Everyday AI Podcast – An AI and ChatGPT Podcast - EP 279: Google’s New AI Updates from I/O: the good, the bad, and the WTF

Episode Date: May 23, 2024

Did Google say 'AI' too many times at their I/O conference? But real talk – it's hard to make sense of all of Google's announcements. With so many new products, updated functionality, and new LLM capabilities, how can you make sense of it all?  Oh.... that's what we're for. Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Ask Jordan questions on Google AIRelated Episode:  Ep 204: Google Gemini Advanced – 7 things you need to knowUpcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:1. Google's Updates and Announcements2. Google AI Evaluations3. Concerns Over Google's AI Development and MarketingTimestamps:01:30 Daily AI news05:30 What was announced at Google's I/O09:31 Microsoft and Google introduce AI for teams.12:38 AI features not available for paid accounts.16:12 Doubt Google's claims about their Gemini model.19:26 Speaker live-drew with Pixel phone, discussed code.22:24 Exciting city scene, impressive Vio and Astra.24:44 Gems and GPTs changing interactions with language models.29:26 Accessing advanced features requires technical know-how.33:45 Concerns about availability and timing of Google's features.36:14 Google CEO makes joke about overusing buzzwords.41:06 Google Gems: A needed improvement for Google Gemini.41:50 GPT 4 ranks 4th behind Windows Copilot.Keywords:Google IO conference, AI updates, NVIDIA revenue growth, Meta acquisition, Adapt AI startup, OpenAI deal, News Corp, Project Astra, Gemini AI agent, Gemini 1.5 pro, Ask photos powered by Gemini, Gemini Nano, Android 15, GEMS, Google AI teammate, Microsoft team copilot, Google Workspace, Google search, Veo, Imagine 3, Lyria, Google's AI music generator, Wyclef Jean, Large language models, GPT 4.0, Google Gemini, AI marketing tactics, Deceptive marketing, Discoverability issues, Branding issues.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. Google just released more AI updates than I can even count, right?

So at their I.O. conference about now a week ago, Google announced literally AI everywhere in almost every single product and service that they offer. So today we're going to be doing a little bit of a deeper dive and talk about what's good, what's bad, and what just has us scratching our head and saying, saying WTF. All right. So we're going to be going over in depth a little bit more today on the Google IO conference and talking about what it all means and a little bit more on everyday AI. So what's going on y'all? My name is Jordan Wilson and I'm the host and this show. It's for you, right? So this is a daily live stream podcast and free daily newsletter helping people like you and me, not just learn

generative AI, but how we can leverage it to grow our companies and to grow our careers. So if that sounds like you, first of all, thank you so much for joining us. You know, always a appreciate hearing from you all. So if you are new here, yeah, we do this live every single weekday, Monday through Friday, covering the latest in AI news. And speaking of the latest in AI news, you can always catch that in our newsletter as well. So make sure you go to your everyday AI.com. Sign up for that free daily newsletter. All right, let's start as we do every day going over what's happening in the world of AI news. So Invidia showed some staggering growth on its earnings call. So, Nvidia reported their net income in the first quarter of 2024 rose over sevenfold compared to a year earlier, reaching 14.8 billion with revenue more than tripling to 26 billion.

Yeah, like sevenfold. That's nuts. So, Nvidia CEO Jensen Wong foresees a new era of AI factories powered by Nvidia chips designed to accelerate the production of artificial intelligence. So, Nvidia's early investment into AI technology has propelled its hardware and software. in AI applications, gaming, automotive, all over the place, right? And as we talk about here on the show a lot, demand for NVIDIA's specialized GPU chips that power generative AI has soared due to the need for generative AI systems among tech giants like Amazon, Google, meta, Microsoft, etc.

Also, with NVIDIA's chips making such a huge profit, they are now planning to release new chips every year instead of every two years. So updating their production cycle, pretty big news there. All right. Next, what's going on in the world is meta may be looking to acquire an AI agent startup, according to reporting from the information. So ADAPT, a two-year-old AI startup founded by X OpenAI and Google AI developers is considering a sale or strategic partnership with tech giants, maybe meta. So investors have valued ADAPT at over $1 billion last year, reflecting the high stakes in the AI startup landscape. So the heavy costs associated with training and maintaining AI models pose challenges for startups like Adept, even with significant initial funding. And obviously, competition in the AI agent field is intensifying like crazy.

Even Google and Microsoft have announced that in the last week with these established players, you know, looking to get involved. All right. Last but definitely not least, Open AI has struck a $250 million deal with a $1,000. with the News Corp for AI training. So OpenAI has signed a deal with News Corp worth a quarter billion dollars. Yes, a $250 million licensing deal that lasts for the next five years. The agreement grants OpenAI access to current and archived articles from News Corpublications for AI training and user question answering.

