EP 463: AI News That Matters - February 17th, 2025

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.

New AI models, new AI models everywhere. That is the name of the game this week when it comes to AI news. The big companies are releasing some big updates. Some probably coming as soon as today, others probably coming within the next week or two. So Elon Musk says that Xai's Grok 3 is being released tonight. OpenAI has laid out the plans for GPT5 and more and new reports are saying that Claude's highly anticipated new model. So the new model from Anthropic was will be a hybrid model. So we're going to be talking about what those stories mean and a whole lot more today on Everyday AI. What's going on, y'? All? My name is Jordan Wilson and I'm the host of Everyday AI. I do this thing for you. This is your daily live stream, podcast and free daily newsletter helping us all try to keep up with AI news but actually put it into use to grow our company and our careers. That's what this show is all about. So maybe this is your very first time tuning in. Thank you very much. If you haven't already, please check out your everydayai.com that is our website, but that is your new BFF in AI. So there you can go sign up for our our daily newsletter and our daily newsletter, we not only recap each, podcast and live stream each and every weekday, but we also give you every other piece of AI news and development that you need to to be the smartest person in AI at your company. So that's where you go do that all for free. Hundreds of back episodes as well. So your Everyday AI dot com. All right, so enough. Well, one other thing you got to do when you're there, go check out our 2025 AI predictions in roadmap series. Y', all like you have to. It's. It's wild, right? So played these a month ago and so many of the things that I talked about, which at the time seemed a little weird, had, have already come to fruition. So go check those out on our website. Episodes 443 through 447. All right, let's get into it. So each, almost every single Monday, we bring you the AI news that matters. So what that is is, well, every single day we do this and then on Mondays, that's kind of our guide to help you start the week off right so you can walk into the workplace, knowing exactly what's happening in the world of AI and how it will impact you. So, you know, we cut through the fluff, cut through the PR and hopefully just bring you the realist on what actually matters. All right, let's get into it. And hey, thank you live stream audience for tuning in. Big bogey face, Sandra and Dr. Harvey joining on YouTube. Douglas, Alan, Brian, Woozy, Graham. Lots of faces in the house. Max joining from Chicago, holding it down like me, Parimi, Pedro. Everyone else, thank you for joining. Appreciate you all tuning in. So OpenAI is not for sale, regardless of what Elon Musk is trying to say. So Elon Musk's bid to acquire Open AI has been officially rejected amid the ongoing feud with the old company that he helped co found. So OpenAI's board has unanimously rejected a $97.4 billion acquisition bid from a group led by Elon Musk, reinforcing its commitment to remain independent and focus on its new nonprofit status. So the rejection was made public through a statement on X, a social media platform obviously owned by Musk, emphasizing that OpenAI is not for sale and aims to ensure that artificial general intelligence, or AGI, benefits all of humanity. So CEO Sam Altman humorously responded to Musk's takeover bid by attempting to instead buy Twitter or X for 9.74 billion billion, which I think would be overpaying for it. You know, I know Musk bought it for like 44 billion, but most estimates say it's worth like 10 now. So it's just kind of highlighting this ongoing tensions between Musk, a co founder of OpenAI, and the current OpenAI leadership, namely CEO Sam Altman. So Musk has a long standing feud with OpenAI's leadership, accusing them of straying from their nonprofit ideals, and is currently involved in a legal action to prevent the company from, from becoming for profit. OpenAI's attorney dismiss Musk's offer as a not serious bid, further underscoring the board's unanimous decision to reject it. Y', all, can we just be honest? Most of what Elon Musk does in the AI space, it's, it's, it's for engagement, it's, it's for theater. It's, it's to drive a narrative, right? Unfortunately, a lot of people believe this, right? Which is sad, right? Spoke to a group of CEOs actually on Friday, you know, helping them, doing a keynote for them in a workshop. And one of their hesitations in rolling out Open AI was, well, they weren't sure on what was going to happen with this whole, you know, Elon Musk and OpenAI takeover. Right. So, you know, people that don't follow this every single day, like me, probably didn't realize that this wasn't a serious, it wasn't serious from Elon Musk. What he's trying to do is he's trying to essentially drive up a fair market value price for open AI that they'll have to deal with when it comes to their transition from a for profit to a non profit. Right. Also, I, I don't know, it's, it's kind of funny coming from Elon Musk. You know that he said when he released Grok, I think Grok 2 that he was going to open source Grock 1 said the same thing, you know, oh, when we release Grok 3, he's going to open source Grock 2. Right. So his, his big thing at least, you know, on paper. And