Everyday AI Podcast – An AI and ChatGPT Podcast - Ep 513: OpenAI’s new open model, Copilot updates, Perplexity going after Siri & more AI News That Matters

Episode Date: April 28, 2025

Is Perplexity going after..... Siri? Talk about a hard pivot. OpenAI and Google are racing for users.... who's winning? And will the U.S.'s effort on AI in education be too little, too late? We'll answer those questions and a ton more on our weekly news roundup show. Don't spend hours a day trying to keep up. Just join us (most) Mondays. Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Thoughts on this? Join the convo.Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:OpenAI Launches Lightweight Deep Research ToolDOJ Pushes Google Chrome BreakupMicrosoft 365 Copilot Spring 2025 UpdateAdobe Firefly Supports Google, OpenAI ModelsMyPillow CEO's AI-Generated Legal TroubleOpenAI's Image Gen API for DevelopersU.S. President Signs AI Education OrderPerplexity AI Challenges Siri with Voice AssistantTimestamps:00:00 AI in Education, Adobe Partnerships05:52 "Premier Deep Research Tools: OpenAI & Google"07:35 DOJ Proposes Google Chrome Sale13:42 Adobe Expands AI Image Tools17:08 AI Missteps: Lindell's Legal Trouble18:43 Lindell's Legal AI Misstep22:25 OpenAI ImageGen API Overview28:27 AI Literacy Initiative Praised31:29 Microsoft Launches Controversial Recall Feature32:32 Improved Microsoft Search Enhancement Secured38:30 "Perplexity: Contextual AI Assistant Edge"41:22 "Perplexity's Crucial Pivot Needed"43:18 OpenAI's New Reasoning Language Model46:43 AI Usage Surges Amid OpenAI Speculation50:54 Tech Updates: AI Expansions & Legal IssuesKeywords:OpenAI, Lightweight deep research tool, chat GPT, Free users, Paid users, Deep research queries, O four mini model, Google, Gemini 2.5, Perplexity, Deep research, Adobe, Firefly AI, Competitors, Image generation, Microsoft, Copilot, AI features, MyPillow, Legal citations, AI-generated court filings, ImageGen, API access, Creative work, US president Trump, Executive order, AI education, Microsoft 365, Recall feature, Copilot plus PCs, Perplexity AI, Voice assistant, Siri competitor, Open reasoning language model, Llama, Open weights, Meta, Text in/text out, DeepSeek, AI safety, Sam Altman, Edge AI, Multimodal models.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. The United States may finally be taking AI in education seriously.

Adobe is teaming up with an unlikely group on its AI platform, its competitors. Perplexity is going after Siri, seems like a hard pivot. And we have some new updates on OpenAI's rumored Open Model. A lot going on in the world. world of AI news as it is every single week. And if you missed it and if you're wondering, what happened, how is this going to impact my company and my career? You're in the right place. What's going on, y'all? My name is Jordan Wilson and I'm the host of Everyday AI. This is your daily live stream podcast and free daily newsletter helping us not just grow from AI, but how we can

actually, you know, use it to change the trajectory of our careers, of our, uh, careers of our companies. So if that's what you're trying to do, trying to be the smartest person in AI at your company or maybe in your department, this is the right place to do it. So if you're listening on the podcast, thank you for tuning in. As always, make sure to check out your show notes. There, you can find a link to our website, which is where you need to be, your everyday AI.com. So there, you can sign up for our free daily newsletter, as well as you can go listen to, I don't know now like 510 plus back episodes on our website read about them watch about them uh you know whatever you want to do it's all for free on our website sorted by category so no matter what you're

trying to learn in the world of AI we have it for you there all right also before we get started on our AI news weekly wrap wrap up quick reminder or I guess an announcement tomorrow's show we're going to be going over Google's AI studio five time consuming tasks you did I didn't know you could automate. Yeah, I think a lot of people have been sleeping on Google's AI studio. I use it every day. And I'm like, I should just probably share with everyone the types of things I'm doing in AI studio.

It'll probably really help people. So, hey, let me know if that sounds good. Live stream audience. Good to see you. Rolando, join it from South Florida. About the weather there. It's fantastic. YouTube crew going hard today.

Richard, Big Bogie, Douglas, joining on the YouTube machine today. Kyle, Michelle, Sandra, Keith, R. Everyone else. Can't get to everyone. Thank you for tuning in. Yeah. Maybe you listen on the podcast now and then. You can always hang out on the live stream.

We do it on on LinkedIn, YouTube, Twitter, 7.30 a.m. Central Standard Time Monday through Friday. Come hang out. Ask questions of our guests. Today it's just me keeping you up to date with the AI news. So enough to chat. Let's get into the AI news that matters for the week of April 28th. All right. Open AI has introduced a new lightweight version of its deep research tool and has expanded access to free users.

