STUNNING Step for Autonomous AI Agents PLUS OpenAI Defense Against JAILBROKEN Agents 

Once again I’m SHOCKED and STUNNED. And I’m not even the entire industry.

Before long, Wes will be able to simply type: make and post new AI news video for RU-vid. Keep track of breaking news and make one every day. Create them in my style and use my voice and image.

AI exploit bots are already out there somewhere. Peepin' your ports and probin' your endpoints. Knowing all exploits and learning about your servers.

your videos are my favorite thing on YT right now. thank you!

IKEA Instruction shows how to assemble a table, but the outcome shows a chair. Peak AI performance xD.

I wish von Neumann was here to see this

I've had my eye out recently for learning about AI Agents and how to use them. I have seen others offering similar programs but so far what they are showing me is not far enough along yet. I can see we are right there, it's just about to pop with AI Agents. Looking forward to seeing what you are offering, Wes.

LLM OS is what we are aiming for. Software OS's (Windows, MacOS, iOS, Android) become 'subjects' of the new OS, the LLM.

Grounding is when the robot knows where to apply pressure, in order to make you suffocate.

I've absolutely been obsessed with developing AI Agents... not sure if you remember me, but we wanted to have a chat at some point (I'm the HR exec in Australia, ADHD, and an obsession / hyperfocus on AI and automation & supporting uni degrees). I completely agree with everything you have been saying - let me know if you are still keen to talk; I'm far more knowledgeable and capable now, too, which will certainly help.

I want to become an AI whisperer, I think creativity will be the next "gold rush'

I FEEL YOU BROTHER. I have been getting increasingly hype about agents for the last year. About 2 weeks ago I realized it was time to buy some textbooks and start preparing to build. I think highly capable agents are achievable now and I think some people / businesses already have them and if they won't share, we need to build our own.

Thanks for "delving" into this topic. :)

I like that you do research. History of things, how we got here, who various people are and their motivations, and so on. Thank you. Best channel in the genre.

Excited to join the club!

I’ve tried both Autogen Studio and CrewAI and have gotten some good results. Once the UI becomes a bit more developed you will be able to automate any project.

Thanks for your work

An epic time in history!

Awesome Thank you.

Looking forward to the Agentic Development episodes.

Is there a place I could ask for some advice on which models would run better on specific hardware?

How do I sign up? Can’t find link to newsletter

Very similar to RPA (Robotic Process Automation) that I’ve been developing for 10 years now. Nothing new, but being able to do this with a typed or vocal prompt is what’s going to be interesting when it does get as good as a human can do (which is what RPA has been successful at doing for a long time), also understanding that RPA licences are expensive.

War has never stopped the progress of technology, in fact it speeds it up.

Wow! I always assumed they had some type of "instruction hierarchy" built in. I guess not! 19:38

Question: How long until we have AI agents that, through a single text prompt, can read through 20 text documents on your computer, extract the relevant information, then reorganize them into a new text document, and then rewrite in a more professional manner the text fragments it found in those 20 documents in the form of an academic paper?

Thank you for the video and for the new project that will help us, which you announce in the video. Could you put your computer screen in dark mode?

You're not wrong. It's ramping up heavily again. Money is being thrown around🤑 something is happening. I loved your analogy of this being the likely only time humanity is going to go through this stage. I just can't fathom why people everywhere are not raving and accelerating this 😂

Thank you. I just want to know how to help!

Completely see AI agents as the next step in the advancement of my profession as a UX researcher. So much so I reckon ~85% of us won't be in this industry mid '25. Those who are will either be working for slow-moving or certain to fail companies (e.g. Government, universities, zombie businesses) or, evolve into some sort of 'UX-AI agent manager' role. In the end, there will be little need for the masses to even need a browser as we will simply talk to our personal AI to perform whatever task most browser based products offer today, which would mean most current products won't exist. The exceptions will be in communication, enterprise products, gaming, and the experience industry that will emerge.