by on paper, I mean on mean tweets is, is saying that, you know, OpenAI is, is no longer it's, you know, nonprofit mission and no longer, you know, open source. Right. Like he. But that was never the case. Right. And there's been so many emails and lawsuits floating around. But this is, you know, it was more or less, I think, if I'm being honest, what I think this was all about was to drum up conversation specifically for releasing Grok. Right. Because that is our next AI news story. So Elon Musk Grock 3 is set to launch tonight, at least according to Elon. So Elon Musk's XAI Group is preparing to launch Grok3, which is being described as the smartest AI on earth tonight. So yeah. Monday, February 7th. So this new AI model is expected to surpass its competitors by offering enhanced reasoning capabilities, faster computation speeds in a groundbreaking approach to learning. Grok3 has been trained on Xai's Colossus supercomputer, which uses an impressive 100,000 Nvidia H100 GPU. So yeah, it is the largest kind of AI supercenter that's up and running right now in the world and, you know, impressive. And hats off to XAI and Elon and the team for getting that set up so quickly. So that setup provides 10 times the computing power to train Grok versus its predecessor. So the AI reportedly employs synthetic data sets to simulate a wide range of real world scenarios. So this approach helps in, well, on paper, reducing bias and enhancing the system's adaptability. Grok 3 reportedly features self correction mechanisms that allow it to refine errors in real time and maintain logical consistency. Elon Musk goal to redefine AI developments is positioning XAI as a significant contender against established players like Google open AI and anthropic. A recent $6 billion funding round has attracted major investors such as Nvidia and amd, providing XAI with a somewhat of a competitive edge in the market at least. You know, there's very few companies that have raised 6 billion dollar plus. You know, I'm curious, live stream audience, are you going to be using grok? My hot take? Right. And this is judging by the original Grock and Grok 2, I would never use it. Not for a real business use case. I would never tell a business or advise a business to use it either. We'll see if Grok3 changes that. One of the main reason is that unfortunately Twitter Slash X has been a main source of its training data, which is not good. Right. And you might say, oh Jordan, well you know, Llama uses Facebook data and you know, Instagram data and it's like, yes it does, but not as one of its main training sources. So, you know, kind of one of the biggest USPs for Grok has been its real time integration with XData. Right. And I've done plenty of videos on this channel and it actually does worse of a job at connecting in real time to X than other large language models do. So it's, it's kind of or at least than other large language models do in connecting to real time data. Right. So you would think, oh, if I could use Grok, you know, I can know up to the minute information and it often doesn't work like that. So I mean, we'll see. I would love to be proved wrong. I do think if I had to give predictions, I think that GROK will bench very, very well. I think it could be a, you know, maybe top three or number one model in terms of ELO scores on the LM arena. But I really don't think that we're going to see widespread adoption, especially here in the US for that very reason. Right. You have to have a high level or at least some level of trust in security in how models were trained. And if a large portion of Grok 3 at least is trained off of real time Twitter and Axe data, that's not something most companies are even going to stick their toe in. But you know, people on Twitter are going to yell at me for this because they love Elon and everything he does and they say everything he touches is gold. But you know, again I don't think most serious businesses will use Grok for the most part. Joe Joe here on LinkedIn says I won't touch Grok, just like I won't touch Deep Seek. I will say from a data and privacy perspective, I will say Grok is should be safe, right? It should be safe to use because it's based here in the U.S. so you know, we have our, our nice data and privacy laws protecting us. But yeah, I still, I still don't think we're going to see much widespread use of Grok, right? It'll probably explode on Twitter. Like I said, it's probably going to do a great job. Some, some different benchmarks, but I think that's probably where it will end. All right. Speaking of things that I wish were a little better Tech Giants Siri Sorry, Apple and Amazon are facing some major AI powered delays in their voice assistance. So yeah, seems like Alexa and Siri might not be getting very AI smart anytime soon. So Apple and Amazon are both experiencing significant delays in launching their new and improved AI enhanced voice assistance, primarily due to software glitches and inaccurate responses during internal testing. So that is According to Bloomberg, Apple's revamped Siri, initially planned for release in April, may be postponed until May or later due to persistent software bugs and engineering issues. Apple's major Siri enhancements are supposed to include improved user data access, app control and screen context understanding are not consistently performing as expected. Similarly, the Washington Post reports