Loading summary
Transcript1 lines
- [00:06]
A
If you are subscribed to any AI newsletters or you're following any AI news sources, you'll know everyone is talking about OpenAI's new project, Strawberry. What is going to be released here. There's been a ton of leaks and the biggest one of course, coming from the information, which is one of the best sources for OpenAI and kind of AI leaks. This is the only newspaper I actually subscribe and pay for. Uh, and it's just because their leaks are so good, I want like the full scoop most of the time. So in any case, some very juicy details about OpenAI's next model. It's leaks who they're showing this to, um, and what people are concerned about. There's a ton of drama on X. I'm going to break all of it down. Before we do that, I wanted to mention that we have a 50% off discount on my podcast course. I've talked about this in the past before. I. I know a lot of people are really interested and what I want to do right now is essentially get more reviews for the course, more testimonials. So if you are interested in starting a podcast and this is something that you'd be interested in this week only, I'm going to do a 50% discount on the course where you can get it. Usually it is 300, around $300 and so it's going to be around 150. If you're interested in the course, I would definitely get it this week. I think podcasting is one of the best things you can do to set yourself apart in your industry. Um, I've used it to raise money, get users for software and see a lot of business success. So this is something that I highly, highly recommend and this course is going to cover it all, how we got 4 million downloads, everything. So link is in the description and I would highly recommend it along with the discount code will be in the description as well for 50% off. So let's talk about the AI Strawberry leaks. Now, the first thing I want to talk about is the fact that there's been some rumors about this for quite a while and a lot of this even links back to Sam Altman. If you go to X, there's an account called I Rule the World Mo. This has been kind of. And the, the name of the account is like three Strawberry Emojis. And this account has been the cause for a lot of the drama. It actually accurately predicted a bunch of white papers, research papers that came out of OpenAI and they essentially has been saying when things are going to be released, when things are going to be happening. It's, it's a very, very interesting account. And currently, and at the top of this account is just the date November 5th. So I'm not sure what that's in relation to. But this was tweeted out on August 18 and many people are saying or speculating this is essentially when opening is going to drop their new Strawberry, which is apparently a new reasoning model or essentially technology that helps it reason. So it's going to essentially be better at reasoning, actually understanding things. Now there's a lot of leaks from the information that I want to talk about, but first I just want to give the background on this. So this I Rule the World account was tweeting a whole bunch of things about essentially GPT5. They call it GPT5. We know it's not going to be called GPT5 because Sam Altman has said that. But they said, you know, the release of GPT5, they've tweeted about it a lot. They've tweeted about a whole bunch of different insider information in the AI space, which leads you to believe like whoever this is running this, you know, quote unquote, Mr. Strawberry account is obviously hasn't knows an insider or is an insider in one of the big AI companies because they have leaks from anthropic Gemini and OpenAI of what's going on. So it's very interesting or probably leaks from all of them or possibly maybe even a journalist over at the information leaking it ahead of a story, I don't know. Anyways, very, very interesting. I personally was in recently a Twitter spaces or I guess an X spaces where this account actually joined. So this Strawberry account joined it but didn't actually say anything. And we were all talking about what, you know, who was behind it and what was going on. And I was kind of speculating and talking about some of the leaks that had been confirmed, making me think there's some sort of good credentials behind who actually is running this account right now. So in any case, a very interesting account, that's who to follow. They tweeted something and the reason why their profile picture is these strawberries here is because Sam Altman responded to them and said, I love strawberries in the summer with this picture of allegedly his strawberry garden. So everyone's like, oh my gosh, Sam Altman's retweeting them. It confirms it. I think he later deleted his tweet. So people are saying like, oh, it was, you know, they're Saying this is just like a troll and I don't know, whatever, but there's really interesting connections. Okay, let's get into what this thing actually can do. So a couple interesting, I guess just bullet points on it and then I want to dive into some more actual information. This allegedly is what people were calling Q Star. So if you remember everyone, when Ilya tried to kick Sam Altman out of OpenAI and everyone's like, what did Ilya see? And then everyone's like, oh, maybe he saw this new thing called Q Star, which is Advanced reasoning. Now apparently we're calling it Strawberry and that's kind of what this is. It's going to be integrated into chat. People are saying as soon as this fall. Well, to be honest, I would say November might be a good date just based off of what we're seeing over on Twitter. So it'll be interesting to see when that actually happens. Apparently it has extra thinking time so it can get really complex subjects. Um, it can solve things like the New York Times connections, you know, a crossword puzzle that's pretty hard. Um, it's able to do the New York Times connections, or I guess not crossword, but a puzzle. Um, it also is able to generate really high quality training data, data for another upcoming LLM which is codenamed Orion, which I'm assuming is essentially GPT5. They're just calling it Orion. And this is actually generating some of the training data for this. It's something we've talked a lot about in the past is how these AI models are essentially getting, running out. People could say running out of data to train on or people are, you know, asking for money or compensation, getting mad. So they're creating synthetic data and you want that synthetic data to be really high quality, otherwise it actually makes the AI model worse. So if you have something that's actually can create really high quality synthetic data, you feed it into new AI model, you essentially can have unlimited data and the outputs can get very good. So assuming Strawberry was able to do that, this might be the solution for GPT5 to get, you know, 10 or 100 times as much data, which could be really interesting. So all of this right now people are speculating is going to be used to help develop AI agents, which is, you know, these AI models actually being able to go out and do their own things, be completely autonomous, complete multi step tasks, do this all very efficiently. So why is this important? If all of these leaks are accurate, it's going to be a really big step in AI reasoning