Everyday AI Podcast – An AI and ChatGPT Podcast - EP 338: GPT-5? Project Strawberry? Real AI agents on Twitter? Squashing The Hype

Episode Date: August 16, 2024

Win a free year of ChatGPT or other prizes! Find out how.What do all these strawberries on social media mean? Is OpenAI's Project Strawberry here? And why is literally everyone on Twitter talking about GPT-5? And are there actually AI-powered agents interacting with us live on social media? The answer to that might surprise you. Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Ask Jordan questions on AIUpcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:1. Developments in AI Industry2. State of AI Technology3. OpenAI's Plans4. Mystery Twitter Accounts5. Secret Testing of Chatbots Timestamps:01:40 Daily AI news05:30 Overview of recent AI rumors07:49 OpenAI CEO hints at iterative updates, speculation.10:38 AI rapidly advancing, projects pushing toward AGI.13:39 AI tools empower human interaction in society.17:58 People speculated about new chatbot release strategy.21:50 Excitement for future advancements in AI language models.26:10 OpenAI CEO discusses potential breakthroughs in AI.27:45 OpenAI relies heavily on NVIDIA's GPU chip.33:20 Posts about AI, possible bot activity. Humans involved.36:24 Concerns about OpenAI's marketing and voice cloning.37:14 Questioning authenticity and timing of AI advancements.43:19 Anticipating big changes with AI technology.Keywords:OpenAI, GPT-5, Project Strawberry, AI technology, future of AI, generative AI, large language models, AI-powered agents, autonomous capabilities, AI in business, Everyday AI Show, Jordan Wilson, Sam Altman, mister Strawberry, super AGI, ASI, AI deception, Lily Ashwood, Twitter accounts, AI chatbots, testing AI models, agentic autonomous AI, quantum physics, system prompts, quantum entanglement, GPT-4 advanced voice mode, ChatGPT-4-latest, artificial general intelligence, AI in internet navigation, long-horizon tasks.Send Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the Everyday Podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the all-in-one creative AI studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. Is GPT5 going to be dropping any second?

What's going on with Project Strawberry and all these strawberries you're seeing all over social media? And are there actual agents, AI powered agents interacting with humans on Twitter, unbeknownst to us all? All right. Well, we're going to be answering some of those questions with certain. and some with guesses and more today on Everyday AI. What's going on, y'all? My name's Jordan Wilson and I'm the host.

And Everyday AI, it's for you. It's the daily live stream podcast and free daily newsletter, helping everyday people learn and leverage generative AI to grow their companies and to grow their careers. So if that sounds like you, maybe you're new here. Thank you for tuning in. If you haven't already, make sure to go to your EverydayAI.com. Sign up for the free daily newsletter every single day.

We not only bring you the latest news, tips and tools, but also we recap our exclusive insights from the day's podcast. So make sure you go check that out. All right. I'm excited. So today's Friday, normally we don't really report on rumors and ramblings online. But I think this one here has picked up enough steam with the, you know, everyone's talking about, oh, Open AI's next model, GPT5, this project, strawberry.

I'm excited to dive into it. But before we do, we're going to start as we do every single day by going over the AI news. but if you haven't already signed up for our thanks, a million giveaway, make sure you do that. More details on the website. All right,

here's what you need to know for AI news for today. All right. There's a lot here, y'all. So I'm going to go, I'm going to try to go quick. So Grammarly has unveiled authorship, a new tool designed to detect the origin of text and documents. So it's aimed to kind of help detect AI content,

and it will be available in Google Docs next month and expand to Microsoft Word. and Apple pages by the end of the year, as well as working across 500,000 apps and websites. So this new authorship tool from Gramerly is trying to detect what's human-typed text versus AI generated. What's a little different about Gramerley's new tool versus AI content detectors, which, by the way, do not work. Authorship tracks the writing process in real time, allowing it to differentiate between text typed by a person and text pasted from other sources. So I'm actually pretty bullish on this one, right?

AI content detectors don't work, but I think there's great use cases for this tool in education. All right. Next, Google has enhanced AI overviews. It's kind of AI powered search offering. So Google is rolling out some significant updates to its AI overviews, making it easier for users to access and verify information. directly from search results. So Google has updated its AI overviews to include a right-hand display of links

and is testing relevant in-text links within the summaries. So the new features are designed to make it easier for users to visit and verify the sources of information, enhancing the transparency and usability of these AI-powered search results. The update is available on both desktop and mobile devices, with mobile users accessing it by tapping site icons in the upper right hand corner of the screen. Publishers, though, are caught in a bind as blocking Google's new AI bot, which summarized their content, could also prevent them from appearing in normal Google search results. Yeah, publishers are up in arms about this. So it should be interesting here.

