerned my games won't work on it as this specific computer is for games only.

@fkxfkx 3 месяца назад

Why not, indeed

@JDM0bsession 3 месяца назад

Could you make a video explaining Life360 and if its bad and if there are any good alternatives to it that are safer?

@robbraxmantech 3 месяца назад

Dump it. If I made a video about it, I would be attacked. It is useless

@JDM0bsession 3 месяца назад

@robbraxmantech Do you have any recommendations for alike apps or would you just avoid all together?

@mjmeans7983 3 месяца назад

AI bad grammar and hallucinations abound.: The beginning of the second sentence, "They are holding up one hand" is grammatically incorrect within the paragraph structure. The paragraph structure begins as framing a single person that appears to be in a protest or demonstration implying that there are other people. To the focus of the rest of the paragraph it would be either the single person, or multiple people in the group. To further identify the subject of the paragraph the next line should be "The person is holding up one hand" (which narrows the focus to that person), instead of "Several of the protesters are holding up one hand (which would broaden the focus to the group). Since the following sentence is also about the person mentioned in the first sentence, then the former option is the correct grammar for this paragraph. Because an alternate beginning of the first paragraph could be "Their hand being held up" would also be consistent with the rest of the paragraph, I think the AI has made the elementary school error of confusing "their" and "they're", and the sentence grammar parser probably replaced "They're" with "They are".

@robbraxmantech 3 месяца назад

This is Llava model btw. One time I think a model argued with me about the grammar, saying it was fine to say "a team of AI researcher" (singular). Could be Llama3. Can't remember.

@Jerry-yk1fh 3 месяца назад

Thanks Rob, but it's way above this 72 year young man. My wife and I both have a brax 2 euro phone and I'm using mine but don't know what I'm doing.. Can I use a degoogle phone as a tablet to watch y/t? Also do not know what apps are safe for the brax2, there are a few podcasts I would like to listen to and what about rumble, locals and patreon? I'm using a normie phone for this functions via wifi right now . Will be subscribing to your email and vpn soon. Thanks Ole Texas Coot Jerry

@JJFX- 3 месяца назад

Maybe we should just use a mini pc or VM on the network logged into MS that's entirely run by AI executing random, benign tasks to overwhelm them with meaningless data.

@phoneywheeze 3 месяца назад

it would just burn out your machine. It's not proven that they AI feature sends anything, and if it would send, like apple does, it would just be a small hash

@jakemelinko 3 месяца назад

Just wait till people find out that brain to brain interface has been commercially available since 2014

@shephusted2714 3 месяца назад

it is not really happening for open source ai right now - in 5-8 years yeah but not in the near future

@MistyB-yv1uw 3 месяца назад

What is Delphi?

@mydetlef 3 месяца назад

Thank you for demonstrating the technology. But, as always, there are two sides to the coin. Every organization, be it a company, government agency, or the CIA, always needs to make sure that none of its employees abuses its trust by spying or sabotaging. And if they do, the extent of the damage can be quickly determined. The same goes for children and young people. Many suicides, school shootings, or the foolishness of teenagers showing themselves naked on social media could be prevented if the signs and patterns were recognized early enough and the extent of the problem could be determined. Of course, on the other hand, it is clear that this is another good way of oppressing people. Whether in a repressive state or simply for a beaten woman trying to escape her tormentor. I would compare it to the invention of the handgun. That, too, was a blessing and a curse.

@mictadlo 2 месяца назад

Hi, Where can I find your code?

@robbraxmantech 2 месяца назад

I did not post the code as it could be misconstrued as malware. But like I said, I had the AI create it. Ask your own AI

@raymitchell9736 3 месяца назад

Creepy and concerning!

@bobtarmac1828 3 месяца назад

Ai agents gone mad. Is it too late to cease Ai? Remember Hansel and Gretal and that scenario? Luring us in with sweets? Just to be… laid off by Ai, then… suffer human extinction? Or worse, imprisoned by an… ai new world order? Enforced by swell robotics popping up everywhere… replacing you with Ai jobloss?

