Open AI's New Statement " EVERYTHING Is About To Change" (Agi + Agents)

So there's been a statement that I really just read that shocked me to my core because this is no word of exaggeration going to change everything. Take a look at this statement right here. Last month Ben Newhouse, an OpenAI employee who has worked on computer using agents at the startup, according to a person familiar with his role, posted on X that he was hiring for his team and building what I think could be an industry defining zero to one product that leverages the latest and greatest for from our upcoming models. He didn't elaborate. Peter Wellinder, OpenAI's vice president of products, remarked on X that the product Newhouse described will change everything. That is the remarkable statement that is the basis for this video. And you really need to watch this thing until the end because there is a lot to unpack because the next couple of months and even the next couple of years with what OpenAI and some of the other companies in competition, including Google Meta and a few others that we will dive into are building, is truly, truly staggering. So that excerpt comes from this article on the information that states that OpenAI has shifted the AI battleground to software that operates devices and automates tasks. And this is something that we did know was coming for quite some time because recently we did do a video where we did a deep dive on agents and Bill Gates entire philosophy on what he thought about that. Now I may include some snippets of that later on in the video, but it is a 33 minute video in which I dive into the future of society. And the reason I say that, not to exaggerate, is because if there is a device that can pretty much control any computer system that is going to be automating large parts of the economy, even according to some people at OpenAI. So now let's look back at this article. Let's truly look at what on earth is going on. Because if OpenAI of are shifting their scope to the AI battleground within AI agents, there are going to be major industry changes. So the article states that there is going to be device takeover. So OpenAI is developing a form of agent software to automate complex tasks by effectively taking over a customer's device. The customer could then ask the ChatGPT agent to transfer data from a document to a spreadsheet for analysis, for instance, or to automatically fill out expense reports and enter them in in accounting software, those kind of requests would trigger the agent to perform the clicks, cursor movements, text typing and other actions humans take as they work with different apps. According to a person with knowledge of the effort. So essentially, if you don't know what these agents are and what they're going to be doing in the future, they're going to be simply taking over your device. You know how when you sign into computer, you click here and you go here, that is exactly what these agents are going to be doing. Now there are some examples of these agents which, which I will get into later. But the first part of this video, I do really want to do a deep dive on exactly what OpenAI and Sam Altman are doing at OpenAI when they are developing this next generation product. Now what I did find interesting about these agents that they did talk about was the fact that there are allegedly two types of agents. So they stated that the software Inworks is one of two types of agents OpenAI is developing. And as it jumps into one of the hottest areas of artificial intelligence, which could soon include Google and Meta platforms. Now, I don't know what kind of, you know, different kinds of two types of agents they're developing, but I do believe that we will gain some more insights as things do develop. As we know, multiple companies are working on this because this is by far the most important problem to solve when we are looking at the next evolution of AI in terms of automating workloads. But I do think that OpenAI Shifting their perspective is quite fascinating. So in addition, they also did talk about they're developing another class of AI agent that would handle web based tasks such as gathering public data about a set of companies creating itineraries under a certain budget or booking flights, said a person with knowledge of that effort. And Google and Meta have said they are developing similar types of agents. Now I want to focus on the bit highlighted in yellow handling web based tasks such as gathering public data about a set of companies creating itineraries, yada yada yada. So this is something that I do want to take a look at because we know that booking flight tickets, you know, creating itineraries under budget, all of these things are rather quite tedious, although it might require someone intelligent to do so. We do know that these are tasks that we wouldn't rather do and we don't particularly enjoy. And if OpenAI is developing an AI agent that is able to do that, I think that that is going to be something that is going to be very, very, very economically valuable and of course very disruptive. ChatGPT was pretty disruptive only to certain writing industries. But with AI agents on your computer, it's not as if we're going to get work that is going to be reviewed. It's going to be a task that simply no one's going to know whether or not an AI system is doing it or not. And that's why I believe that handling these web based tasks, unlike writing an article which we can almost always tell is AI generated, doing something like that is going to be providing immense value and there's going to be no kind of human intervention that needed unless the model fails. Now something I also did find quite fascinating as well was that Google and Meta are silently working on this. So it says here that Google and Meta have said they are developing similar types of agents powered by conversational ally known as LLMs. Now I think that Google and Meta have been pretty remarkably silent about their developing of their agents because potentially they're working on something that could change the game as well. And, and also they probably don't want to give away any kind of advantage that they do have. Because whilst LLMs are good, we've even seen that the recent Gemini advance now is on the level of GPT4 and there isn't really so much difference that there means to be a reason to switch from one over another. It really does seem like we are hitting the very, very upper limit of what is possible with LLMs in terms of everyday usage. But agents are going to be the very next evolution of, of this kind of AI software and we are witnessing real time this evolution take place. Now of course, like I said, Google and Meta developing this, I'm not sure who's going to be the first person to do this, but we do know that whichever company does deploy, it will have a huge competitive advantage because they're going to be able to get to the market first. Because as of right now, there isn't really a good agent that works effectively that we can use on a day to day basis. So it is just a matter of time because before that seal is broken. Now in addition, they did state something about a super smart assistant. So both forms of agents OpenAI is developing could help CEO Sam Altman turn ChatGPT into what he has privately called a super smart personal assistant for work. They may also bring his company into more direct competition with Microsoft, which is also using OpenAI's LLMs to automate features of its enterprise apps. So, so they can help people quickly create new documents or draft email responses. And I guess kind of, kind of the kind of agents that we have sort of seen are things in Copilot where you can essentially create a PowerPoint presentation in around a Couple of seconds, I will have a few demonstrations on screen. But it is something that does seem rudimentary in nature. But whilst this is still in its early phases, it does show us the direction that we're going to be able to get to. Whilst Copilot right now might just be a browser add on for Microsoft Word on all these other applications in the future, it's likely going to be the largest thing that we do use when we go to these applications. And I can't imagine that other AI companies are not going to be developing some things. Now, here's where things start to get interesting. If we take a look at this, we can see that they talk about a new operating system. So LLMs underpin productivity. Chatbots such as ChatGPT and some at OpenAI also view them as having the potential to be a kind of operating system, a including for personal devices, because of their ability to write code, make sense of images and retrieve files. And agents could further buttress this potential. Now, one thing I really want to focus on is of course, the new operating system. And as luck would have it, there was a recent company that kind of developed their own internal operating system foundation model for their new product. You've likely heard of it, it is actually called the rabbit R1 device. This. This is of course the first AI agent device. Now, if you don't know what this is, this is a product that's been rapidly selling out after its major debut and hitting the markets with a very positive reception. Now, essentially what they do talk about and what they do showcase is something called lams or language action models or large action models. I'm going to show you guys the video so you can completely understand why we are about to enter the next stage in evolution. But after you see that, I'm then going to show you guys why. Why Rabbit is absolutely crazy in terms of what they're going to be able to do. And if their proprietary software is as good as it claims, it's really going to put some spanners in the works for OpenAI and other large conglomerates, because maybe they've managed to corner the market in the AI hardware, which has hardly taken off just yet.

