Open AI Insider Just LEAKED GPT 4.5...

So what we're going to be taking a look at is of course some of the leaks regarding a GPT 4.5. And a lot of the information that you're about to see is most certainly going to surprise you because a lot of the information that is regarding GPT 4.5 was simply buried. And I'm here to bring some of that to light because it is rather fascinating with as to what GPT5 could potentially be. So if you don't know, some people on Twitter have been predicting earlier this year, well, not earlier this year, around maybe two weeks ago, that GPT 4.5 would be coming in December. You can see that this is a tweet from Jimmy Apples. If you don't know why I'm referring to someone who is on Twitter as jimmyapples. This is someone who has predicted the AI release of many large language models very, very accurately. Previously he did manage to predict the release date of GPT4 with 100% accuracy and it was only after the model was released that people realized he tweeted about it. And then of course people started following this account. So he's tweeted here. I'm more excited on some open source models coming out soon, but keep an eye out for a potential end of December GPT 4.5 drop and a new multimodal from Anthropic. So this is essentially giving us some information regarding GPT 4.5. And he was not only one of the people that did this, there was also someone else that that also tweeted about GPT 4.5 this year around two weeks ago. And essentially he also did tweet this. So I'm going to show you guys this tweet here. He said that Sama, what are you waiting for? Are you going to take the mitts off yet and release what you are sitting on or keep drip reading? So essentially this is in response to the fact that Google actually last week released Gemini and Gemini 1.0 was absolutely incredible. It did take the industry by storm and it has everyone talking about Google in AI again. So, so he's basically saying here is OpenAI going to let Google take the lead or are they going to sit on what they're releasing? And essentially what we have here is a very interesting situation because Jimmy Apples and I wasn't able to find the tweet. But someone else also who also actually predicted the Google Gemini release date on exact dates, weren't also confused. They actually said the same thing that Jimmy Apples was saying in the fact that GPT 4.5 was, would be released today and it might be released maybe while I'm making this video, maybe even just after I've uploaded this video. But I guess we're going to see because what actually happened today was rather, rather interesting. What you're currently looking at is apparently a OpenAI employee leak. So essentially according to some sources, and their sources are very, very unknown to say the least, this is going to be GPT 4.5. And you definitely want to share this around to kind of verify because not many people are really talking about this. So you can see here that that it says GPT 4.5, our most advanced model brings multimodal capabilities across language, audio, vision, video and 3D alongside complex reasoning and cross modal understanding. Learn more about GPT 4.5. So this, if it is true, is going to be absolutely incredible because this would be a large leap from GPT4 to GPT 4.5. Now there are some reasons as to why I do believe that this is true and although that this is speculation, I'm going to show you guys some points that actually show you guys why this could be actually true. So some of the things that you need to pay attention to here is of course language, audio, vision and video and 3D. Now I have seen some rumors about that, but of course I'm going to show you guys about those rumors in a moment. So remember the fact that this is a multimodal model which has audio, vision, video and 3D alongside complex reasoning and cross modal understanding. So you can see right Here there's GPT 4.5, GPT4.5.64k and GPT4.5 audio and speech and and then of course we have the inputs and the outputs. Now of course like I said, this could be completely fabricated. We don't know where this has come from. It's been circulating on Reddit, but like I said, we will also see within the next couple of days or even today if this is actually true. Now I'm about to show you some screenshots as to why I do think that there is some credence to this screenshot potentially being true. Just some. Not a lot, but just some. In case you guys were wondering. So this is the person who, who previously did tweet about Google Gemini. So he tweeted. The content of the screenshot seems to be largely correct, but none of my sources can currently verify whether it is a genuine draft. So this is the person who tweeted correctly about Google Gemini's date. So that's why I'm bringing this tweet up because this is someone who clearly I'm not sure how, I'm not sure where from, but does have some insider information on AI models being released to the public. Now, essentially this person also did state that I don't know what to make of it, it could be a fake, but I'm not sure. No one I know has heard of this draft, which of course means nothing, but be careful lest you fall victim to a quick endorphin rush. And yes, I know some of you are using my account in the same way. So essentially what he's talking about here is the fact that although this could be true, don't just quickly assume that it is true just because it's circulating, because these rumors are have been circulating quite a bit. But as you watch this video, you'll start to see why this rumor does seem more and more likely. So remember the screenshot about language, audio, vision and 3D across complex modal and understanding? So the reason that the screenshot does gain some credence is because if I do look at some of the stuff that I was just looking at when I was just looking into a lot of AI news, and this was some of the stuff that I didn't cover in other