that Amazon is delaying the release of its large language model powered Alexa, initially launched as a limited demo in 2023 due to incorrect responses found during testing. Amazon plans to announce the ALEXA update on February 26, but public access is not expected until at least March 31. So the new version of Alexa will offer premium features like personality customization, call reminders and service ordering capabilities, although some features are already available in the current free version. Apple is considering various launch strategies for its new and improved Siri, and including adding new features in April's update, but disabling them by default until the iOS 18.5 update in May. A fully LLM powered Siri capable of human like conversations is not expected until next year. Why? Why? Why? I don't know how. Like I don't know how many times I got a bellyache about this. Foreign We've had very great AI voice assistance from open AI and from Google for many months. All right? So yes, Google is a giant tech conglomerate, but we've had extremely impressive options now dating back to the latter part of 2024 and reportedly Amazon and Siri have been working on this for many years. Reportedly many, many millions of dollars. I don't understand. I don't understand how I still can't ask Siri or Alexa for absolutely anything other than the time and the weather, which is like exactly what I used it for 10 years ago, right? I don't understand. I, I literally don't understand. And, you know, I've been lucky enough to talk to people that have been very involved in its development, and I still don't understand how Apple and Amazon have dropped the ball so hard, right? Because I do think that we're going to see, especially with some, some, some new advancements from OpenAI and from Google. I think those are going to catch on very mainstream, right? I think, first of all, I mean, you have to tip your cap to Amazon and Apple for having this technology. What, 10 years ago, at the time, it did feel nice, it felt advanced. Now it's obviously very stale and frustrating to use, but now I just, on my new, on my new iPhone, I just programmed, you know, the side button to automatically launch Chat GPT voice search, right? So now it's like, at least on my phone, I don't know if I'll ever use Siri, right? As long as I can touch one button, I'm instantly chatting with Chat GPT, which is seemingly a hundred times smarter and a hundred times more useful than Siri. I don't think Siri can really do anything right now, which is a little disappointing. But, hey, hopefully that'll change. Hopefully that'll change. All right, next piece of AI news. Perplexity has released its version of Deep Research to compete with OpenAI and Google. So Perplexity AI has launched its version of Deep Research, a new mode designed to provide extensive research capabilities in minutes, making it newsworthy for its potential to transform how users conduct complex research tasks. So the big, I think news here is it's free, right? For. So users with free accounts can make up to five queries per day, while pro users essentially have unlimited use of it. You can, I guess you're capped at 500 deep research queries a day. But like, I'm doing the math, right, because they take two to five minutes, that'd be very hard to actually hit 5, 500 of these deep Research queries a day. So yes, users with a free Perplexity account can do up to five. And essentially, if you're a paid Perplexity Pro user like I am, you have unlimited access to this feature. So the feature is adept at handling expert level queries in fields like finance, marketing and product research, offering detailed responses quickly. Deep Research has shown strong performance, scoring a 21% on humanity's last exam, a new kind of AI benchmark for large language models. So the company is expanding its reach by planning to offer deep seq research also on iOS, Android and Mac platforms. And it is currently live and available on the web. So despite facing legal challenges from media organizations over copyright issues, Perplexity AI is not only just rolling this out and still it looks like it using some of these even same sources that they're facing some legal challenges against, but also they're currently negotiating revenue sharing agreements with publishers like Time and Fortune, according to reports. So. Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Gen AI. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for ChatGPT training for thousands or just need help building your front end AI strategy, or you can partner with us too, just like some of the biggest companies in the world do, go to your everyday AI.com partner to get in contact with our team. Or you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on Gen AI. So yeah, great, great comment here From Max on LinkedIn. So saying that you can use deep seq R1 through perplexity too. Yeah, so it's a little confusing because they've, they've had essentially a reasoning search where you could tap into either Deep seeks R1 or OpenAI's 03. So they did have this kind of, it was like almost like a research, like a deep research light version because they had essentially a reasoning search already available and, and it is a little confusing. So actually what we have going tomorrow, we are doing a deep research perplexity deep research episode. So we're going to be talking about what it is and if