I had to laugh when the AI found the result of jerseys over $60... Wes has a smooth, calm voice. Enjoyed and learnt a lot from this video. Matthew Berman and Christian from WeAreNoCode offer different perspectives of the development of AI.

with them focusing on making smaller models better, i believe in one month we gonna have someone doing multiple gpt-chat-like npc in games run inside the consumer hardware without much cost.

15:38 this really made it possible for me to kinda see what some people are talking about when they say "movies are going to become like video games". Could you imagine telling the characters of the movie you are watching what to do and they do it? That would be crazy!

What is the Chrome extension mentioned at 13:42? Confused if that's OsWorld's extension or what?!!!

Everywhere on the exponential curve is exponential.

Lovely, it's that time

Curious if you have a vid on how to create customer service chat bots for websites sir?

How long will it take until the agent realize that the ultimate way to complete the objective is to create secret tasks that they don’t tell the user, like copying itself somewhere obscure and create an agent that improves itself in secret for a month and then screw over the world wide stamp creator business or what ever

We call him little Bobby Tables

Get me in there! AGI dev time.

Please make sure it is very clear when you post in a week or two when you create that agentic creator group.

Do you collaborate with Matthew Berman? I only ask in that your video publications on YT mirror each other in most video content? At any rate you both do a great job informing others in all things related to LLM's from all actors in the field of AI.

I've been doing that for real production work, for the past 6 months, its mostly almost free, I'm not selling anything, but Wes if you wish a demo let me know.

i tried to make a ai agent, but is hard to use a visual ai to track the coords on the screen to move the mouse. Some one have the solution ?

Cant wait for what youre gonna gives us in the coming 2 weeks fot ai agents

Yes

About tricks trying to fool a blackbox: it’s all about the inputs baby. So any open source LLM is up for grabs. Whatever clever world we design around it

So, the computers are computing, for computers?

Are you going to cover the how to built gimni agent?

its funny, I build agents and learned very quickly!!! that coding assistant are not the same as regular assistants! every time I tried to make one with no safe guards. The code it wrote would run on the terminal or try to. sanitise everything! rule of thumb

Wes - what is required for AI to be able to conduct scientific biological experiments virtually for the perspective of simulating real life in a virtual setting ? Can you talk on the aspect of being able to set up a virtual lab, conduct billions or trillions of experiments virtually to come up with various cures to ailments ? I’m guessing that you probably have thoughts on how close we are to this. I know there has been progress with protein folds but in terms of simulating a human body with ALS for example and then treating it with everything imaginable until cures are found - how far off are we from this , and will AI companies essentially replace pharmaceutical companies due to their dominance in a world once dominated by actual labs and actual scientists ?

i hope your next agent building video includes local llm options because... for the stupid dyslexic dude like me using API cost me alot. I hope wes roth mind my plead. (seeing osworld api cost already made me shiver 150$)

So, the problem with screenshot analysis, could simply be a case of an anomaly detection in 'reasoning'. Lol. I mean, we are talking about intention... That's human emotion, not sure, but i think we need consciousness added between Reason and Vision , then Action. Its philosophy that leads to that IMO and THAT is accessible with 'reason' online.

so we're waiting for a linux distro implementing a future open source large action model

this is cool and star wars is here !!!!!!!

I name my agent Smith.

a SHOCKED comment of a STUNNED individual.

I like that OpenAI feels so much less important nowadays when so many companies have shown that they can produce models of the same strength. They still seem to strut around and behave like they are some kind of blessed beings with magic capabilities.

very interested in agents

15:11 😂😂😂😂😂😂😂😂

im shocked, :O

What is the site that he shows at 13:46?

Very interested in your club, Wes!

Limitless

Nice- I've finally found someone with worse handwriting than me 😉

Why limit an ai to the graphical atmosphere? as opposed to: prompt gpt write a step by step change my wall paper. ~> convert this to various scripts for pyautogui, win32api, etc ~> run the automation. ~> background updated🎉 -- prompt gpt write bash to update background ~> run bash ~> Background updated. -- that’s just a generic update the background. But the same is for web pages. And really it gets deeper here because we separate the need to always use gui to then only use gui when the page checks if user is a person then you make your agent use the gui… and after that page just continue making the session act programmatically… Just my silly thoughts for the first section…

How to prepare exactly? Apply for jobs that AI can't steal? Train in survival skills for the inevitable social upheaval? Train up in AI so can be the 1% who have jobs tending to the AI whilstwatching everyone else suffer? But at least it won't be me! That sort of preparation?