Large Action Model, or as we call it lam, is a new foundation model that understands human intentions on computers. Through it, we can teach Rabbit OS how to use specific applications. In this video, I'm teaching a Rabbit how to book an Airbnb while I'm operating normally as a human on the left screen, watch closely on the right as the large Action model is learning all my inputs and imitating my behavior in real time. So I'm trying to plan a trip to Barcelona with my wife and my daughter. The first thing I'm going to do is navigate to the anywhere option and I'm going to type Barcelona in the search field. The system is suggesting Barcelona, Spain, which is exactly where we want to go. Using the website's calendar tool, I'm going to mark our check in on the 15th and check out on the 21st. Now I'll click Add Guests and adjust the members accordingly. Now let's hit the search button and.

See what pops up.

Since we love the beach, let's make sure to select the beachfront option. And for a more private experience, I'm going to select entire home so we have the whole place to ourselves. For the budget, I'll set a maximum at 400,000 won and a minimum of 100,000 so that all the options are within our price range. We're going to need at least two bedrooms to make sure we all have our own space. Finally, with all of our preferences set, we've got plenty of units and does not require access to an application programming interface, also known as an API. Nor do you need train each workflow once. Let's try to use Rabbit OS and instead book a room in London. My extended family is going to London. It's going to be eight of us and four kids. We're thinking of December 30th to January 5th. It's not set in stone yet, so I just want some general options. Can you look it up for me?

Sure, I can help you with that.

And I think you guys do get the gist of the entire video. But things are really going to change with computers and operating systems because likely with what we're seeing here with Rabbit, this is truly where things are headed and it really does make sense. I mean, the average person trying to do something on a computer, it actually does take quite a bit of learning, especially with all the online tutorials for various different applications. Imagine if you just had a system that you could easily ask exactly what to do and it knew how to do it straight away, out of the box. You wouldn't need to learn pretty much anything. You would just ask it to do something and you'd immediately have the thing you needed. Now, if you weren't familiar with how exactly the device actually does work in terms of what it kind of does Agents. This kind of agent is going to help you quite a bit. I'm not going to play the fourth thing, but I want to just jog your memory on why this is really, really Next level agents.

It has caught the eye of many open source projects and productivity software companies. What remains to be solved is for these agents to perform tasks end to end, accurately and speedily. The problem is forcing a model to perform a task it is not designed for, whether for a language model to reason about webpage using super prompts or screenshots. At a philosophical level, if we can make an AI trigger actions on any kind of interface just like a human would, we will solve the problem. This insight led us to create the large action model and of course this.