capabilities. Potentially moving us towards what OpenAI has said. There's like five stages on the roadmap to AGI. They have this outlined in some papers that they've talked about before. So if all of this is accurate and we're actually seeing these breakthroughs that people are talking about, then this could be quite big in the sense that we're, we'd be moving to stage two of this five step process. If this comes out in the fall, I think this is going to be really big. As far as disruption goes now, as far as the dates people are saying totally speculative. And who knows, if it was November 5th, it'd be interesting to see like how this is, how this has played out, you know, before or after the United States election. If people, if this comes out because they say they don't want it to be, you know, used for interference or election, you know, have any sort of impact on the election. I don't know. I think it's interesting and probably something people OpenAI are currently talking about as far as the actual leaks go and some of the things that the information has been talking about. There's a couple interesting things I wanted to bring up. The first is that OpenAI right now is growing incredibly fast. So sales of LLMs to essentially their corporate side, so the ChatGPT but for premium for businesses has about tripled. Right now it's at 2.83, $283 million in monthly revenue. That is astronomical. So right, we're talking about like over $3 billion just from that alone. Some people are speculating though that the monthly losses are actually a lot higher than that. So they're actually not profitable, probably losing money. Their valuation is about $86 billion right now. And like, you know, it says a huge valuation, but they are literally generating quite a lot of revenue. So I think that's always something to keep in mind. Right now with the, you know, this new flagship Orion AI model that they're developing, it's going to be interesting to see how this plays out because there's a lot of other players releasing a lot of other things. Some people are speculating that they already have it ready and they're just waiting for Anthropic to drop a new model so that they can immediately drop it after and dominate them. And some people are saying Anthropic already has another model just waiting to do the same thing to them. It's, it's kind of interesting to see how this goes. Everyone wants to have the best model for the longest, so they kind of like hold out to drop something. Should be interesting. Allegedly, Strawberry is going to reduce hallucinations by quite a bit, which is very interesting. I think the reason, by the way, that this is called Strawberry, getting, getting to the actual name all boils down to something that a lot of people have been talking about on Twitter, which is chatgpt not being able to spell the word Strawberry. Now, I did just go and try it myself and actually it. It was able to. Okay, so essentially what people do is they say, how many Rs are in the word Strawberry? And a lot of people on Twitter are showing that ChatGPT responds and says there are two Rs in the word Strawberry. And they're like, you're incorrect. And it's like, spell the word Strawberry. Give me every single letter. Letter individually. And then it, you know, it says S, T, A S T, R A, W, B, E, R, R, Y. And then it's like, okay, now look at every single one of those letters. How many Rs are there? Like, people really break it down before it can do it. And this is something that's been going viral for quite a few months. People have been like, it's the Strawberry problem, right? And so allegedly, the ability for ChatGPT to actually go and solve this problem, be able to tell you accurately how many hours are in the word Strawberry, is an indicator of its intelligence and would be an indicator of when this Strawberry system gets implemented or added. Now, what's interesting to me is I actually went to ChatGPT today and tested this out and it told me there was three R's in the Strawberry. So, I mean, I don't know if this is already implemented or if it just got slightly better on that, because there's a lot of people that have been essentially saying this and talking about, like, how it's essentially a sign of, of how not advanced the AI model is. The reason why this is important is not just for text. Like, definitely reasoning for text is really important, but beyond just the reasoning for text, when you have some of these AI models that are generating images, for example, and you need it to like spell out words or do menus or do or do text or fonts or like, it needs to be able to spell things, it needs to be able to understand what the actual letters are. So that's kind of its own problem, beyond just the reasoning, which hopefully this is going to be solving in a much bigger way. The last things I wanted to bring out, bring up was that allegedly some people are saying that right now. OpenAI is showing off the capabilities of Strawberry to some national security officials, essentially trying to get them, you know, give them the inside scoop ahead of time, you know, giving this to the, giving this to the feds. And they're saying that Strawberry right now is actually going to be. This element is what's going to make ChatGPT 5 or aka Orion, so much better. It's not just that the new model is going to be so much better. It's like this piece of technology independent of the model, which is kind of interesting. It feels like maybe the AI models are just predictive things like we've been saying for a long time, but it's hard to actually make them those. That element, better reasoning. It's like outside software technology that we bring in. Something else people have been saying is that there is a smaller version of Strawberry that could launch in the chat bot Chatbot forum very soon. So we'll, we'll see how this plays out or if they're going to roll out the full thing. All in all, there is a ton of speculation, a ton of rumors going around. A lot of these are coming from the information though, which is a very reliable source and it's been right about a lot of things. So this is great. Is definitely a lot of competition from Google Xai, Anthropic and meta right now. So when we start looking at, you know, the Chatbot arena and a lot of these other areas, it's going to be interesting to see how this plays out and who is going to come out the winner, if this is really going to boost them significantly ahead or if it's a bit of a dud and other people are going to be able to catch up to this new Strawberry technology, I'll keep you up to date on all of it. If you enjoyed the podcast today and you want to start your own podcast, make sure to check out the link in the description to my podcast, which is 50% off this week only. There's a, there is a discount code in the description, so you can go check that out and get my podcast course, which is normally $300 now 150 for this week, which will cover everything from start to finish growing your podcast, scaling it, how I got 4 million downloads and how I was able to raise hundreds of thousands of dollars for my business, all from podcasts in the last year or so. So go definitely check that out. Thanks so much for tuning into the podcast today. Hope you all have an amazing rest of the day and I will catch you next time.