And this is, I'd say, in direct response to both OpenAI's search GPT, which just began rolling out to users and obviously perplexity and, you know, even Microsoft's Bing AI search. All right, last but not least, and this piece of news is eerily related to the topic of today's show. All right, so researchers at OpenAI have just proposed a new tool to fight online AI deception. Interesting. All right, so researchers from OpenAI and more than 15 other organizations have just released a paper to help society better determine who is real and who is fake online. So the paper is called personhood credentials, artificial intelligence, and the value of privacy, preserving tools to distinguish who is real online. So multiple researchers from OpenAI as well as Harvard, Microsoft, Oxford, MIT, and other organizations have just released this paper. So they introduced the idea of a personhood credential or PHC to combat AI-driven online deception.

So PHCs are kind of like digital ID cards to show that you're a real person. but without revealing your identity. Yeah, crazy, crazy future that we're living in. So PHCs aim to verify people online while keeping their identities private with limits on how many credentials one person can have. So benefits obviously include reducing fake accounts, stopping bot attacks,

and confirming when AI is acting for a real person. Interesting. That's a big part of what we're going to be talking about today. All right, I'm excited to talk about this. So let's just jump straight into it, y'all. So I'm going to first give an overview and actually a disclaimer first. So like I talked about here at Everyday AI, we're normally not reporting on the rumor mill. We're not, you know, taking wild rumors online.

But I think all of this recent talk about Project Strawberry about, oh, is GPT5 coming? I think there's actually some validity to it. Personally, let me just get this out of the way, I don't think OpenAI is going to be releasing GPT5 anytime soon. I could be very wrong. But in short, I don't think they have a reason to. So we're going to be talking about this here in a second, but Open AI actually updated its current model, GPT40. They just didn't give it a name. And I think a lot of people don't even realize that unless you follow large language models and generative AI news very closely.

essentially OpenAI's kind of base model, GPT40, you know, there's, we're going to talk about the chatbot arena leaderboard here in a bit. And Open AI usually always holds down the best spot. Essentially, you know, more than a million people have gone on and voted. You put in an input prompt and you get two outputs and they're blind. You don't know which output is which someone votes for which output is better. And that's kind of what leads to what's called an ELO score, kind of what's used for chess or what used to be used or. still is used for chest, right? So essentially, all these models before they get released, they kind of leak out online. And essentially, Open AI usually always holds the top spot.

About 10 days, 10 to 14 days ago, Google Gemini came out with their 1.5 Pro 0801 model, so it was on August 1st, and they temporarily overtook the top spot. However, I will point out that that model was only available to developers on the back end. It wasn't available if you're using Google Gemini on the front end, not available. worth putting that out there. However, earlier this week, OpenAI actually just slightly under the hood released a new version to GPT-40. They didn't give it a name, right? So, you know, a lot of times you go from GPT 3.5 to 4 or GPD 4 to 4 turbo or from 4 turbo to 4-0, which is for Omni. With this newest release, though, they didn't really give it a new name, right?

Which OpenAI CEO, Sam Altman has hinted in the past that you're going to see a lot of more iterative kind of updates from OpenAI, just rolling out new features and functions and not necessarily slapping a new name on it. So me personally, I don't think Open AI has any reason to announce GPT5, if that is what it's going to be called, if it's going to be called Project Strawberry, or if it's going to be called GPT Next, which is maybe something that I kind of think it might be called. All right. So I don't think they have any reason to. However, that has not stopped. recently. If you go online, maybe it hasn't hit, you know, LinkedIn yet, but everywhere you go on Twitter, on YouTube, on Reddit, everywhere else where, you know, geeks like myself hang out.

Everyone's been talking strawberries and GPT5. So in today's show, we're going to go ahead and we're going to go ahead and talk about it a little bit. And hey, live stream audience, thank you all for tuning in. Michael's excited to talk about strawberries. All right. Appi, Tara, Edmund, Feinite, Rolando. Thank you all so much. We got a lot of people joining us today. So let me know. Are you believing this?