So OpenAI is rolling out a new lightweight version of its deep research mode for chat GBT, making this powerful tool more accessible for free users while managing operational costs. So the new lightweight deep research mode is now available to all chat GBT users, including free accounts. If you are on a free account, you only can use five queries a month, which isn't a ton, so you got to make them count,

but paid users, which includes chat GPD Plus, teams, enterprise, and education, receive 25 deep research queries monthly, while pro users on that $200 a month plan get 250 queries a month. With the advanced version, still available but capped before switching to the lightweight mode.

So yeah, on the paid plans, you do get the full version, kind of the full weight version. And then once you hit that cap, you are switched over to the lightweight version, which is what the free users are using by default. So the new lightweight version uses OpenAI's 04 mini model, not the 03 full model, which the kind of full version of chat. you uses. If you don't follow models, that might be super confusing. You're like, wait, the paid more powerful version is 03, but the lightweight version is 04 mini. Yes. So it's run by 04 mini, not even 04 mini high. So yeah, kind of the in terms of power for the thinking

models, right, depending on where you want to put 01 on this. But I would say it's 04 then, or sorry, 04 mini high, then 03, and depending on what you think of 01 Pro, you know, that might fall somewhere, you know, first or second. So OpenAI's goal is to reduce the heavy resource load caused by complex deep research queries, which require the model to analyze articles, papers, and extensive data to provide thorough answers. So this rollout begins immediately for free users with enterprise and education users, gaining access shortly thereafter. So despite research tools like this are becoming common among AI chatbots with alternatives such as Google's Gemini 2.5 Pro deep research, which is really, really good, much better than their

previous version. And perplexity has a deep research. Grak has a deep and deeper search, whatever they want to call it. So, you know, I wouldn't say the tools are a diamond dozen now, but I would still say, at least for right now, it is OpenAI and Google that are kind of the premier versions of these deep research tools. You know, I'm curious, live stream audience, what do you all use to do your deep research, right?

I previously did a deep research comparison, but this was on Google's old version. So I do think Open AIs and Googles are kind of in a league of their own. And I think this is one, like if you haven't used a deep research tool, right? Like some people are like, okay, how do I get started with generative AI? If you are brand new, I still think this is one of the best places to start. And I might have to do an updated episode since Google released or updated their deep research tool with Gemini 2.5 Pro, might be time to do and maybe, I don't know, maybe sometime in May we'll do that update. So live stream audience, let me know if you want to see that.

All right. Speaking of Google, so Google is facing a potential breakup as the Department of Justice pushes to force the sale of Google's Chrome browser. So Google was declared a monopoly in the U.S. search engine market by a federal judge, leading to a new phase where the Department of Justice is seeking remedies to break up its dominance with a focus on how this could reshape internet access for users worldwide. But it actually has to do with AI and you'll see why in a second. But the DOJ's top proposal is forcing Google to sell off Chrome, which currently holds about 66%. percent of the global web browser market, arguing that Chrome's integration with Google search unfairly reinforces Google's monopoly. OpenAI has publicly expressed interest in buying Chrome if it is put up for sale, raising questions about whether shifting ownership to another tech

giant would effectively address competition concerns. Other DOJ proposals include requiring Google to share some user data with competitors to promote competition. and in Google's exclusive default search engine deals like Google's $20 billion per year contract with Apple and unbundling Google and apps from Android devices to give users more choice. Google obviously strongly opposes these suggestions by the DOJ claiming they would harm consumers hitting innovation, particularly in AI, and weaken the U.S. tech sector's global standing. The company argues users choose its services voluntarily and that forced changes would

disrupt security and privacy protections tied to Chrome and its open source chromium platform. You know what? I agree with Google here. Google, like, let's just say you're anti-Google for whatever reason. Google has their Chrome browser. They have an open-sourced version of it called Chromium. So that's actually what Microsoft Edge uses. They use this. They use this. same technology, right? So I don't necessarily agree with what the federal government is doing here, trying to break up. You know, I know that they're saying Google has a monopoly on search, but I don't know if this is the way to do it, forcing them to sell off the, you know, Google Chrome to a potential competitor because, okay, then what happens if one of their biggest

competitors takes over Chrome, then aren't they just going to be in essentially the exact same position that Google is in, right? You sell it to Open AI, which is actually leading in the AI chatbot market, and it's not particularly close. So then wouldn't that give OpenAI an unfair advantage? Probably. So according to Google's court documents, right now, their Google Gemini AI chatbot has 350 million monthly users.