Was a very simple description on exactly what they've done. Now there's a lot more to unpack here because trust me, this thing continues to go on. So another thing that Sam Altman OpenAI have done is is they've collaborated with some former Apple employees on of course, things that are very compatible to the R1 device that I've just previously discussed. And essentially what they're doing is creating an AI hardware project and I'm wondering how good this will be considering that Sam Altman is involved in so many different things due to the rise of OpenAI. Now essentially this is a little bit different because what this screenshot here is showing you is Outman's ties to one consumer hardware startup which is called Humane, and surprisingly I didn't actually know this until I did some research for this video, but that he is one of the largest investors and the company was founded by both former Apple employees and it plans to develop a screenless wearable device that can be directed using voice prompts and gestures while projecting information onto surfaces in front of it. Now, this device is available for pre order and there was something that I guess you could say didn't go down too well. They did a video demo that many people did kind of take the mickey out of on Twitter. But I'm thinking even if Humane's product didn't go as well as R1s, is Sam Altman going to be able to get this done? Now I'm going to show you guys a very small segment of the Humane thing because it didn't really get as much coverage. But once you see how Humane's device is, you're going to see potentially where Sam Altman and this former Apple employee are going to be taking their next device potentially on a different company. Now, just to be clear, the Humane company that I'm talking about isn't a collaboration between Sam Altman and Joni. I've this is a Completely separate investment that Sam Altman has made into a company that former Apple employees have done. And Sam Altman is looking at a new venture with Joni.

I've just reached two pieces. This is the Humane AI pin. It's a standalone device and software platform built from the ground up for AI. It comes in three colorways, got Eclipse, Lunar and Equinox. There's two, a computer and a battery booster. The battery booster powers a smaller battery inside the main computer and this is how we achieve our all day battery life.

So if you ever exhaust the booster, you just reach into your pocket or bag and hot swap it. This is a perpetual power system that allows you to use your AI pin for as long as you want to.

There are no wake words, so it's not always listening or always recording. In fact it doesn't do anything until you engage with it and your engagement comes through your voice, touch, gesture or the laserink display.

It also comes with its own connectivity built right in our own Humane network connected by T Mobile.

We were able to pack a lot of technology into something really small. It runs a Qualcomm Snapdragon chipset and that means it's really fast which is crucial for AI.

It's got an ultra wide RGB camera, depth sensor and motion sensors.

And it's got this really special speaker that creates this bubble of sound. The Trust light indicates when its input optical or audio sensors are active, ensuring full transparency and data security.

Now that was a small segment of the Humane's AI pin. There is a lot more to unpack there. But going back to what OpenAI are doing, which is rather fascinating, one thing that they did talk about was stating that OpenAI's plan for their agent, not one that's on a of course physical device like some kind of hardware, will operate in a personalized fashion and respond quickly in the way that Apple Siri does on the iPhone. The prospective OpenAI computer using Agent may need to be partly stored on users devices, which is rather interesting because we haven't seen that from many of the AI stuff. I mean the only thing we've really seen is that AI moving forward could potentially be stored on devices due to the fact that we are being able to make them more efficient. Now they're essentially they're saying that the company may also need to get permission from users to train the software on personal data such as an individual's emails and contacts. And I don't think that's going to be a problem considering that most people do want to do less work Anyways, but here's the thing that did make me quite surprised. And this is the statement, of course, last month, Ben Newhouse, what I talked about before, an OpenAI employee who has worked on a computer using agents at the startup essentially was stating that they are building what I think could be an industry defining zero to one product that leverages the latest and greatest from upcoming models. And he didn't elaborate. And of course they stated that this will change everything. So even if you think that they don't, the people that are working on this technology certainly do. Which means that whenever this product does launch, we could definitely see another mass engagement in this kind of technology where people start to realize the true ramifications of AI. Now, these ramifications are so crazy that certain people are even leaving Google. Okay? Some longtime AI researchers have been leaving companies such as Google to start companies that develop AI agents. Which goes to show that this is a very, very, very big thing. Now, Sam Altman has actually talked in a few interviews about what he thinks agents are going to be like in the future.

The answer will be like, start slow, start and you know, set the training wheels very low. And then as we understand the risk surface and where it's okay and where it's not, we can make thoughtful, risk adjusted decisions about how to move those training wheels up. But the upside here is going to be tremendous for people. If you think about having an agent that you can give a task to to help you with whatever you're trying to do and be confident it'll go do it, that's awesome.