So talked about this a lot on the show, but major companies like already the Associated Press, Financial Times, and political owner Axel Springer have already partnered. nerd in content deals with Open AI. And on the flip side, other major outlets, such as the New York Times and my newspaper up the street here, the Chicago Tribune, have filed lawsuits against Open AI and Microsoft for alleged copyright infringement. So the News Corp will provide journalistic expertise to Open AI to uphold journalism standards. Also, so this partnership, if you don't know about News Corp, that includes outlets such as

Barron's MarketWatch, Investors, Business Daily. FN, the Sunday Times, the Sun, the Australian, a lot of others. So pretty interesting. And, you know, I said this about six or seven years ago. I said the only way this, or six or seven months ago, I said, the only way this shakes out is a ton of lawsuits and a ton of huge partnership deals because, hey, let's face it. The reality is large language models are largely built on copyrighted content. So, you know, more on that later.

But let's talk about what we came here for today is to go over Google's new AI updates from their IO conference. last week, go over the good, the bad, and the what the freak is this. All right. So thanks to our live audience tuning in, as always. Peter joining us from Belgium, Tara from Nashville, someone joining us on LinkedIn here from Florida. Woozy joining us from Kansas City.

Thank you all for joining. I would love to hear your questions and comments on what you thought of Google's announcements. So let's just start with a recap, right? So if you didn't catch our newsletter last week where we recapped it all, let's just go over what was kind of announced because, yeah, it was a lot. Like I started at the top of the show, it was maybe too much AI, right? And that's a lot for me to say. I like AI everywhere, but I'm like, yo, this is, is this too much? All right.

So a couple of what I think are going to be the highlights and some of the AI initiatives from Google impacting everyday people like you and me. So Google unveiled Project Astra, which. is an impressive new kind of AI agent of sorts powered by Gemini. This is very similar to what OpenAI announced in their GPT40 and kind of the ability to see here react and interface with humans in real time in the app. So we're going to show an example of that here in a minute. Google also announced enhancements to Gemini models, including Gemini 1.5 Pro with between 1 and 2 million.

tokens of context, right? One thing Google is absolutely crushing right now is the context window. Also, they announced Ask Photos powered by Gemini. It was revealed for Google Photos, providing enhanced photo memory summaries, which I thought was a pretty cool piece there. So if you use Google Photos, I do. Once this is released, you'll essentially just be able to talk or ask Google Photos to, you know, like let's say, hey, I want to see, you know, my pictures of my cat, you know,

growing up over the years, right? And once it identifies, hey, this is your cat, it's going to show you. And then you can ask questions, right? Like, where has my, you know, what toys has my cat played with over the years or whatever you might use it for, right? That's a fun example. But imagine what that could do for, you know, even business, right? Like your screenshots, you know, if you're taking pictures, you know, out on a construction site or something like that, pretty big. So I do think Ask Photos will be a pretty popular feature. So Google announced the rollout of Gemma. I have multi-modality on pixel phones. So yeah, a lot of this, you know,

geared toward edge AI on Samsung devices and pixel phones. Android 15, another huge one, was presented with AI powered search with Gemini as the new AI assistant and on-device AI for new experiences. Yes. So that cannot be overlooked. So, you know, everything that we're kind of talking about with, you know,

Apple working on in the future, well, Google already just announced it, right? I would say that at least for smartphone dominance. I do believe that's Apple, right? Or maybe, I don't know, maybe I live in a bubble here in the U.S. living in a big city like Chicago. But, you know, I don't really know anyone personally that has an Android phone. I'm trying to think maybe like a, you know, a cousin or something like that.

But basically everyone I know has an iPhone, which is why people are looking, you know, at WDC in June for Apple to see what they announced. But hey, Google just did it, right? So Google just announced Edge AI. It's already been rolled out in a phone. I believe it was the S-24 a couple of months ago, but, you know, talking about Android 15 here, bringing it to the operating system. Google also announced gems, which are customized versions of Gemini,

its large language model for, you know, specific tasks. So more or less, this is GPTs. This is the version of, you know, GPTs that the world has already had access to inside of OpenAI's chat GBT, essentially creating a, you know, version of Google Gemini with some specific features, you know, specific requests, etc. So calling that gems. So that I think is going to be pretty big. Also, Google demonstrated Google AI teammate, a virtual teammate for work tasks. So we saw something similar with a team co-pilot from Microsoft this week, you know,