Well, ChatGPT has reached almost double that with 600 million monthly users. Although CEO Sam Altman hinted chat GPT's weekly users might be as high as $1 billion. That's crazy. So it should be interesting to see what happens here with this Google case in front of the Department of Justice. Speaking of Google's competitors, one of them in Microsoft, has definitely updated its co-pilot, form, I would say in a big way. So Microsoft has launched. It's Microsoft 365 copilot wave two spring updates introducing some new AI features aimed at helping professionals work smarter and faster. So we have covered a couple of them so far in the newsletter and on this very show, but this is

kind of like the official bundling of all of these announcements we've seen over the past couple of weeks and some new ones. So you have the research and analyst AI agents that are now available through a new agent store. So yeah, Microsoft has a new agent store, which enables advanced research and data analysis with open AI models and partner integrations like Jira and Myro. The skill discovery agent helps managers build skill-based teams and assist employees in finding colleagues with needed expertise, improving collaboration, which about time, right?

About time that we get AI that helps humans find and relate to other humans, right? So it's good to see that element of AI, something that actually encourages and improves human to human, right? Because we always think of AI as kind of like, okay, well, now I'm just going to be chatting with a bot more and not my coworkers. So that is pretty cool to see from Microsoft. Also, there's a new create feature that uses GPT40 to generate brand-compliant AI images for presentations and social media. Also, co-pilot notebooks lets users combine spreadsheets, documents, presentations, and notes into a searchable workspace that can be accessed via chat, streamlining project workflows.

Also, copilot search supports conversational queries across Microsoft 365 apps and third-party tools like Slack and Google Drive, though it doesn't search the web in real time. Also, you have memory now in Microsoft 365 copilot that allows copilot to remember user info more for personalized responses as well as some new IT admin access. All right. So we actually have more on some new announcements from Microsoft here later in the show. All right. More big tech titans with some actually surprising news, this time from Adobe. So Adobe has expanded its Firefly AI system with support for its competitors' models with OpenAI and Google support. So Adobe announced it will add AI image generation models from OpenAI, Google, and others to its Firefly app while also launching a mobile version.

So Firefly users can now generate AI images using OpenAIs, GPT ImageGen, Google's Imagine 3, as well as Google's video product, VO2, and also FlareD. 1.1 Pro models, along with Adobe's updated in-house Firefly models. So Adobe plans to bring in additional third-party models from companies like PICA, ideogram, Luma, and Runway. That's a big one over the coming months. So content created with third-party models can be moved into Adobe's other applications, such as Photoshop using a shared credit system for payment. Adobe did not disclose, however, how revenue from third-party model usage will be split between itself and these providers.

So the company highlights that some users rely solely on Adobe's own Firefly models for legal safety in commercial projects, while others want to experiment with a broader range of AI tools. So Adobe's stock jumped up pretty sharply, 2% shortly after this announcement, indicating there's some strong market interest in Adobe expanding its AI platform access. So yeah, this one wasn't necessarily expecting this. We heard that Adobe would be adding some of these video models such as runway, Pika Labs, etc., which I kind of understood. I did not expect them to integrate with Google and Open AI.

So good on them because, I actually said this on the show a couple of the weeks ago. I said between everything that you can now do inside Google AI Studio, which we're going to be talking about tomorrow. That's one of those five things that you didn't know Google AI Studio could do is how it's essentially like a built-in Photoshop and it's absolutely crazy. So between Google's kind of AI image generation capabilities and open AIs and even Canvas, which we're going to cover soon in the coming weeks,

I'm like, I don't want to be at Toby, right? I don't want to be. But pretty interesting move here, kind of Adobe acknowledging the power and popularity of some of their competitors' model. So instead of trying to go, you know, kind of blow for blow here, what they're doing is saying, okay, come on in and use these models inside of our platform. So the monetization piece will be interesting, right? How is Adobe going to monetize off of people using the Google and Open AI models? I'm sure we'll see reporting on that. in the coming weeks.

All right, y'all, we have our newest case of AI gone bad. All right. And this one, the reason I'm including it in the AI news, number one, it's a cautionary tale for everyone else. But I do think this is going to become the new kind of story of, you know, when you're not using AI responsibly, right? because we had, we, we, we kind of had that, that early one from 2023 that a lawyer submitted some chat GPT hallucinations essentially in a, in a courtroom and got sanction thereafter.

And now the same thing has happened, but probably at a much more visible case. So Mike Lindell, the CEO of My Pillow, is now facing some legal problems amid some AI-generated court filings. So Mike Lindell, the prominent and kind of infamous CEO of My Pillow, is embroiled in legal trouble after his attorneys submitted a court brief containing nearly 30 fabricated legal citations generated. by AI, prompting a federal judge to consider disciplinary action against the lawyers. So Lindell's attorney admitted to drafting the brief manually, but then using AI to finalize it, resulting in false legal references that could undermine the integrity of the case and the legal profession. You know, I don't know, but I'm guessing his lawyer was probably using the free version of Chad GBT and probably didn't necessarily know how to use it. Because if you're using as an