@thexfile. 3 месяца назад

Building an AI-Virus from the ground up.

@robbraxmantech 3 месяца назад

You can't do if the OS itself has embedded the agents

@non9886 3 месяца назад

go for it man! snowball ai with misinformation. if you create few ai agents as misinformation agents which can replicate itself and learn how to deal with anti-actions, it should feed ai quite quickly. and we will see if ai is real ai or there are backdoors and censorship happening there. we need some olgoy chorchoy which will eat ai from within 🙂 maybe that people will volunteer to host such agents in their comps and as a part of anti ai botnet...

@somepoliticalgamer6459 3 месяца назад

Don’t be a ping ping tom.

@scottwatschke4192 3 месяца назад

I think with this video, you're helping new hackers. get better skills.

@robbraxmantech 3 месяца назад

I'm not showing the source code. So no. But would you rather I not tell you what is happening on your computer?

@Skolotoi 3 месяца назад

I get it that you can build these systems yourself, but how would Microsoft spy on you then? Serious answers only 😂

@robbraxmantech 3 месяца назад

They already did this (and 'recalled' it) and this is the same methodology explained by Apple. Have the AI do the scanning and report only the results. Thus your data never leaves the device. More efficient. Just understand the seriousness of this. There would be no End-to-End encryption then

@Skolotoi 3 месяца назад

@robbraxmantech Dear Rob, I was only being sarcastic. I've been following your channel for a while, and understood maybe about 40-50% of what you are laying down, but it was enough to see that NPU = end of in-device privacy. I think that there is still hope for achieving privacy through external gadgers. What do you think of an idea of Augmented Reality wearables that will interact with corporate phones in such a way as to allow for encryption? Let's say it's 2028, and I have Telegram on my Verizon phone, but it's looking like a messy QR code thingy that is constantly transforming, and only through my AR glasses I can actually see the app the way one sees it on the phone today. Will need a 3rd party input device as well, but will potentially allow for encryption ran through corporate networks? Thank you so much for your wonderful work! ❤️❤️

@robbraxmantech 3 месяца назад

If you use Linux (as I did in the video), the only AI is one you add yourself

@linsqopiring6816 3 месяца назад

@@Skolotoi It's kind of a cool idea, but it implies that you trust the AR glasses more than the phone. As soon as people started doing this they'd be sure to implant spy modules and back doors into the AR glasses firmware itself.

@boirfanman 3 месяца назад

An NPU is not needed. They are just to speed up the calculations.

@zwarst 3 месяца назад

A eye cannot watch Linux...

@mediocreape 3 месяца назад

fascinating video!

@hybridinfodesk409 3 месяца назад

Copiliot was developed for developers. Not end users. Not a fan of AI. Tested a opensoure one and interrogated it. Ask what was the difference in the different AI models coming from manufacturers. Copilot is for developers so they will be developing tools for their usage, defined and undefined. Gemini is an advertising AI with the sole purpose of making advertising easier for Google's advertising imagine having a cpu AI chip. Siri, Alexa, Bixby and the rest are all AI. They just threw in assistant .

@h.766 3 месяца назад

You sound liked bot. ,😂😂😂

@hybridinfodesk409 3 месяца назад

@h.766 I test drove an open-source GPT, and that's what it said. I reset my case. You just said AI answers are unusable and nonhuman. Lol 😆

@Whit3hat 2 месяца назад

Hi, okay, this is weird considering local AI is supposed to be private. So I am new to Linux. Basically, Llama3 helps me to learn my way, about 8 or 9 days ago, we had convo to help me setup Snort on Ubuntu, today I wanted to check out its log file and asked Llama3 on its default location, and which search cmd to use it supplied me with I can use to search sudo grep -r "snort.log" amongst the result was text that I recognized as being the convo I had with it help to setup snort why on earth would this be saved? When i asked it why it evaded a direct answer mentioned it must be convo i had with snort?? I pushed it for a answer and stated I cannot provide information on how Meta's conversations are stored, sooooo it does store it !Are you guys aware of this is this normal? These would be located in your home directory /.ollama/history this looks like a windows recall trick to me