So Microsoft just had their build conference, which is actually wrapping up today, I think, is the last day. So Microsoft announced something similar. But we really are moving into this space now where you are literally assigning an AI, right, like to your teams. Like it has a license, you know, you grant it access to certain software. Think of it like, you know, if you've ever managed, you know, network access or something like that. You know, it's literally giving an AI a seat on your team and giving it access, you know, to, you know, depending on if you're using, you know,

Google Workspace Suites of Product, you know, obviously now you have that with, you know, this new kind of virtual teammate, you know, within Google AI teammate, similarly for TeamCopilot that Microsoft just announced as well. Also, updates to Google Workspace were highlighted integrating Gemini to enhance efficiency at work. Yeah, just about every Google Workspace product, they announced some new AI feature, some Gemini,

Don't have time to go over them all. Google has like dozens of products and they rolled it out. They rolled, you know, AI integration or Gemini integration into just about everything. Also, new features for Google search were introduced leveraging generative AI for what they're calling organized search results. And that's actually gotten some mixed, some kind of mixed reaction so far, which has been kind of interesting. And we'll share about that in today's newsletter as well.

Two other things worth noting, I think Google revealed VO, a text to video, and imagine three in AI image generator. So VO will in theory be a competitor to SORA. And I do think the quality we saw from VO was pretty good, right? Probably better than what we've seen from companies like runway, PICA Labs. But I would say still very far behind Open AIs SORA. But still, this was kind of a surprise to most. I would say the, you know, text to video, you know, there wasn't a lot of reporting out there.

on this. Obviously, we knew that, you know, Google already had their image generator. So whether you want to call that a, you know, Dali 3 competitor or mid-jorney competitor, you could stay, you know, you could say stability competitor, even though they may, you know, get bought out. But still, so imagine three, some, some updated image capabilities, text to image within Google Gemini and VO. VEO is not released yet. And then Leria. So last but not, least on our on our update list, Liria, which is Google's AI music generator, was demonstrated with musician Wycleft John. So yeah, and that's just some of the highlights that we picked, right? There are literally, you know, I went through, I watched, watched the presentation,

had dozens of bullet points of like main announcements. So that's just the ones that I think ultimately are going to be affecting a lot of everyday people, right? So if your company, especially, right, if your company, you, uses or if you personally use Google, you know, Google's suite of products in workspace, you know, that's what our company, you know, my, my company's Accelerant Agency and everyday AI right now we use Google workspace, right? So some of these features have already been rolled out. Some of them for whatever reason are available on, you know, my free Gmail, you know, my personal Gmail, but not my work accounts, even though we pay extra, right? So not only

do we pay monthly for a seat on those, but we also pay extra for all these AI features, and they still haven't rolled them out. So more on that here in a bit. But I'm curious from our, you know, live stream audience. What are your thoughts so far on the Google Gemini or sorry, the Google's I.O announcements, literally sprinkling AI everywhere. I'm going to go over now what I think is the good, the bad, and the WTF. But, you know, I'm curious from you all or if you have questions, get them in to our

podcast audience. We always leave them the show notes, ways that you can reach out, send us an email. You can just actually send us a text message now straight from the show notes. So if you are listening on the podcast, we'd love to hear from you as well. And we'll probably feature some of our favorite comments or feedback in today's newsletter. So let's just go ahead and let's talk a little bit about the good, the bad, and the what the frick. All right. So the good, Project Astra, right? So Project Astra, I'd say was the headlining feature because it was something that I think at the time was competing directly with Open AI, right? So Open AI, obviously, and we cover this, well, maybe not obviously, let me tell you. So, you know, essentially Google had their I-O date set three months prior.

Three days prior, Open AI came and sprung a quote-unquote surprise announcement, even though it was kind of report. the day before Google, right? And one of the big things that OpenAI release in its new GPT4-O model is what some people are calling, you know, her. There is no name for it. We're calling it kind of Omni Live. But, you know, that's the GPD-40 is for Omni. So we call it Omni Live.

They didn't give it a name. You know, OpenAI didn't. But essentially, it's the same thing here as Project Astra. So this is a new way to interact with generative AI, with large language models. And essentially, it has the ability, Project Astra, to see things in real time. And to also, you can interact with it in real time, ask it questions. Essentially think of how you FaceTime someone. Imagine FaceTiming an expert via a large language model.

It can see what you see in real times, in real time, identify objects. You can talk to it. It can hear you. It can process, tap into the large language model and spit back answers. So think of no matter, you know, there's so many applications for a project Astra type product or, you know, an Omni live product in our everyday lives. But I think maybe it'll make sense once we play a little bit of a clip here. So I do, hey, and people might roll their eyes at me. I do have to preface this, right, which I feel kind of bad doing because I'm not ragging on Google.