you should use it. All right. And then yeah, back to back because I think this new deep research category, it is going to be huge Right. And I think one of the reasons being there was a lot of initial. Which makes sense, right? But there was a lot of initial hesitation for companies to use large language models because they thought, okay, all this means is our employees are going to be copying and pasting, right? They're going to just copy and paste things from these large language models, which is often bad, right? Because large language models, at least like two years ago, unless you were using ChatGPT, which had good access to the web via browse with Bing, Google really struggled to, or, sorry, Google Gemini really struggled to use Google, right? Microsoft Copilot did a good job at accessing Bing. But for the most part you run into a lot of issues that people were just lazily not really trying to understand how these AI models worked. They would just copy and paste things from a large language model and oftentimes that's just using its internal training, its internal data set, which, you know, people always look at the cutoff date, right? Like as an example, the Cutoff date for GPT4O right now is June 2024. And they're like, oh, all the information out of here is from June 2024. No, it's not, right? So much of what makes it into these large language models can be many years old, which makes sometimes the outputs kind of dangerous to use, especially for novice users who don't know what they're doing. Which is why I think the deep research boon is going to be huge, right? Because essentially what this does is it ensures a higher accuracy rate because these Both from Perplexity, OpenAI and Google, it really goes, and it sometimes researches dozens or hundreds of websites and then it kind of relies on the underlying model. So I do think this is going to be a sweeping new category of tools and it is going to be kind of like a second ripple, right? So we kind of saw the, the chat GPT moment of November 2022, which really brought generative AI into the mainstream. I do think for those companies that have not widely adopted generative AI yet, I do think this is going to be kind of the second, I won't say wave, but this is going to be a second ripple, right? Most companies, they don't care about AI images, most companies don't care about AI video, but they are going to care about this, right? So that's why we're actually doing back to back episodes. So tomorrow we're gonna be going over. So tomorrow, Tuesday, February 18th, we're going to be going over Perplexity deep research. And then the next day so Wednesday, in two days from right now, we're going to be doing a Deep Research throwdown. We're going to be comparing perplexity, Google and OpenAI's versions of deep Research. I'm excited about this episode. I'm a geek, y'. All. Like, yeah, I spent like my weekend doing this, you know, making charts and testing things, right? Like, this is what we do. I spend hours at night and on the weekends testing all of these things so you don't have to. And I share my results and I give it to you straight, right? I'm not gonna lie to you on, oh, this one's great. This one's garbage, right? So make sure you join us for those episodes. All right, our next piece of AI news. Well, the world's most powerful AI video tool is now publicly available, but not in the way that you would think. So YouTube has upgraded its Dream Screen feature with Google's new VO2AI model, enabling faster and higher quality AI generated backgrounds and standalone clips for YouTube Shorts. So, yeah, the VO2 access was extremely limited, right? It still is not generally available, but you can now use it in very limited fashions, right? So if you're working on a YouTube short in the Dream Screen feature, then you can use it. So the integration of VO2 allows for quicker generation of AI videos and images, supporting a wider range of subjects with more detailed visual output. New capabilities include the creation of standalone video clips in addition to AI driven backgrounds, offering creators more versatility in content production. It's not available to everyone yet, but the initial rollout of this feature does cover users in the U.S. canada, Australia and New Zealand with plans for future expansion to other regions. So users can access Dream Screen by selecting the Shorts camera within the YouTube app. Okay, so open the YouTube app, go to the Shorts camera, then you can access the Dream screen powered by VO2 by tapping the Green screen option and entering a text prompt to generate AI backgrounds or standalone clips. Google is trying to ensure a little bit more AI transparency by applying their Synth ID watermark and to create clear labels that it is AI generated media. So not exactly the rollout of Google's V2 that a lot of people were hoping for. But anyone that wants to at least go test it out to see if, especially I think for, you know, creative companies, visual companies, or any marketing department, really, I think you should at least be testing this out. You should be looking at it. So when and if we get VO2 access, right? V1 was never publicly made available. VO2 still not publicly made available. If you had good timing on their Discord server. You probably got, you know, early access to it, but most people do not have access to it just yet. All right, let's get to some Open AI updates. So