videos because I just thought that the speculation was just way too far fetched. But this is an article from the Information. Now if you remember, the Information was the place that got the first drop on many of the things that were going on in OpenAI. And if you don't know what I'm talking about, that's when OpenAI collapsed and then reforged again. So this actually talks about OpenAI dropped work on a new model called Arrakis, an AI model in a rare setback. So essentially it says late last year, around the time that ChatGPT became a global sensation, the engineers at OpenAI began working on a new AI model codenamed Arrakis. Although OpenAI was preparing to boost ChatGPT with a different model, now known as GPT4, which it had completed earlier in the year, the upcoming Arrakis model would allow to run the company Chatbot less expensively. Success with Arrakis would also help OpenAI show Microsoft how fast it could create successive LLMs, which would be valuable as the two firms finished negotiating a $10 billion investment and product deal by the end of the middle 2023. OpenAI had scrapped the Arrakis model launch after the model didn't run as effectively as the company expected. Now remember, this is Just discussing Arrakis, but there's also some other information that we did talk about. There were some rumors that there was a model called Gobi. Okay? And I've only ever heard about this model through Jimmy Apples because that was the only person I ever saw a tweet about it. But you have to remember, Jimmy Apples was the person that predicted many other different things from OpenAI. And I'm going to show you why they're so crazy. So there was this article, right? And this article essentially talks about OpenAI's Gobi. So it says, apparently two Reddit users claim they have access to two of OpenAI's internal models called Gobi and Arrakis, and they go way beyond of what anyone ever dreamed. This summed with some cryptic tweets of OpenAI personnel. Got the rumor real running it then shows us a tweet from Jimmy Apples where he says, AGI has been achieved internally. We did a video on this. Then, of course, you can see Sam altmand saying, Sure, 10x engineers are cool. That. But them, those 10,000x engineers, researchers, and I'm pretty sure a 10,000x engineer researcher is going to be some kind of AGI model that can do research for you at Lightning quickspace. All completely automated. Now, remember, okay, Jimmy Apples did talk about Gobi, and he did talk about Arrakis. And of course, we do know that the information source, the same place that got really good information on OpenAI's destruction, when the company was breaking up, did actually talk about Arrakis. They did actually talk about this Arrakis model. So that means that this information about Arrakis is likely true. Okay? And that also does give us some credence to the model called Gobi. And Gobi was actually apparently going to be a video 3D model. So let's take a look at some of this stuff. Okay? So it says, let's study the rumors to see what we can possibly wait for here or now. OpenAI is in possession of a powerful model codenamed Arrakis. It. It is everything to everything, all modalities to all modalities. Arrakis exceeds GPT4 capabilities and performs very close to human experts in many different fields. Hallucination rates are much lower than different than GPT4. Half of its training data was synthetic. We also did a video the other day on synthetic data. Should check that out. It says inference cost is around the same as GPT4 because of the conditional MOE multimodal weight offloading. And it's also a very, very Good autonomous agent and its release date is scheduled for 2024. So you can see here that we got so many different points with as to what this model is going to be. And this was something that many people didn't even see. Now, of course, this is of course something that is purely speculation, so we do have to take this with a grain of salt, but this information does pretty much line up to what we did just see. And everything to everything is something that is possible. Meta also did release a model like this where you can do everything to everything. I think it was audio to text, image to text, like you can just transfer in between. It was pretty crazy, but it's something that's still in its early phases. But this is something that Meta did do. Also, hallucination rates, we know that they're working on something like that. Synthetic data, we do know that Microsoft was working on that. And an autonomous agent we did see earlier with that previous tweet, a 10,000x engineer researcher could definitely be in reference to an autonomous agent. And in the other video where we did talk about synthetic data and autonomous agents, we showed you a video where Sam Altman was essentially talking about how autonomous agents are going to be the future. So definitely I'm not sure if I'll include a clip of that, but it is something that is there. So can it be possible, or is it possible that this is what GPT 4.5 is going to be? I mean, I guess we're going to have to see, and I wouldn't be surprised if that is going to be the case. Now, in addition to this, there was a clip of Sam Altman actually talking about the Future models of ChatGPT. And essentially he was on the Time interview and this was really cool because he was just talking about the past few weeks at OpenAI and what he said in this was pretty shocking because he said, as we get closer and closer to super intelligence, everybody involved gets more and more stressed and more and more anxious. And we realized the stakes get higher and higher. And Jimmy Apples quotes this by saying, we are further ahead than you think. This should be obvious by now. So the reason this is pretty crazy is because he's not talking about, are we going to get to AGI? Are we going to get, you know, a good level of AGI? They said that as we get closer