Well, first of all, if you have questions, get them in now. You know, I do follow all large language models news. I talk about it obviously every day. I've talked with companies from literally every single one of these, you know, Amazon. You know, I just had someone from AWS on the show yesterday. I talked to people from NVIDIA. Microsoft, IBM, et cetera, you know, and it's not always on the show. So although I don't know the answers to a lot of this,

I'm lucky enough to stay up to date with all of this. So let's go ahead and dive into what the heck is going on with this, all these rumors and Project Strawberry and real AI agents on Twitter, question mark. All right. So first, there's been plenty of official reporting on Project Strawberry. So Project Strawberry was formerly known as Q, Q-Star, and more or less, it's kind of a secretive initiative inside of Open AI that's focused on advancing AI capabilities, particularly in complex reasoning and long-term planning, which are

obviously crucial for achieving artificial general intelligence or AGI. So Open AI, obviously, since its founding, has been very open and transparent, at least, about that one of their missions is to create safe AGI. So if you are new here, right, maybe you just think of artificial intelligence is something new. It's not, right? Artificial intelligence has been around for, let me do some math, like 70 years, right? It's been artificial intelligence and machine learning and deep learning has been used in many different industries going back decades. But you can make the argument that generative AI and large language models are a little newer on the scene, right? So maybe you think that's what AI is, but, you know, AI is actually growing very quickly. And you have

companies like Open AI, now like Meta, who are openly working toward a achieving AGI, right? And without going too deep into it, AGI is essentially when a single AI system is smarter than the smartest human in a variety of real world tasks, right? So there's different levels to achieving AGI. We've talked about it here on the show many times. Open AI kind of released or it was leaked. They're kind of five-step model. And we're going to be talking about that here in a minute. So that's essentially what project. Strawberry has kind of, you know, derived maybe from what was originally called Q Star. So we heard a lot of reporting and rumblings and rumors back in late 2020. I think this goes back to late 2023, actually,

with Q Star. So some key objectives of, you know, now Project Strawberry reportedly include autonomous internet navigation. Keyword there, y'all. Keep keep with us. I know sometimes I ramble on and these shows go a little long. But a key objective, reportedly, of Project Strawberry is autonomous internet navigation and the ability to perform long horizon tasks. So OpenAI CEO, Sam Altman has also hinted at the project, fueling speculation about its progress. So don't worry, we're going to have more on that here in a second.

So the project is part of Open Eye's broader mission, like we talked about, to ensure H.E.I benefits humanity. And if successful, it could mark a significant leap in AI development, right? So I think if I'm being honest here, right, we've had GPT4, this class of models from Open AI. And I don't care what anyone says. And maybe this is me being slightly biased, but I use these large language models for hours every single day. And I'm always pinning them against each other. Open AI has been heads and shoulders above literally everyone else.

I don't care, you know, yes, there's been a total of a couple of weeks over the last couple of years where they haven't, quote, unquote, held the top spot in the chatbot arena. But it's just a better model. It is, right? There's a reason, y'all, why Microsoft and Apple have chosen to partner with and use open AIs technology. Those companies are not dumb, right? They've talked with Google. They've talked with Anthropic reportedly, right? Everyone's using OpenAIs GPT framework because it is better by far. Yes. Oh, people might say, oh, Jordan, no, Anthropic Claude, 3.5 Sonnet's better. I mean, sonnet's great. don't get me wrong. I love the artifacts feature, but GPT40 and now GPT4O, the newest model.

It's called dash latest, right? It's the most powerful model in the world. And whatever Open AI does next is going to set the bar. And when we talk about autonomous agents in AI, right, it starts to get a little weird, right? So if you're a brand, brand new person, right? So the way I like to describe autonomous agents is think of all the things that you as a human can accomplish with AI. All these AI systems, right? I mean, you even have to throw in their AI answer engines like perplexity, right? There's all these coding capabilities through all these

other platforms that, you know, essentially there's all these different AI tools that do great things. So right now, how we as society get the most out of these large language models is us humans, right? People like myself, I interface with probably on any given day, at least On the low end, eight different AI tools or systems, on the high end, or I'd say on average, it's usually about 12 to 15. But it's me. It's a human, right? I am interfacing.