So the question here is, where are we with agents now? Like, where are we with agents? Like, if you wanted to go ahead right now and go and download an agent and use it on your computer, where are we? Well, the thing is, OpenAI truly does have some serious competition and you really do need to take a look at them because a lot of people are not paying attention to some of these companies. And there are two in particular. Of course, the first one that we already previously discussed was the first AI agent device which is sold out. And we are waiting for this to hit the market. But there are two other companies. Number one is Imbue. I'm not sure how you state that, but IT is a $1 billion valuation company that has recently raised $200 million to, to build AI systems that can reason and code. And they stated that they're excited to announce their Series B at a valuation of over 1 billion with participation from companies like Nvidia. Now, in addition, this is Their video, which I think you should watch, because this talks about how they're going to be building their agents. And potentially we could even get these before OpenAI's.

My name is Kanjun and I run Imbue. We build AI systems that can reason so that they can help us accomplish much bigger goals in the world. I've always believed that technology can liberate people. If I were born in 1900, as a woman, I'd be spending all my time preparing food and stoking fires to cook it and mending clothing. Because we have the refrigerator and the loom and the dishwasher and all these technologies. I'm free. Well, actually, the modern day is not that different from the 1900s. Back then, we didn't have machines to do physical labor. And today we don't really have good machines for doing mental labor. We have computers, but they need to be micromanaged. It's hard to even imagine our computers accomplishing larger goals, like planning a family vacation or making a PowerPoint deck, without us basically doing all the work ourselves. But we're starting to see glimmers of that possibility with these large models that are behind chatbots and generative AI. Today's models are actually really exciting because they surprise me with how good they are sometimes, but I really do not trust them to do anything bigger. They just generate an output like text or an image or button click, and then they put the onus back on me to actually do stuff with it. Trying to do is actually get to reasoning models that we can ultimately build agents we can trust on top of all sorts of agents. It's hard to see because we're so in IT today, but AI models are like the very first electronic computer, which was just a calculator. What we're going to see is this explosion of enablement over the next 50 to 100 years. If we do things well, we have a world where we don't have to be glued to our screens anymore, where computers can help us thin the barrier between idea and execution so that as a person, I can go off and start exploring my curiosity. Discovering the laws of the universe, creating magical experiences for other people, understanding each other and loving our work. Creating those masterpieces inside of us that we're dreaming of. These things that feel very beautiful and very human. That's what I hope technology will enable for us to become much more human. I forgot what. I forgot what to say next.

Then there was another company that I do want to show you guys, and you probably have heard of this one, but this is adept AI. And adept raises $350 million to build AI that learns how to use software for you. And there was actually a recently good interview in which he talked about how they're actually going to do this. So this was an interview with the CEO of Adept AI, and it's important to take a look at a few of his statements here because it goes to show just how far ahead they are in terms of building.

Why don't you show me what you're building at Adept AI Labs, and thanks for the little diversion down history land there. So, super fascinating.

Yeah, for sure. So first, let me tell you a little bit more about what we're up to and why I'm really excited about it. And then we can quickly flip through some demos of a release we actually recently did last week, the North Star for Adapt from day one actually has been that in the long term, the thing that'd be the most valuable thing to build for work is an AI agent that does much more than reading and writing and drawing images, but can actually handle for you arbitrary work tasks and workflows. And those two things are very different. Reading and writing is not the ability for you to, say, be able to delegate your entire payments process to a neural network. In the latter case, what you really want is you want a system that knows how to use all the software you already have on your computer as if it were you. And in order to get there, you need to train these models that deeply understand not just the text, but also the pixels on your screen, and also what actions lead to what outcomes in the world. And so we've been hard at work on this, training this model that could do anything a human can do on a computer. And we've been building, effectively a product that enables knowledge workers to arbitrarily delegate tasks to the system. Here's a quick demo example. In this case, let's say you're responsible for paying invoices, and you get your plumbing invoice, you fire up a debt that pulls up the invoices, all being done by the model right now. Pulls up the invoice, reads the pixels in this PDF, realizes what it's about, stores some interesting facts about this, and then pulls up QuickBooks and then correctly enters, who is the payee savant? Plumbers. How did you pay? What category is this? And it realizes the category was never written in the PDF, but it realizes a plumbing invoice, so it should go into repairs and maintenance. And this task that you probably would have had to do 10 or 20 times a day for your job, you show Adept how to do it once. And now every time you get a new email invoice, you just fire up Adept and it handles this task for you.

It's also, where does Adept live? Is it in your system tray there? Or how does it intercept this? Coming in by via email? Because you have this invoice come in via email, you got to get it paid. You're in.

So after consuming all of that information about AI, Agent Sam Altman, OpenAI and the vast majority of companies tackling the problem, what do you guys think? Is this going to change everything, or is this going to be something that just makes a small dent in the ecosystem of AI? I think truly this will change the game, and we will largely be using these systems well within the next five to 10 years. So with that being said, I think.