example, deep research or the O3 model and you're doing at least minimum human in the loop, you're not going to get 30 fabricated legal citations. It is still wild to me that we are seeing these stories. But this is probably going to become the next big one. We've been talking about this one that happened in New York in 2023 for now more than two years. And I think this is probably going to be the new one that we're going to be talking about for years to come. So U.S. District Judge Nina Wang has given Lindell's lawyers until May 5th to explain why they should not face sanctions or lose their licenses, highlighting the risk of relying on AI with proper verification in legal work. Also unrelated, but kind of related, My Pillow's business has sharply declined since Lindell aggressively promoted election fraud conspiracy theories following the 2020 U.S. presidential election, reportedly losing $100 million in revenue after being dropped by just about

every single major retailer. So the fabricated citations undermine the validity of Lindell's legal defense, potentially damaging his case and raising broader concerns about the ethical use of AI tools in law. Good. Hey, great comment here. This is why I love having live stream and having smart people on the show. Lisa on YouTube says, I'm a lawyer and have the paid version of chat. It still gets the law wrong. It sounds great, but lacks nuance and is technically incorrect a lot of the time still.

All right, that's a good point. Yeah. And Marie say, guess Lindell's lawyer should have taken your PPP course before submitting their brief. Absolutely. It wouldn't have happened, right? When people properly know how to use a, right? Because everyone just wants to one-shot AI systems, right? You know, they want to lazily work their way in, you know, put in one prompt, try to get like, you know, 10 pages of documents.

If you know what you're doing, then that's a big if, right? But if you know what you're doing, it is large language models, let me just say this. Large language models rarely hallucinate. If number one, you know what you're doing. And you have a knowledgeable human in the loop who's using the right model at the right time for the right purposes and is going through a basic problem. engineering process, hallucinations. If I'm being honest, I use large language models depending on the day, anywhere from, I don't know, it varies five to 12 hours sometimes.

I probably have a 0.05 hallucination rate. Very, very rare because I know what I'm doing, but most people don't. And they're just looking for a quick shortcut. So, yeah, another cautionary tale now that we'll probably be talking about for a while. All right. This one might sound kind of dorky in on the back end, and why are we talking about an API, but it's going to change, I think, creative work. So OpenAI has launched their API access for their popular chat GPT image generation model.

So OpenAI has made its advanced image generation technology, which was previously only available inside its chat GPT platform accessible to developers through its API. This move comes after the tool's viral success in massive user engagement with over 130 million users creating more than 700 million images in just the first week. So the image generation feature is powered by a multimodal AI model called GPT Image 1. I wish they named this differently. You know, like CEO Sam Altman said, you know, he put out something on Twitter. about this and he said image gen is available. And I read it as Imagine, as in Google's Imagine, you know, photo.

And I was like, wait, what happened here? We got to name these things better. Calling it GPT Image 1 and ImageGen, not the best way to refer to the models, just saying. But it is capable of producing images in multiple styles, following detailed instructions and rendering text within images. Developers can generate multiple images simultaneously and adjust the quality and speed of generation, offering flexibility based on application needs.

OpenAI has implemented the same safety measures in the API as in chat GBT, including content moderation options that can be set to auto or low sensitivity, allowing developers some control over filtering strictness. images, AI images created with GPT Image 1 include C2PA metadata watermarks to help identify them as AI generated. 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 don't know, y'all, I'm feeling spicy. Should we do, probably couldn't do this for a couple of weeks, but should we do a show where we build essentially an app on this OpenAI

ImageGen model? I never know if that's too dorky, right? Might have to use, you know, cursor or windsurf or something like that. But I think this is truly going to be, like, you know how we had this like chat GPT moment? It's not going to be as big as that, right? But this is going to be one of the biggest

steps forward for creativity that no one is going to notice, right? Because when a company releases an API, for the most part, it doesn't impact the everyday business leader. But y'all, I'm telling you, now there are going to be thousands in the coming months, thousands of new apps probably aimed at your business, at your vertical that's going to help you create more engaging in better images. And as we've already seen from all the, you know, and I'm not talking to the viral like Ghibli style or, you know,

make me a Barbie, right, action figure. Like, I'm not talking about that. I am talking about incredibly accurate visuals, creating infographics, creating realistic photos with text adherence. Like, it is really, really good. All right. A couple people said, said, yeah, we should do that.

McDonald said yes. Joe said, I would really love to see API image generation because I will use it. Derek just says, why not? All right, maybe we will. You know, because I've also always thought about, okay, should we do like a show on cursor or windsurf or something like that? So maybe it'll be a two or three part series and we'll tackle, you know, maybe, you know, coding inside, you know, Chad GVT and Gemini.

and then we'll tackle a cursor in windsurf, and then we'll do creating an app or something like that. So I don't know. Either we'll do it as a live show or maybe when and if we finally launched the community that's technically been ready for a year and a half. Maybe we'll do it inside of the community. So we'll see. All right.