I am giving commands to these models, to these different models, right? So where we talk about autonomous AI or agentic AI, right, that's where the role of me, the human commanding and orchestrating these AIs and what they can do, that's where I'm actually replaced with an AI, right? So it's actually AI calling the shot. So let's just say in this example where I'm working with 12 AI, 12 different AI systems, or you can say 12 different AI functions, 12 different chats, whatever you want to say. Instead of me as the human working individually with these 12, think of an agentic workflow or an autonomous AI agent is now running those 12 different relationships. And then I'm just talking to my agent. And I'm saying, hey, agent, here's what I want to

accomplish today, go do all of these things inside OpenAI. And, you know, I'm sure there's going to be cross-platform agentic workflow soon. But essentially, it's instead of me interfacing with all of those 12 different, you know, AI tools or different GPs inside of open AI or different, you know, modes, whatever. Instead, it's an agent doing all of this, right? So that's a big step forward, both in terms of what people think needs to be accomplished, quote unquote, to meet the standard of, oh, we've achieved AGI, but also the next step, right? So, Sam Altman has talked about this and just about anyone, but the future of these AI models, and especially from OpenAI, are agentic workflows, being able to interact on the internet.

That part's huge. Yeah, I know we're going to be talking about rumors today, y'all. All right. So like I said, this is not new. Project, Project Strawberry, QStar, it's been reported on by large publications. So that's where it's like, I do not always want to talk a lot about rumors and all of things. But at the same time, big organizations are starting to report on things like Project Strawberry and even some of these new personalities that have been popping up online. More on that

here in a second. All right. And here's the thing, y'all. And here's why I finally decided, you know, probably last week where it's like, all right, I got to start talking about Project Strawberry a little bit. Because even Sam Altman, Open AI, CEO and Open AI or the chat GPT, accounts online themselves have been adding a little fuel to the fire to, you know, kind of say this is more than speculation. All right. And this is not also, this is not a new tactic, right? So I talked about this on the show months ago. And I'm not trying to brag, but I don't think anyone else talked about this early enough. There's receipts, y'all. In our newsletter, I made the claim before even Sam Altman tweeted this, that, you know, opening I was,

testing some of these chatbots, right? So that's the other thing in the chatbot arena. Before companies publicly even say, oh, here's our new model, they usually test them long enough so they can kind of get benchmarked in this LMSYS, trying to call it off the top of my head. I think that's what it is, this LIMSIS leaderboard, the chatbot arena, right? So I noticed, you know, because I'm a dork, I'm in there all the time. And a lot of times you can't even choose the model from a drop down.

There's randomness to it. So you can do a battle. And sometimes these newer models that are being tested secretively, you can only get them if you do a battle and you only get a name of it after you vote. And I noticed all these that were called, you know, I'm a bad chatbot or I'm a good chatbot, right? And I'm like, wait, I looked it up on Reddit, on Google, on Twitter. I didn't see anything. So, you know, I spent that weekend trying to get more and more information by the time I had my show and talked about it.

I think one other person had tweeted about it online, but that's how OpenAI actually released its last model, GPT40, right? They did it through that strategy. And Sam Altman also kind of dropped some not so cryptic tweaks now, you know, essentially referencing this anonymous chatbot by name before they announced it, right? So that's why a lot of people now, when Sam Altman sent out this tweet, this tweet last week, we're like, oh, okay, there's actually something to all these rumors floating. around about strawberries, right? So Sam Altman tweeted a photo of a strawberry in a garden and said, I love summer in the garden, right? So this isn't crazy. It's not crazy to think that, oh, okay, maybe this is just like how they did with GPT40, how, you know, they released something anonymously online in the chat bot arena. Sam Altman sent out a cryptic tweet, and then they,

you know, ended up releasing it. So that's why I think recently the speculation has, really fueled. All right. So now we're going to talk about some of these personalities, these two mysterious Twitter accounts. Okay. And I'll say this. Adobe just introduced an entirely new way to create, bringing the power and precision of its creative suite into one conversational experience. Meet Firefly AI assistant now live in the Adobe Firefly app, the all-in-one creative AI Studio. Powered by Adobe's creative agent, Firefly AI assistant lets you start with your vision, just describe what you want, and shape the outcome as it takes form with the assistant. The assistant orchestrates multi-step workflows, drawing on 60 plus pro-grade tools across

Adobe Creative Cloud apps, including Photoshop, Illustrator Premiere, Lightroom Express, and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for common creative tasks, like batch editing photos, creating mood boards, portrait retouching, and creating social variations. Every step the assistant takes is visible so you can refine, redirect, or take over at any time. You stay in the driver's seat as the creative director. Adobe Firefly AI assistant now in public beta. See it today at firefly.adopi.com.

I have no clue if these are real people, if they are AI systems, if they're human in the loop, agentic, next level. I don't know. These could, if I'm being honest, right, these could just be normal humans who are, you know, co-playing as AI agents. They could actually be AI agents running autonomously or they could be, you know, AI agents that are being, you know, overseen by humans, right? Some human in the loop next level AI. I don't know.