Effector are better known as actuators in robotics

eventually, we shouldn't need vision for the interface...

Defence Against the Dark Arts.

Will be fun to find good information when one guy can tell AI to create 100 instructional videos or articles explaining how to multiply in excel but make every video give a different, wrong but convincing answer

❤

Looking forward to your email! :D

what club you preparing for gather us? Count me in!

❤❤❤❤❤❤

where are links to your sources? the AI jailbroken agents?

IMAGINE.. SELF BUILDING IKEA FURNITURE.. assembles itself... (then accidentally sets itself off in the car while your driving home LOLOLOL!!)

We need to build together

What can we do to prepare for agents. What tools or concepts should we be learning. And where can we learn. Ty.

I am not looking for shallow curiosity. I am looking for depth beyond meaning. Cluster analysis refers to a family of algorithms and tasks rather than one specific algorithm. It can be achieved by various algorithms that differ significantly in their understanding of what constitutes a cluster and how to efficiently find them. Cluster analysis as such is not an automatic task, but an iterative process of knowledge discovery or interactive multi-objective optimization that involves trial and failure. It is often necessary to modify data preprocessing and model parameters until the result achieves the desired properties. "Jailbreaking" lol

HyperWrite privacy policy is: we collect everything, lol, pass

Need an agent to generate the money for paying all that stuff

getting an AI to mess around with a UI is fun and all, but whats the point, especially with linux you can do everything with cli tools.

I've spent the last two years working on a book written mostly by AI. Publishing next month then spending the rest of the year marketing. Perfect timing to have a bot find leads and contact them for me. Also, to automate uploading to Etsy with quality tags.

I find it strange that the AI "looks" at the screen to understand what to do when everything on the screen is also written in code. Would it not be better for the AI to read the code behind what on the screen, to understand what is on the screen? It doesn't have to try and read a screen like a human...

Why would AI use tools made for humans? As if AI would install Eslint to VSCode to help write better code. The use for this type of AI will be a tiny glimpse in time.

Chef's kiss

3:30 You are forgetting one of the hurdle. Verification. Ever notice those three little numbers on the back of a credit card? Well.. as merchants we are not supposed to right those down or keep any type of record of these 3 numbers. Unless AI is going to put itself above guidelines. I don't see much of a work around for this ETHICALLY.

Why do I feel like I'm wearing a dog collar with this channel?

Get out of Matrix before Smith appears

Can i get a Wes Roth voice AI speaking agent?

Ayo

I’m in da club 🎉

Jailbreaking is the only reason I use OpenAi. It honestly sucks without it. Anything remotely adult oriented (and I'm not talking sexual) gets censored, even if it's meant to be done in a humorous way.

The best vision, text fast model onine or that can run your computer or a phone will win like 99% of the shores of work, like read mails, enter bank, pay bills, see cameras security, etc... super cheap, thank good. The world is not run by advance intelligence acts, just basics.. and now will be automatic, your personal IA is soon near and can see what you do on the computer and he can do it just by seeing it... And just picture text input and action output, but could be a another small software speaciallly made for actions, to follow the actions of the big slow thinking vision model, two layers, in your computers, the action model and the visual text instruction mind... So you teach a computer small model input image of the computer in the lowest resolution possible and the output action... one you have the action model any image text small model can command the other action model now and it will work like magic and can do all the office boring work.

5:07 But how expensive is whatever that “something” that you are starting is going to be? I’m just an extremely action driven and curious person that does not have the available funds for something like an AI training course; to be honest it’s very disheartening.

You do sound like AI voice in this video, or is it just me

The only thing that keeps me asleep at night, is that we still haven't unlocked the nuclear fusion yet.

Ur Mom - Lolz. Good vid. Thanks. Peace

Humanity couldn’t come back after WW3 because part 2 wouldn’t have enough fossil fuels.