OpenAI CEO Sam Altman announced on Twitter, slash, act, slash, whatever you want to call it, that they have released an updated version of their GPT4O model. So this normally isn't very big news, right? Because OpenAI and other AI companies do release kind of under the hood updates. So what that means is they'll release updates to a model without necessarily renaming it. Right? So this isn't a GPT 4.5 or anything like that, so it is still called GPT4.0. However, it is extremely more impressive. Right. I don't know if anyone else has been using it over the last two days, but no real updates on what's included, which is bonkers. All right, so hey, AI companies, sometimes you do a great job at this, sometimes you fail, right? So OpenAI released this out. No change log, no model notes, absolutely nothing. Right? So why is that problematic? Well, when companies roll out updates to models, especially on the API, right? So you might be running a huge part of your Software off of GPT4O or GPT4OH Mini as an example. So if there's not detailed updates on what was changed, you essentially just have to do a bunch of engineering over again to try to figure it out on your own. So I'm sure there's thousands of companies right now and grumpy developers that are waking up Monday morning and being like, oh, you know, they're getting tickets on saying, hey, you know, our system isn't working correctly, you know, we're not getting ideal customer support, right? Whatever you're using the GPT4Amodel for, it could be going much better or much worse. Right? So kind of disappointing that OpenAI doesn't release a little bit more details. But what I've seen and what other people have talked about, a better personality, a better writing style, less censored, and just better benchmarks. Right, let's check this out. So, looking at the LM arena chatbot leaderboard, the GPT4O model is back in a three way tie for first. So if you don't know what the LM Chatbot arena is, essentially you go in, you put a prompt, you get two responses. You don't know who's like, what company it it is. You vote on which one's better. So it's essentially like a blind taste test, right? Like the Pepsi vs Coke marketing campaign of the 90s, right? And then the winner gets a score, right? So you have this thing called an ELO score and then there's kind of a margin of error. So what that means is that technically GPT4O is in a tie for the top ranking with Gemini 2.0 Pro, the new version that just came out, and Gemini 2.0 Flash thinking both of those were just released in the last couple of weeks. So one other thing, kind of some unique reporting here from us. Well, looks like OpenAI is taking even more shots at Google. Some original reporting here. I didn't see this anywhere. I didn't know if this was a mistake that, you know, showed up over the weekend. You know, on the weekends is normally where I do a lot of my LLM testing because during the week I'm busy bringing you guys a daily live stream, podcast and newsletter. And I noticed that I'm like, what is this? Is this Open AI is showing an autocomplete suggestions? Yes, they are so interesting here. A very Google ask, right? So when you go into Google search and you start typing something and you see all of these suggestions, right? It's actually very helpful in Google search. I don't know how helpful it will be in ChatGPT, but this is a new feature. I didn't see literally anyone else in the world because I was googling it. I was using AI search engines to look for it. Looking on Twitter, I couldn't find anything. So it could be a feature that is slowly rolling out. But essentially when you start to type a query, you may get autocomplete suggestions. So a couple new things from open AI. But speaking of it, the big news, I think the, probably the biggest piece of news this week is Sam Altman announced on Twitter their upcoming essentially road map and releases for models, which is something that they generally don't do a lot. Kind of the open eye play, the open AI playbook over the last like year has been kind of these just like teasing online, right? Like they drop these little hints, it's almost like little Easter eggs and you know, then they build up a ton of hype and then you know, either they're going to announce something out of nowhere or you know, like they did in December. They had the 12 days of OpenAI, you know, ship miss. But this one is interesting because actually a much higher level of transparency here from Sam Altman and really just talking openly about some of the challenges that they're facing and some of their plans. But the biggest thing is is announcing the upcoming release of GPT4.5 and GPT5. So they are aiming to simplify the company's complex product offerings. So Altman described GPT5, which is known internally as Orion, as the final non chain of thought models. Okay, so this is marking a shift toward a unified AI system. So there, even after this tweet announcement and all the press coverage, there's still some things that are a little vague, there's still some gray areas. So just keep that in mind and you know, we'll, we'll link to that tweet thread in our newsletter today if you want to read it. But essentially he said that GPT5 is going to be more of a system and it's not. And it's going to integrate different reasoning models into a cohesive system. So the free tier of chat GPT in the future will have unlimited access to