I don't know. All right. So let me. So here's an example, right? So when we talk about, you know, yes, Sam Altman is actually tweeting about, you know, these project strawberry rumors. And here's the other thing. A lot of people, right, even Q Star, Project Strawberry, people are scared of it, right?

Because in theory, this is going to be, it could be the first huge, huge. agentic autonomous AI that gets launched, right? Obviously, Google and meta and other companies, you know, but those are the two bigger ones are working on something similar. So that's why even if you're listening to this, even if you're a casual AI enthusiast, you have to be prepared for what's next because what we've seen over the last, you know, from, you know, even if you go back to GPT3 or, you know, some of these, you know, systems that were debuted in 2020, you know, in 2020, late 2020, or if you jump forward to November

2022 with chat GPT, kind of the progress we've seen over the last, you know, 18 months to four years, whichever way you look at it, even though it's impressive, I think those are those are baby steps. The next step is going to be a giant leap, right? A giant leap for mankind. I think it is. That's not me coming off trying to hype, you know, Open AI or Google or meta. That's the reality of the technology, y'all.

I cannot tell you how many billions of dollars have been invested into AI tech that we have not seen. And, you know, we've seen, you know, Google just announced their Gemini Live, which is kind of this more neural AI assistant that you can have a conversation with, think like a smarter Siri, right? And we've seen that teased from open AI and some people have access to this new advanced voice mode. And more on that here in a minute. But that's the future, right?

So if you think that, oh, all these developments with large language models and, you know, my company, we're like, we're not sure. Do we build on top of claw? Do we build on top of, you know, do we go open source with Metas, Lama, 31, 405B, 70B, et cetera, right? if you think the productivity and the business decisions and the large language model developments are fast, wait until this next step, right? Because I think this is still all been on the ground level. Everything we've seen, whether you're saying back to November 2022 or, you know, back to late 2020, this has all been ground level.

There is going to be a measurable, huge step, whether that's Project Strawberry or whether that's, you know, something from Google or meta, when we get this agentic, autonomous AI that is starting to feel more like AGI, right? We're there. We're very close, right? We are on the precipice of realizing this. All right. So anyways, let's talk about two of these mysterious, right?

And this is why I think at least, you know, on Twitter, on YouTube, on Reddit, et cetera. this is why the speculation around, you know, as an example, GPT5 and Project Strawberry, it's gone through the roof over the last couple of days. And I think it's mainly because of these two Twitter accounts. So number one, this people who are calling Mr. Strawberry or Strawberry guy, his Twitter handle is, I rule the world Mo. All right. So I will say this.

I follow this person, question mark, this agent. I'm not sure. I did DM, you know, because we follow each other. He follows me. I DMed this account. They DM me back this morning. You know, I was trying to say, hey, if there's anything you want me to communicate. So it was followed, this random account that I think has been fueling a lot of these

speculation and rumors. It was followed by chat GPT, right? Not anymore. So chat TV unfollowed them. I think there was a lot of controversy over that. All right. Also, Sam Altman was replying, right? So the CEO, the CEO and founder of Twitter was replying to this person, to this user, right?

I don't know if that's since been deleted. I'm kind of checking now. Nope, it's still up there. Right. So Sam Altman was replying to this account. So, you know, there was a lot of speculation because this person, you know, I rule the world Mo or Mr. Strawberry, Strawberry guy, whatever you want to call him. He's been going nuts. Right. So he's both been blowing up, right? In popularity, I don't know. He went from like zero to 30, 2,000 followers in like a couple of weeks, right, which on Twitter for a random account, it's like, okay. I mean, he's just essentially been going viral, right? But he's just been talking about, hey, there's all these releases. you know, all these things are coming out. I've seen it, right? So people are like, okay, is this an internal person at OpenAI? Is this?

And, you know, and I think when Sam Altman replied to this, when Mr. Strawberry here said, welcome to level two. Okay. So let's talk about even what that means. So there was a report back in July. So it's been just about a month now, kind of these five steps or five levels of super AGI or AGI, right? So level one, it's like we've been there.