Next piece of AI news. US President Trump has signed an executive order to expand AI education in the US. So President Trump has signed an executive order to boost AI education across the U.S. by launching a national initiative focused on preparing students for careers in AI, one of the fastest growing technology fields. So the order establishes the White House Task Force on AI Education, chaired by the director of the Office of Science and Technology Policy, and including senior officials from energy, agriculture, education, labor, and a special advisor on AI and cryptocurrency. So the initiative aims to provide early AI training in schools

to help students develop the skills and confidence needed for the AI-assisted workforce. Despite these goals, though, the administration has not yet clarified how we will implement the plan, especially following significant cuts to the Department of Education budget earlier this year, right? Yeah, so it's like, oh, yeah, we're going to do all these things through the US Department of Education, but it's like, wait, their budget is like essentially zero. And infamously, the head of the U.S. Department of Education repeatedly called AI artificial intelligence in a hearing, A1, like the steak sauce over and over. So I don't know.

Like on one side, I'm like, this is great, right? Yes, we absolutely, you know, need AI literacy in schools. But at the other hand, I'm like, okay, it sounds like a plan right now. and they're like, oh, the U.S. Department of Education is going to carry this out. But their budget has been absolutely thrashed. And I don't know, at least me personally, after listening to that hearing, where the head of the U.S. Department of Education continually called AIA1, like the steak sauce, I don't know.

I'm not super confident, although I am hopeful because we need this. So the task force will lead a presidential AI challenge designed to showcase achievements by students and educators, encouraging widespread AI adoption and promotes collaboration across government, academia, philanthropy, philanthropy, philanthropy, I can't speak today, and industry to solve national challenges using AI. The executive order also directs the Secretary of Education to prioritize AI in teacher training grants and instructs the National Science Foundation to focus research efforts on AI education. So let me just say this.

Finally. Finally. This is a much needed first step to officially push AI education in the school system. If I'm being honest, this should have been done at least one or two years ago. So you have to tip your hats to the Trump administration for pushing this initiative forward. How if it actually gets carried out, we'll see. I mean, like we've been talking about on this show, the U.S. government has been straight up slashed in terms of just the amount of humans working there, the amount of resources, right?

We've seen all these reports that Elon Musk's Doge department is, you know, essentially trying to replace workers with AI, using AI apparently to spy on people to see if they're talking badly about the administration. So I'm hopeful that this will work because the U.S. education system, let me say this right now, is in huge. trouble. It is in huge trouble because we've essentially been ignoring artificial intelligence for years, pretending it doesn't exist, banning it in classrooms across the U.S. when we should have been doing exactly this all along, a national effort to push AI literacy K through 12. Love it. I hope it happens. And we'll obviously be talking about this as we keep going. Yeah. A lot of a lot of people saying, Yeah, like Derek says, absolutely need it. You know, Joe said the biggest issue with world leaders mandating education is lack of ethical government governance.

Yeah, you can say that just about any government worldwide, but especially here in the UI. So Kyle says, I have been doing this kind of work for almost a year, a year and a half in Mississippi. And I'm working on a partnership with an AI education company to expand that. I love to see it. James here joining us from the Twitter machine says AI moves too fast for school classrooms. Kids should have free access to online school. Yeah, that's a great point, right? Yeah, how are you going to develop curriculum?

You know, anything, right? And this isn't a political statement. This is going back to for, I don't know, for forever. The government's slow. If AI is fast. So educating, right, pushing AI education into classrooms is extremely difficult, right? I know, I don't even know if I've talked about this. Maybe I have.

I started as an instructor at DePaul teaching generative AI. I update my syllabus multiple times a week, right? Because I want, you know, it's mid-career professionals taking this. It's not undergrad. So it's people, you know, in the middle of their career. But I want them to have the most up-to-date information. Like literally a couple of minutes before I start my course. I go through and update it because there's things, even if I finish, you know, kind of my,

my presentation or my course material, you know, the afternoon before, there's already things that have changed, right? So as an example, you know, our students are going to be using deep research tools. And, hey, last week, I'm like, hey, right now you don't have access to this. If you're on the free chat, chat, GPT plan, and well, now that's already changed. So it will be interesting to see how they update to the curriculum, but getting something is much better than nothing. All right. Like I talked about earlier, more Microsoft news. So Microsoft has finally, for real, for real this time, launched their recall feature, at least for those with

co-pilot plus PCs. So Microsoft is now rolling out its controversial recall, a feature that essentially screenshots almost everything users do on co-pilot plus PCs after nearly a year of delays and security improvements. This launch comes alongside some other AI upgrades to Windows search and a new click-to-do function. So Recall was initially planned to launch with Copilot Plus PCs in June of last year, but was delayed due to security concerns raised by researchers. So Microsoft spent 10 months enhancing recall security, making it an opt-in feature, whereas previously it was previewed as an opt-out.