and closer to superintelligence, everybody involved gets more and more stressed and more and more anxious. And this was in reference to OpenAI collapsing. So I guess what we have here is the fact that Sam Altman could be. And once again, let me warn you, this is pure speculation. We don't have any direct statements. The fact that he's saying that as we get closer to superintelligence, everybody gets more stressed and anxious and we realize the stakes are higher. And he's talking about the last few weeks of OpenAI have been pretty crazy. And what did we know happened the last few weeks at OpenAI that were pretty crazy? There were many people saying that the discovery could essentially change the world and there was this Q model and essentially now he's saying that as we get closer to superintelligence, things are getting more and more crazy. Does it not seem as if they already probably have AGI and are edging towards superintelligence? Of course, once again, it's pure speculation. And the problem with AGI is is that clear definitions of AGI despite Google DeepMind's levels of AGI paper are quite a struggle because some of the key AI researchers, including even some of the notable figures like Geoffrey Hinton and Yan, can't even agree on what AGI is or what an AGI system is going to be. The problem is, is that every year we move forward in AI development and every year it seems like AI development is just being redefined with us to what AGI is going to be. Because if GPT4 was released maybe 10 years ago, since some people would have declared it AGI and now people are saying this is not AGI, it's not smart, it doesn't know. So I think the problem is, is that we don't have a clear definition of AGI. But of course we will do a video on Deep Mind's levels of AGI because it does give you insight into what AGI could be. Now I will try and play a clip from the interview so you guys can see exactly what Sam Altman says. But it definitely is shocking now to.

Say something like empathetic. I think everybody involved in this, as we get closer and closer to super intelligence, everybody involved gets more stressed and more anxious and we realize the stakes are higher and higher. And I think that all exploded.

In addition, if you think that GPT 4.5 isn't real and you think that this is just pure speculation, that is completely fair. But something that did kind of shock me was the fact that this internal message slash memo from Google reveals that Google acted on the information that you. GPT4.5's release was coming today. Take a look at this. So you can see here that we have Gemini releasing Gemini Provision API and Gemini's API, because the GPT 4.5 release is apparently around the corner. So it says rumors circulating within the department has taken decisive action to address the potential impact of GPT 4.5. In response to these speculations, a strategic decision had been made to expedite the accreditation of of the Gemini API, effective as of today. This proactive measure aims to mitigate any unforeseen consequences and reinforce our position in light of emerging advancements. This is an internal memo tweeted out by Flowers from the Future, essentially saying that, look, we know that GPT 4.5 could be just around the corner. And I think if that is going to be the case, that is going to take away from Gemini's spotlight. So what do we need to do? We need to release this. Absolutely, as soon as possible. And you can see right here, and you can see right here that Gemini was actually released. Take a look at this. So it says, today developers can start building with our first version of Gemini Pro through Google AI Studio and aigoogle.dev. developers have a free quota and access to a full range of features, including function calling, embeddings, semantic retrieval, custom knowledge, grounding, chat functionality, and more. It supports 38 languages across 180 countries. And Gemini Ultra is coming early next year. We're excited so to see what you build. So this is why it's pretty shocking because people are now starting to build with Gemini Pro and their API. So, I mean, it seems like, you know, Google is rushing to get this out the door just in case OpenAI is about to release. Now, I don't know if Google has insider information, which is why they were like, look, we know that this model is coming out. We want to release this straight away. But we can see here that people have started to build some things with Gemini. Now for a link to this tweet right here. This. This was by a Twitter user. I don't actually remember the username, but I will leave a link to it down in the description. But essentially, remember how everyone was hating on Google for faking the demo. This essentially just shows that this is how Google's API does work. And this is a Gradio Pro. So it's not as quick because Gradio sometimes is quite slow when there's a lot of users. But it just goes to show how good Google Vision is with its API. So with all that information being said, what do you guys think about GPT 4.5? Do you think GPT 4.5 is released? Do you think in terms of they're just waiting to release it, maybe in a couple of days. Do you think that they're going to delay the release just so that these leakers are just proved wrong? Do you think that they're going to release it later today? Do you think that it's even going to be like this screenshot, what we have right here? Either way? I have no idea. I just wanted to show you guys this piece of information because I thought that when we look at everything together, it definitely paints a more cohesive picture, rather than just looking at one initial screenshot. So, with that being said, I'm still excited for GPT 4.5. And I. I do think that this AI race is heating up, so much so that all of these models are being rushed out the door. And largely, I do think that we still need to just remember that AI safety is a top priority.