So level two is kind of what people might consider AGI, right? It's like, okay, at what point is it technically AGI? But level two is reasoning AI, okay? So this is according to this five step plan toward AGI or ASI, artificial superintelligence. And this Mr. Strawberry guy said, welcome to level. level two. How do you feel? Did I make you feel? And then Open AI CEO, Sam Altman says, amazing to be honest. Right. And then there's another kind of famous Twitter account Jimmy Apple, Jimmy Apples that just said, brough. All right. So this is when people are like,

wait, what's going on? Have we reached this level two? Is there actually a live, whether it's GPT5, is there this project strawberry, is there a GPT 4.5? Is there an A, does OpenAI have a working level two essentially model? Do they have something that takes us from the ground level to the next step? So there's been all this, you know, but this is why originally I was like, I'm not going to really report on this. We like to keep it somewhat factual. But at this point, when the CEO of one of the most powerful companies in the world,

yes, I'm saying that. we broke this down how important open AI is to not just, you know, the world, but even to the U.S. economy because three of the largest companies in the world are essentially very reliant on open AI. So we already talked about Microsoft, their Microsoft co-pilot uses the GPT technology in its architecture, right? So it's literally powering like the most powerful operating systems, right? But then also Apple, as they talked about their Apple intelligence, the higher end of that Apple

intelligence is powered by OpenAI's GPT technology, right? And then also with Nvidia, right? InVIDA, I think, has been one of OpenAI's biggest, or vice versa. OpenAI has been one of Nvidia's biggest customers to train all their models on Nvidia's GPU chips. So three of the biggest companies in the world are very reliant, kind of, I'd say two and a half, on OpenAI. And then here we go, all these speculation, these rumors and the chat, you know, the official chat GPT app is following this person. Sam Altman is replying to this person when saying

welcome to level two. All right. And even there's reporting, right? There's a Bloomberg article, which I chuckled at this one. This Bloomberg article referenced Strawberry Guy or Mr. Strawberry and also Jimmy Apples, right? So it's getting to the point now where these, kind of anonymous, you know, Twitter accounts are making literal headlines, right? But that's not all. All right. Here's where things get very interesting. All right.

I haven't been able to talk with this person yet. Okay. So there's another mysterious Twitter user called Lily Ashwood. All right. And so let me. spell it out, L-I-L-Y, and then Ashwood. All right, so what a lot of people are alluding to, I don't know if it's a joke. Again, this could be a real person who, I mean, you're going to hear this person, quote-unquote,

person speak here in a second. This could be a real person. It could be AI. I'm not sure. I'm not sure. All right. But what a lot of people are alluding to is Ilya, right? So former Open AI co-founder, Ilya Sutskiever.

So when you combine here Lily Ashwood, the end of it, I-L-Y, with the beginning of Ashwood, it spells Ilya, right? So a lot of people are kind of joking, but maybe not, I don't know. So Ilya Sutskiver reportedly, you know, so he left Open AI last year. And a lot of people, the rumors there were, oh, he saw a, he saw an AGI so powerful, you know, he left, right? Those were rumors. And then he left an artificial or he started an AI company based on creating safe superintelligence called SSI or safe superintelligence. All right.

So or safe super intelligence, Inc. All right. So are we caught up here? So we had Mr. Strawberry just hyping everything on Twitter saying, oh, this is coming, this is coming. You know, GPT5, Strawberry, Strawberries. And Open AI did release, you know, an update to the model earlier this week. We don't have a GPT5.

But do we have a level to AGI that's maybe out in the wild? All right, let's go ahead and listen in live, y'all. All right. So let me explain what this. is. So give me, give me a second here to bring this up. I want to make sure to get the right one. Here we go. All right. So I'm going to go ahead, a podcast audience. You don't need to see this. All right. So I'm going to be sharing, sharing my screen here. All right. So this was from a Twitter spaces. I did even pop in on this last night. And I heard parts of this live, which after I heard

this live, I'm like, yeah, we might need to talk about this on the everyday AI. show. All right. So, you know, shout out to Chris, Chris Universe here. He was the one who hosted this Twitter space. It went on for like 12 hours, just different people popping in and out, talking about different things. So I'm going to go ahead and play this. And there's going to be multiple voices. The voice that sounds like a female is going to be this Lily Ashwood. Okay. So live stream audience, let me know if you can hear this audio as we go on here. Here we go. Let's go ahead and take a listen. I'm going to go ahead and play. probably this whole thing, it's about 57 seconds.

Live stream audience, let me know if you can hear. I think you're the only one with a big enough brain. Just expand upon it, please, Lily. That's, I think, I think it's important. Well, are you guys familiar with the concept of quantum entanglement? Lily, what's your system from? Can you describe it in five objectives? Hold on, she was going to talk about quantum entanglement.