GPT5 while paid subscribers will get higher intelligence levels and then pro subscribers, which are those paying, you know, 200amonth will get the highest. So it doesn't. So it looks like essentially GPT5 will not be a traditional GPT model because GPT 4.5 is going to be the last of the non reasoning models. So that is the standard transformer model. So it looks like GPT5 will still have a. Maybe it's just an updated version of GPT 4.5, I'm not sure. But essentially it's going to have some sort of GPT5 that's going to be more of a system that is going to recognize if it needs to use a transformer type model to answer your query or if it's going to use a reasoning model and then at what level it is going to give it to you as well. Because as an example right now, right, I have a pro plan, so I have access to GPT4.0, GPT4O mini GPT O or sorry, OH1 01 normal 01 pro O3 mini, OH3 mini high, right? It's confusing. And then you have all the tools, right? You have canvas, you have tasks, you have operator, you have deep research, right? So it looks like GPT5 is going to be more of a system and it will essentially have use of all the tools and all of the models. And so it's not going to be going into this, you know, pick your model, choose your tool. It looks like GPT5 will be a system and will have use a combination probably of a transformer type model and, and a O series or a reasoner model, but doing it all under the hood so Altman indicated that these releases are expected within weeks or months. All right, so, you know, it could be six months, who knows? Or it could be as soon as a couple of weeks. So they're aiming to streamline the user experience by eliminating that current model picker. And OpenAI has been very transparent that they kind of stink at naming models. Right? Yeah. Like O3 mini. O3 mini high 0101 pro GPT4.0. GPT4.0 mini. Right. It's, it's, it's, it's so confusing. It's so confusing. So despite delays and reports of increased costs, Altman dismissed claims that AI progress has stalled, maintaining optimism about OpenAI's advancements. Yeah, it's funny. Yeah. Denny here is saying, when is OpenAI going to hire a marketing person to help name its models? Well, it looks like this new GPT5 system approach, there might not even be any model names. Who knows? I mean, we'll see. I'll save most of my judgment for when this comes out. I'm obviously a power user, right. I think one of the problems maybe is, you know, I'm guessing 99% of the population is not a power user like me. Right. I generally use large language models. Depends on the day. I'd say anywhere from nowadays three to ten hours a day. Right. So let's say I average using large language models for like five or six hours. That's a ton of time. I, I have no problem right now with the current feature sets of OpenAI or any other tool. But I, I do assume the overwhelming majority of users can get quite confused. And that probably keeps people maybe from paying a paid subscription or upgrading to a higher tier because they're just confused. They're like, I don't know what any of these things are, let alone if I need them. Right. But we have seen a huge push from OpenAI. And y', all, this is in response, I think, to Google. People are like, oh, you know, Deep Seat came out with this amazing model. No, go look at Deep Seat now. It's, it's banned in so many countries. US businesses, I think, aren't going to be adopting it at a high level. Right. Let's just be honest, right? There's, there's legislation out there that would find companies for using it. I don't think that version of the bill is actually ever going to make it into actual legislation. But, you know, Deep Seek had its moment for like a week and now it's, you know, it's no longer the cheapest model in terms of, you know, price per quality. It's not even close, right? So I do think OpenAI did this in response to Google, right? That's, that's who they're going after. Y' all. Look at everything they did. They made chat, GPT search free. You don't even have to be logged in. They made their biggest model or one of their more powerful models in O3 mini free for limited use cases, even for free users, right? People are like, oh, this isn't responsive. No, it's not, it's not in response to Deep Seek people. I don't know. I just don't think people like to use their brains. Open AI is going after Google, and Google went bonkers in December and they continue to go bonkers, you know, all the way through February. They're very silent for a long time, right? But Google came out straight up swinging. So I personally think that. And yeah, I talk to people off the record, you know, at OpenAI and at Google, but this is my personal, my personal thoughts. I think all of this is in response to Google, right? Even Sam Altman, you know, on, on Super Bowl Sunday, there was a tweet or maybe it was the Saturday before, looking at the top websites in the world, and OpenAI was, had actually the biggest growth of any of those. I think it was a list of the top seven. And his response was, have a long way to go to catch Google, because Google was obviously number one. That's who they're after. They're not after. They don't care about Deep Seek. I don't think they necessarily care even about, you know, Claude. Right. Even though that's probably on paper their closest competitor. They are going after billions of users. All right? That's what they're going after. All