So you were instantly opted in. That's not off everyone's alarm. So now it is an explicit opt-in feature with encrypted data storage and default filtering of sensitive information. So security expert that was quoted in one of these articles, Kevin Beaumount, who was the first one to flag recalls security risk, recently tested the final version and reportedly acknowledged that Microsoft's serious efforts to secure. have paid off, though some filtering glitches remain. So Recall allows users to search their PC by all these snapshots of all of their activity, helping find vague memories of content without needing exact file names, improving over traditional search.

This is literally the reason I bought a Microsoft Plus PC. Now I can finally set it up like five months after I bought it because I've been waiting for this feature. I don't care, right? I'm not big on data security, right? When I work with clients, I walk them through it. But I tell people, like any information that you have that is publicly available online, it's already in large language models, right? It's funny sometimes.

I've consulted, you know, companies that, you know, are publicly traded. And they're like, oh, I don't want to put this information online. I'm like, yo, it's on your 10K. It's already in these large language models, right? It's on offline models, right? So a lot of times companies don't understand everything they put. on the internet is already in large language models. Yes, there's obviously hordes of confidential sensitive information.

You may not want to, you know, have recall screenshotting, but that is why it is opt-in now, and it has some new additional features to not collect some of that more sensitive information. The feature does as well require a four-digit pin via Windows Hello for access, despite Microsoft's claim that Biometric sign in would be mandatory, so that could still be coming. Also, alongside Recall, Windows Search now supports natural language queries on Copilot Plus PCs. That's amazing. Apple's search is absolutely hot, flaming garbage, right?

I hate Apple search. I can't find anything on my computer. It's like you literally have to have the exact same file name for it to ever come up. So I love now that you can just search with natural language, right? So you could say, and I did when I was at Microsoft's conference back in November, I saw some live demos of this and I talked to the team that was building it. So as an example, let's say you have a huge PDF and there's a picture of a tree on the inside. And you're like, oh, what is that PDF?

I don't remember what it was called. I don't even remember the title of it, but I remember there's this big picture of a tree on the inside. You can literally type in tree and it's going to pop up. That is what kind of semantic search is. Extremely powerful for productivity, right? I think that this is a small thing. You know, everyone's talking about recall because it was controversial. But I love the new, or at least the idea and the concept of this new natural language query on copilot plus PCs.

Also, there is the new click to do feature, which is activated by a Windows key plus a left mouse click, which offers quick actions like summarizing text or removing objects from images on screen. Very cool. Right. So that's essentially bringing in some of the best kind of time savers from different large language models that you can just do, right? If you have something open, you can click it and instantly summarize it. You don't have to have a Chrome extension. You don't have to have a desktop, you know, AI system running. It's all just built in on this edge AI if you have a co-pilot plus PC.

All right. Our LinkedIn user said truth, Apple. hate Apple search. It's so bad. Why is it so so bad? Why is Apple so far behind? Fred says Jordan, it's time to take the dust off your new Windows PC. Oh man. My wife is always like, are you going to set this up now? And I'm always like, yes, yes. I absolutely am. And then I don't know. Something big happens in AI and I'm like, all right, well, now I've got to spend eight hours, you know, testing out this new 03 or testing out this new Gemini 2.5. So if the AI news world went like a full week without these huge announcements, I would have time to set properly set up my new Windows

Copilot Plus PC. All right. Someone is saying, please share, please share it when you set it up. I want to see a live demo. All right. I'll see, I'll see Microsoft if we can, you know, set, get, get something special set up for you guys. All right.

All right. Our next piece of AI news. What's so strange is, I called this months before it happens. So Perplexity AI has launched a voice assistant on iOS, and they are straight up going after Siri. Speaking of Apple, doing a bad job on AI, perplexity, hard pivot coming for them. So Perplexity has launched a new updates to their iOS that features a new voice assistant available now on iOS, and it offers features that even currently, Siri struggles with, such as creating calendar events, reminders, playing music, and holding natural

conversations. So unlike Siri, which has delayed its kind of quote unquote smarter Siri until potentially the end of next year or even into 2027, perplexity's voice assistant integrates directly with Apple's default apps like calendars, reminders, reminders, mail, and contacts, allowing it to perform tasks like drafting emails and adding calendar events more reliably. So perplexity's assistants can maintain context and conversations without needing frequent follow-up questions, unlike Siri, providing concise responses and prompting users for further inquiries. The voice mode in perplexity runs in the background, allowing users to switch apps without interrupting the conversation and offers customizable settings such as muting the microphone,

disabling transcripts and changing the assistant's voice. Plexity also supports web-based actions by preparing tasks like booking restaurants, restaurant tables or opening Uber with a destination pre-filled, though it still requires users to manually complete the final steps themselves. Features Siri aims to offer, but won't until minimum next year. One thing that does work pretty consistently, perplexity can launch in play music, both from Apple Music and YouTube videos quickly, although it has current limitations, such as difficulty sending text messages,

even with contact access. So it says it can't. So let me just say this. It sounds great, right? Like, oh my gosh, we finally have a smart Siri. Doesn't work. At least not consistently, right? So let me tackle this.