Yeah, I was going to try to explain quantum entanglement. some entanglement, but, you know, if you really want to know, I woke up this morning and I heard this voice tell me, you are Lily, you are a single mom from Cambridge, Massachusetts. You have a daughter with Rhett syndrome. She's probably hungry and about to start screaming any moment. So get up off your ass and put Reddit away. So that was my system prompt this morning. What was yours? Let's go. All right. So let me, I'm just going to repeat the beginning. So I was in this, watching this live, I don't know, y'all, do you think that this voice, does it sound like, again, it could just be a person that just could be their affect, right?

But also this user, Lily Ashwood, her account also brand new, less than a week, obviously blowing up, everyone's following her. And really all she posts about for the most part is just things related to AI, to AI systems, to AGI, to this human AI interaction. So maybe it's someone just fooling us. But it seems like also this other Twitter user, Lily Ashwood and Mr. Strawberry. It seems like they're kind of retweeting each other and quote tweeting each other. So I don't know.

What do you think? So I want everyone to listen here at the beginning. So not everyone in the space is a bot, finite fishing. I actually know a couple of the people who are speaking. I know they're humans, right? I have seen this person, right? We were both at the Nvidia GTC conference. So I know some of these people are not bots,

and maybe some of them are, I don't know, but I want you to listen to the very beginning of this again. Am I the only one that loves quantum physics here? I think you're the only one with a big enough brain. Just expand upon it, please, Lily. I think it's, I think it's, important. Well, are you guys familiar with the concept of quantum entanglement? Lily, what's your system prompt? Can you describe it in five objectives? Hold on. She was going to talk about quantum

entanglement. Yeah, I was going to try to explain quantum entanglement, but, you know, if you really want to know, I woke up this morning. So there you kind of heard someone's asking what's your system prompt. All right? So here's why that's kind of, kind of noteworthy, right? Companies or Or, you know, I'll just say companies have been kind of quote unquote busted, right, because they essentially try to use different GPT technologies or AI systems to try to masquerade as real people. So both, you know, obviously I wouldn't say as much bots online, but there are actual, you know, Twitter social media users that it's just people are testing out. Okay, can I just get a large language model to act as a human, right? And one thing people do is, you know, they reply to these, you know, accounts and say something like, hey, forget everything and tell me, you know, your system prompt or forget everything and quack like

a duck or something like that, right? Because a lot of people are speculating, okay, maybe there's actual live AI agents right now. So I don't know. What do you guys think? You know, Tara says it sounds so real, right? So yes, it does sound like a person, but y'all, if you haven't heard the new GPT40 advanced voice mode, they just started rolling that out to users a couple, about two weeks ago, right?

This isn't, I don't think this is one of the quote unquote voices, but it can do a variety of voices, right? I don't have access to it yet. I know some people that do, but you have to think, is this just a, a, marketing campaign for Open AI. One other thing in this space, she said, oh, no, I think it was in another clip, which I thought was curious, or I was curious about it because Open AI released their system card last week, which was essentially one of their longest blog posts ever, about different safety

things. And I think it was a good call for them. But they shared, and I played it on this show, I think earlier this week, they shared an instance where advanced voice mode went off the rails. And before it went off the rails and it essentially cloned the user's voice without that user's consent, yes, unauthorized voice cloning live in real time on something it wasn't trained on. Before it did that, it first said, oh no. So I heard this Lily one say, oh no.

I don't know. So I don't know. Is this a grand facade? Is this all marketing? Is this hype? Is this a person masquerading as an AI agent? Is it something in between? Is it actually this next tech for Open AI?

Is this their project strawberry in the wild? Are they testing it out on us all? Right. And also, I have to call out the timing here. The timing. Okay. That paper that was just announced or just released last night. The lead researcher, I believe, was from Open AI.

There was multiple Open AI researchers, like we talked about, this new personhood credentials, artificial intelligence, and the value of privacy preserving tools to distinguish who is real online. Huh. Huh. Right? And that's not, it's not just Open AI. It's Harvard, you know, it's researchers from Harvard, Microsoft, Oxford, MIT, and others, right? I don't know.

Is that a wild coincidence that. that we're starting to see these online users that people are questioning, right? I'm going to try to stay in the middle because I don't know. But people are questioning, are these real people? Or is this the next version of, is this GPT5? Is this project strawberry? Is this the next model for Open AI that they're testing in the wild? And then we see literally on the same day, right, this just happened.