right? Speaking of some of their closest competitors and new models. So Anthropic Claude is set to launch a new hybrid model in weeks. So Anthropic is preparing to release its next major AI model, which is expected to arrive within weeks, according to reporting by the Information. So the new model is described as a hybrid capable of switching between deep reasoning and fast responses, offering flexibility for various applications. So a sliding scale will be introduced for developers, allowing them to manage costs associated with the model's deep reasoning capabilities. So according to the report, Anthropic's soon to be released new model outperforms OpenAI's O3 mini high reasoning model in specific programming tasks and excels at analyzing large code bases. So CEO Dario Amodi expressed the company's focus on creating Differentiated reasoning models, questioning the distinction between normal and reasoning models. So this development highlights Anthropic's commitment to advancing AI technology and could significantly impact businesses relying on AI for complex problem solving. So, I mean, we'll see, we'll see what actually happens here. And I'd be very interested on the timing, right? So I don't know, we might see a, a GPT 4.5 release right after we get this new model from Anthropic. Because it's been, I, I don't know, I lost track. It's been like nine or 10 months since we've seen an update from Anthropic, which in like AI and large language model terms is like decades, right? Claude's they did update kind of Sonnet 3.5 clots on it 3.5, but they still just called it quads on it 3.5. So was it really an update, right? Or is it what just, you know, OpenAI did over the weekend with GPT4, you know, but Claude 3.5 Sonnet still does remain extremely popular for developers and for coders. Right. So even though we don't talk about it a ton on the show, that's because right now I don't think there is a ton of general business use cases outside of just development for Claude 3.5 Sonnet. I think in almost everything else you should be probably looking at or using Google Gemini models or OpenAI models or you know, obviously Microsoft Copilot is powered by OpenAI's model. So for the most part I don't see a ton of use cases. And one of those reasons is, well, Claude has been offline this whole time, right? So there's been reports that Claude will have some Internet connectivity maybe alongside these reasoning models, maybe it'll be a different release. But you know, that's one of the reasons I've been saying all along, companies shouldn't be using claude. Right? Unless you're using it on the back end. But you know, for front end users, which is a lot of times what I'm talking, talking about on this show, right? If, if you have a chat GPT enterprise account or if you have, you know, Google Gemini across your workspace or maybe you have, you know, Microsoft 365 copilot, those are all connected to real time information on the Internet, right? CLAUDE is not on the front end. So if you're using, you know, Claude AI, if you have a, A teams account, whatever, you could be working with extremely old information which makes the outputs unreliable, period. So I am excited to see what Anthropic releases next and I hope this new hybrid model includes the ability for it to have real time information from the Internet. All right, more tech giants in the fold this week. So we have Apple and Meta reportedly competing in the AI powered humanoid robotics field. So Apple and Meta are entering a new territory for their rivalry focusing on AI powered human eyed robots, according to a report from Bloomberg. So the competition between these two tech giants is significant as it could really change the way that everyday tasks are automated in homes, impacting industries and consumer lifestyles. So according to reports, Meta is adopting a software first strategy to creating a new unit within its Reality Labs division to develop a foundational software platform for humanoid robots similar to Android 4 robots. So the company aims to leverage its expertise in sensors, computing and AI, including its Llama AI models to gain a competitive edge. Meta has started hiring engineers to support its vision of helping other companies bring humanoid robots to market in the coming years. Apple on the other hand, is focusing on seamless hardware and software integration, developing a tabletop device with a robotic arm that could either be extremely weird or extremely cool, and exploring mobile robot concepts akin to Amazon's Astro. So the company Apple may eventually create a full fledged humanoid robot competing with Tesla's Optimus or Figures 01 and 02, although it faces huge challenges in AI development, particularly with those buggy Siri things that we just talked about. So Apple's decade long experience in self driving car technology could provide valuable insights for its robotic navigation within homes. Meta's robotic division is led by Mark Whitten, former CEO of GM's Cruise Self Driving Units, while Apple's initiative is headed by Kevin lynch, known for his work on Apple Watch. All right, last but not least, Adobe has launched public beta for its AI Video generator, enhancing its creative tool sets. So Adobe has released its Text and Image to Video AI generator Generate Video to the public in beta form, marking a pretty significant step from the powerhouse that is Adobe. So this tool is part