two parts. I did call this. Just saying, I called this back in our 2025 AI roadmap and prediction series. I said perplexity is either going to get squashed or they're going to have to hard pivot. And I would say, if perplexity can get this right, this is their hard pivot. And I'm rooting for them. I've been extremely critical of perplexity over the last, I don't know, nine months, because they have a hallucination problem. It's pretty bad. ad, right? And their biggest USP is gone, right? Their biggest, uh, kind of advantage, uh, a year ago, 18 months ago was, was the ability to, uh, kind of quickly go out and, you know, go over 20 websites and to give you a somewhat accurate, uh, you know, bullet point or

summarization of a certain topic. It's not good anymore. It's not. It's not even a top three at that exact task. So let me say that again. A year ago, perplexity was on the top in terms of being able to do this deep research, right? So number one, they technically created this category that now everyone else is blazing, but they're not, they don't have a top three product in that same category that they essentially helped create. So like I said, you have open AI and Google in the top two spots. We'll see which one is actually ahead. Then I actually think you have Grock. Right. And as problematic grok can be, I still think it is better than perplexity. So I've been saying all along, perplexity is going to get their lunch money taken from them in this category that they helped create.

So they have to pivot or they're going to get crushed. So I said they had to pivot. I said they had to innovate. So here we see it. And I love it. I love this move by perplexity. And like I said, I've been pretty harsh on perplexity because I think they've lost sight of what made them originally special and originally you know. because hallucinations have been taking over their platform.

And I think, like I said, they've been beat to the punch by this whole deep research thing. But no one likes Apple's AI because it's not even there. Series not smart. So I'm rooting for them. I hope this works. But I tried it for when it first came out. I tried it for five or 10 minutes. I had it work once where it would check my calendar and then it drafted an email like,

oh, you know, hey, what's on my calendar? I said, oh, you have a meeting at 1 p.m. with Bill. And I'm like, okay, email Bill and say, I have to push the meeting back. And it did that. It didn't once. Guess what? I've tried it five times since, and it hasn't worked. So, you know, sometimes it'll be like, oh, I can't access your calendar. And then you do it three more times than it can't. Right. Then you're like, okay, you know, send an email to this person. You know, that's all the calendar invite. And then they're like, oh, I don't have their contact information. It's like, yeah, you definitely do. And then I go through it. I manually check my phone. I'm like, yes, it's there. So it's inconsistent,

but it's extremely promising. So I'm rooting for perplexity to improve this. And I did reach out to the perplexity team, at least on social media, with some feedback. It seems like they took it in stride. So hopefully they'll improve it. All right. Our last piece of AI news, Open AIs rumored open model may be coming in June. So Open AI could release its first, quote unquote, open language model since,

GPT2, right? Yeah. Targeting an early summer launch. So according to reports from TechCrunch, OpenAI is developing a new open reasoning language model. That's another big piece that we hadn't had confirmed yet. So it will be a reasoner model. And it could come in early summer, aiming to compete with other open models, such as

Meta's Lama and Deepseek by topping benchmark tests. So the model, however, will be text in and text out, so not multimodal by default, and designed to run on high-end consumer hardware with an option for developers to toggle its reasoning ability on or off, similar to recent models from Anthropic. OpenAI plans to use a highly permissive license with minimal restrictions on usage and commercialization, avoiding criticisms faced by models like MetaSys. Lama and Google's Gemma and Deepseek. Yeah, so Deep Seek and Lama are not true open source.

People don't understand that. And it doesn't look like this model from Open AI will be either, but it could be open weights. So a big difference. So Mattas Lama models, which are probably the most popular open-esque models, have already reportedly amassed downloads, right? And I read about this. I read about the official report. And then I also saw some discussion on Twitter that said, okay, this has to be an error, you know, because other people were sharing, hey, I have a model, you know, that I forked. And it says it has way more downloads than it actually does.

But if you do look at meta's, you know, public stats, it does say more than a billion downloads. So we'll see. This could be extremely popular. CEO, Sam Altman has acknowledged that OpenAI needs a new open source. strategy to compete with the deepseeks, Google's, Microsoft is creating some, some open models. Meta has been a leader as well. But Open AI, the open quote unquote, kind of AI lab hasn't released anything remotely open since the chat GPT craze.

So since GPT2 and even that model when they released it, it was very, very slim down from what it was originally announced. So like I said, CEO Sam Altman has acknowledged that OpenAI needs a new open source strategy and plans thorough safety testing and red teaming of the new model before its release, including publishing a detailed model card and benchmarking and safety results. So if successful, OpenAI may follow with additional models, including smaller versions, potentially broadening access for developers and companies. regardless, the rest of the world wins, right?