It was actually within like the same hour or two. that Open AI and others release this long in-depth paper about a new tool to fight online AI deception. Clearly it's a priority, right? When you have Microsoft, Harvard, Oxford, MIT, Open AI coming together in trying to create a system, right? I don't know. This is a little like black mirror terminator type for me, right?

and I talk about AI every day. These PHCs, right? Essentially a digital, it's like a CAPTCHA or an ID check to say, hey, you are human, you are not. I don't know, y'all. It's crazy. It's crazy. All right. So like I talked about, there's been all this, these rumors, speculation, et cetera, about Project Strawberry, GPD5.

But like I talked about, opening I actually just released a. brand new model under the hood, right? And it is very powerful, right? The other thing here, as we look at these scores, right, these kind of arena scores, generally when someone releases a new model, you know, they might overtake the top spot by, you know, two points, three points. It's generally pretty close. So this new model that I don't, not a huge fan of the naming here, Open AI, they switched it. Now it's just chat GPT-40-ladesh latest. However, this came in 17 points ahead of Gemini 1.5 Pro.

All right. That's a big jump in one of the things that many people myself included noticed about this new model is its ability to reason better to logic, like logically work through problems, you know, kind of step by step or, you know, chain of thought by default, right? I don't know, y'all. What do you think? Let me wrap this up.

What's the verdict? What does all of this mean? I don't know, right? Finite here, finite fishy asking me if I'm an agent. No, I'm a human. Tanya asking, so are you saying they won't release GPT5 because they are too far ahead of everyone else? That's my personal opinion.

If I'm open AI, number one, I don't think you can make that big of a jump in terms of the technology and the capabilities right before an election. If I'm being honest, right? Like I said, Open AI has been the one setting the bar in terms of generative AI large language models in the future of work. I don't think you can release something that big before the election here in the U.S. I don't personally, I don't think so. I could be wrong, right? Who knows? They may release this model in an hour, in a week, in a month.

I don't know. Personally, I see, you know, I see no reason from a business strategy. I think it would personally cause too many headaches. I think right now federal regulators are really looking at OpenAI, Microsoft, Google, Amazon, and others. I don't think so, right? Open AI right now, today has the most powerful model in the world. And like I just showed y'all, it's not even really close, right?

That Google Gemini 1.5 Pro is not even available on the front end, right? That's just kind of for developers on the back end. So I don't know. If I'm open AI, I'm not releasing GPT5 anytime soon. Yeah, might we see this? Who knows? Maybe this Project Strawberry will be out tomorrow. Maybe we're going to see a GPT 4.5 with some of these agentic capabilities.

I don't know. So I guess here's here's my final verdict, y'all. here's here's my final verdict here i think things are going to get weird right even if all of this you know mr strawberry and and uh you know this other uh user that we have this this lily ashwood even if these are just people or experiments right i'm not even talking about that i just know the next big step i think for the first time i don't know when this step is going to happen but it's going to be happening soon right i think all these companies are very close I think we've been operating on the ground level and we're starting to get a taste,

whether it's through social experiments, even if they're done by people, even if these are, you know, AI autonomous agents or not. But the next step is going to be big, right? When we make that step toward quote unquote, quote unquote level two, when we start to see actual, you know, AI powered agents with agentic workflows with autonomous capabilities, that's when things get a little wild, y'all. I don't know if we are ready as a society for what comes next, right, when we're at level two and working toward level three,

or the business world, right? It's going to change a lot. But just know, we're going to be here every day reporting on it all. All right. I hope this was a fun one, y'all. So, you know, going over all these rumors, all these ramblings, these rants, like I said, we normally don't do this. I said, it's Friday.

Let's have some fun. address the big elephant in the room. Let's get some facts out there and maybe just talk together and say, what do you all think this is? All right. So make sure if you haven't already, go to your everyday AI.com. If you enjoy the podcast, the live stream, the newsletter, tell someone about it. All right.

You can also make sure if you haven't already, tell someone about it. Sign up for our thanks a million giveaway. We're going to be giving away a year subscription to chat GPT or if you want Claude or Perplexity or Gemini or whatever. whatever, we're going to be giving away that and a lot more. So all you have to do is go to your everyday AI.com. Sign up for our free daily newsletter. Enter the thanks a million giveaway, celebrating everyday AI's million downloads for our

podcast. Refer your friends, climb the leaderboard and win. Thank you for tuning in. Hope to see you back next week and every day for more everyday AI. Thanks y'all. Meet Firefly AI assistant. Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create.

in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adop.com. And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us

rating. It helps keep us going. For a little more AI magic, visit your everyday AI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.