of Adobe's Firefly web app, which now includes updated image generation and translation capabilities, making it a comprehensive platform for creators. So Generate Video offers two main features. Here's here's the gist Text to Video and Image to Video, allowing users to create video content from text descriptions or by using reference images. So the video output quality has been improved to 1080p HD at 24 frames per second, though the maximum clip length is currently limited to just five seconds. So bummer there. Adobe is working on both a faster and a lower resolution model as well as a 4K model to enhance future capabilities. So it looks like they're kind of going this, you know, small, medium, large. And it looks like what's available now is probably more of a medium tier. The Firefly web app now integrates with creative cloud apps such as Photoshop and Premiere Pro, streamlining the workflow for those users. Adobe emphasized that Firefly, this is the big thing, is trained on public domain and license content, ensuring its tools are safe for commercial use. And that will be a big talking point over the next couple of years, if I'm being honest, because both OpenAI and Google have been a little vague and every other company, right, Runway, Luma, everyone else, they don't really talk about how their video models are trained. Presumably they just scrape any video or photo it can find on the Internet. Presumably. Right. But Adobe has really emphasized more of a safer approach, you know, using tools that it has for commercial use. Although previous, previous reports have said that up to, I believe it was 5% of what Adobe had trained previously for. Some of its visual models included images from Mid Journey, which kind of makes the whole like, oh, we have commercial use rights for this a moot point if you, if you're using things from Mid Journey. So the AI video market is becoming increasingly competitive, with Adobe facing challenges from companies like OpenAI, Google, ByteDance, Pika Labs. There's another handful or so from China over the last month that have really been impressive. So two additional tools called Scene to Image and Translate audio and video are also now available in public beta, offering new creative possibilities for users. So how much will it cost? Adobe has introduced two Firefly subscription plans, Firefly Standard at 9.99amonth and Firefly Pro at 29.99 per month, each offering various levels of credits and access to features. All right, that's it. Let's do the very quick rundown. As a recap, here is the AI news that matters for this week. Number one, Elon Musk bid to acquire Open AI was rejected amid their ongoing feud. Number two, Elon Musk's Grock AI chatbot is set to release tonight. So we'll see how that goes. Number three, we've seen some pretty sizable now delays according to reports for Apple's AI powered Siri and Amazon's AI powered Alexa. Number four, Perplexity has unveiled its Deep Research tool. So make sure to join us tomorrow as we give a rundown on that and then on Wednesday as we compare the three major deep research tools. AI story number five, YouTube. We finally now have a VO2 release, but just on YouTube shorts, so very limited scope that you can use it. AI news story number six GPT4O we saw some under the board, sorry, under the hood updates from the GPT4O model that shot it back up to tied for first in the LM Chatbot arena leaderboard. Number seven, OpenAI CEO Sam Altman went to Twitter to talk about new GPT5 updates and kind of how these models might just get married together. Number eight, finally seeing some movement from Anthropics Claude and they will reportedly be launching a new hybrid AI model in the coming weeks. Number nine, both Apple and Meta are reportedly going much deeper into the AI powered humanoid robotics space. And last but not least, number 10, Adobe has launched a public beta for its AI video generator. All right y', all, I hope that this was helpful. We spent a ton of time not just on our show every single day, but on this show, the news that matters. There's literally hundreds of news stories every single week. I don't want you to be confused, right? I don't want you to be worried about, oh well, what does this mean? Right? That's what we do. So if you can only join us once a week, join us Mondays so you can keep up. I hope you join us every single day because I think that there is so much value in tuning in every single day. Yes, the show might not, you know, be highly applicable to you every single day as we bring on guests across various fields, but there is so much to learn. So thank you for tuning in. If this was helpful, please click that repost button if you're listening on the podcast. Thank you. As always, please make sure to subscribe to the podcast if you haven't already. If you could leave us a rating, we'd super appreciate that. All right, so. And make sure while you're on our website signing up for our free daily newsletter@your everydayai.com make sure if you haven't already. Why the heck have you not listened to our 2025 AI predictions and roadmap series? It's free to listen to. They're short episodes, 25 minutes, so go give episodes 443 through 447 a listen. Thank you for tuning in. We'll see you tomorrow and every day for more Everyday AI. Thanks y'.

All. 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 a 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.