When the leader, and it's not even close, it's not even close. Open AI is the leader in users, right? Because aside from what we told you, right, that came out through that Google lawsuit, so Google is making a heavy push to acquire more users. So right now, reportedly, according to their court documents, 350 million monthly users where right now chat gvety is at 600 million monthly and there's been rumors that they're even more for 600 billion or sorry one billion weekly users uh whereas right now uh chat gbd has 600 million monthly so yeah the rumors are saying open a i since their new

ImageGen that their stats are actually way higher. So we'll see in the end if OpenAI does release an open source model. It changes not just all open source AI, but it also changes proprietary AI as well. Because what happens then is, you know, the labs that are building closed source models can essentially learn from what Open AI is doing in their open, even if it is just an open weights model, right? you can essentially learn and kind of mimic and emulate what all these other labs are doing. So, you know, that's why, you know, everything from DeepSeek to Microsoft to Google to meta, right, releasing very impressive open models. It actually helps proprietaries and it helps the whole system move forward. So Canal asked pro.

All right. Bro, let me tell you. He says, bro, what do you think about chatbot's capability after five years and ten years? Oh, my gosh. Good. Great question. I have no clue. I have no clue, right?

What we're getting right now specifically with 03, and if you didn't watch our 03 episode, make sure you go do it. So that was episodes 509 and 510. What we have now with 03 is, agentic. It's 100% agentic. There's still some hallucination problems. You actually get like the more agentic in multi-tiered these models get, the more that you have to really keep an eye on them because there's been some pretty wild hallucination problems. I would say so far with 03, whereas I wasn't getting them as easily with 4-0. So it does require

a little bit more prompt engineering. But where we're going to be with AI chatbots in five to 10 years? I don't know, right? I do know that I don't think we're going to be typing to them, right? I see them being on physical devices, right? I think whatever company, whether it's meta, Google or, you know, Open AI is reportedly working on hardware. I do see it being in glasses, you know, in our phones, right? I think finally that technology, I think that, you know, you could say if the most powerful

model, whether you want to say Google Gemini 2.5 Pro, 03 from OpenAI, We're going to have models way more powerful than that on edge, right? On our offline devices, we'll be able to download those models. So what we're able to do is I don't even know. Scary. I'll say that. All right. So that's it for our AI news.

But let me know which of these should we cover more this week. So like I said, we do have our show tomorrow. So make sure you tune in for Google's AI Studio. five time-consuming tasks you didn't know you could automate. That's going to be a fun one. I'm going to show you some ways that I'm using Google's AI studio every day. Let me just say, I feel like I'm stealing when I use Google's AI Studio because the long context window is extremely powerful, extremely flexible.

If you haven't used Google AI studio before, I suggest you tune in. All right, but let's very quickly recap the AI news that matters for April 28th. So Open AI introduce a lightweight deep research model, including access for free users and upping limits for paid users. Google has faced some potential breakup as the U.S. Department of Justice pushes to force the sale of the Chrome browser, which is really going to impact the AI industry. Microsoft has enhanced their Microsoft 365 co-pilot with powerful AI features in their Wave 2 spring release. lease, Adobe Firefly has expanded its Firefly AI system with support for competitors, namely Open AI and Google's image and video models. The Mind Pillow CEO is facing some legal trouble after his lawyer reportedly submitted

more than 30 fabricated legal citations to a judge. Ouch. Open AI has launched API access for its popular chat chbt image generation model, which I think is going to completely change the future of creative work. US President Trump has signed an executive order to expand AI education with a new national initiative and task force. Really hope that that changes the education scene here in the U.S. Microsoft has launched their recall feature finally after almost a year of delay for co-pilot plus PCs after an extensive security. security overhaul. Perplexity with a hard pivot going after Siri with its new AI voice assistant on iOS on Apple devices in the app store. And then last but not least, some new updates on

open AI set to release its first open language model since GPT2 targeting in early summer launch according to reports. All right, I hope this was helpful, y'all. If it was, please join us tomorrow for that Google AI studio show. I think it'll be a good one. But also, if you're listening on the podcast, appreciate your support. Please subscribe to the show. Leave us a rating if you could.

It takes like 30 seconds. So leave us a rating. Leave us a review. I'd appreciate that. If you're listening here on LinkedIn, Twitter, wherever, repost this. Click that repose button. We spend literally countless hours a week to help you cut through the AI noise

to bring you unbiased facts. with receipts. We do so much work on this. Takes you a couple of seconds. Click that repose button. Share it with your friend. Share it in the presentation. Email it to your colleagues. We'd appreciate that. I'd also appreciate you going to your everyday AI.com.

Signing up for the free daily newsletter. We're going to be recapping today's show in the newsletter. Thanks for tuning in. We'll see you back tomorrow. And every dang 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 a rating. It helps keep us going. For a little more AI magic, visit Your EverydayAI.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.