However, we did see in December when Google, announced Gemini, they released a marketing video, and it turns out that most of it was a fit. Most of it was made up. Most of it was overly produced. And Google, I mean, yeah, they lied, right? I can say that because just about every news organization literally said Google lied, Google, you know, deceptive marketing, whatever you want to say. So I do have to say that out loud.

And it feels like I feel bad saying it because, you know, I don't want to come off too biased. here on the everyday AI show, but, you know, ever since that happened, right, you have to take everything, I'm sorry, you have to take everything that Google does with a grain of salt, right? So if you miss that whole situation, you know, when Google announced their Gemini model, they showed this marketing video that essentially was Project Astra. It showed that Google Gemini had the ability to kind of see and interact with it in real time, which it didn't. And then later they kind of released a paper that said, oh, actually.

it wasn't live. We just took all these screenshots and then humans did a bunch of text prompting based on these screenshots over and over and over to produce the result. And then, you know, we essentially did some nice marketing and made it kind of look real time, right, which it wasn't. So I have to say that because, you know, again, Google says that this is, you know, all, you know, one X unedited, but yet you have to preface this, right? So I don't 100% trust Google. You know, I don't know if I'm alone in that. Maybe someone out, someone else out there doesn't. But still, I have to put that out there before I show, show the live stream audience,

this and also on the podcast. So I'm going to just preface this here real quick. I'm not going to play the entire piece here. All right. So here's what we have. This is, I'm going to play just about maybe 45 seconds of it. So this is showing this project Astra. So like I said, the ability for Google Gemini to essentially see, to hear,

and for you to interact with it in real time. So they said this is in real time. So someone is using Project Astra, and let's just go ahead. Take a watch, take a listen. We're going to let it go a little bit here. Okay. Let's do some tests. Tell me when you see something that makes sound.

I see a speaker, which makes sound. What is that part of the speaker called? That is the tweeter. It produces high frequency sounds. Give me a creative alliteration about these. Creative crayon. colors cheerfully. They certainly craft colorful creations. What does that part of the code do? This code defines encryption and decryption functions. It seems to use AESCBC encryption to encroption to

encode data based on a key and an initialization vector, IV. 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, Premier, 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.adopo.com. So hopefully, let me know live stream audience. Hopefully y'all could hear that audio. And you weren't just sitting there in silence for 40 seconds. So let me just kind of explain to our podcast audience what happened there.

So it started with someone going up to a speaker, right? And then they on the screen, which I think this is a great feature. And if this works, this is pretty amazing. And we didn't see this out of Open AI's announcement. But it essentially, it looks like there is the ability to kind of draw live. So, you know, there was this speaker, and she kind of live drew with with her finger on the pixel phone an arrow and pointed to a certain part of that speaker while it was still live in real time, which I think is really cool. Next, she didn't say this, but the thing she was asking for creative, like alliteration was a jar full of crayons, right? And then it said, creative crayons color cheerfully.

They certainly cracked colorful creations, right? So showing it a jar of crayons and, you know, responding reportedly in real time. And then last but not least, it looks like one of her coworkers was at the desk there writing some code. And then that's what she said, you know, what does this part of the code mean? And then Gemini via Project Astra said this code defines encryption and decryption functions, et cetera. So there's a couple other kind of examples of Project Astra that just actually dropped about a day ago. So really all we had was one or two examples early on, but a couple more have dropped. And we'll share those in the newsletter today as well.

All right. So I'll say the good right away. Project Astra looks pretty amazing, right? Again, if this is actually what is going to be rolled out, you know, this looks like it does have some real-time capabilities that OpenAIs, GPT40, Live Omni, might not have out of the gate. So pretty impressive. Something else that I think was good. Vio, yeah.

So let's take a look at this. So similarly, I'm just going to play just a portion. So again, Google says this was all created, you know, with VO, all stitched together. So let's just take a quick look. I don't even know if this has sound. It doesn't matter. It's just visuals. So I'm going to try to do my best to kind of explain what's going on here.

All right. So let's go ahead and click play. I forgot if this even has sound. All right. It doesn't have sound. So I'll kind of narrate it. So we have kind of what looks like a drone shot over like a neon type city. The visuals look pretty crisp.

This shot looks very smooth. All right. So now it's speeding up. And it is almost like you're kind of flying through a city that goes straight into this, you know, futuristic kind of tron-looking, you know, race scene, you know, again, in a futuristic city. That seems to be the theme here. A lot of neon lights, futuristic city. But again, this video presumably was all from text prompts.

So it almost looks like a car race scene. You can obviously tell this is not quote unquote, you know, real life. But it does look like extremely high quality, right? Like whether that was CGI or animation, it looks like something that would take a lot of money and a lot of skill to build. here. So again, we're just kind of going through this race car scene. And now it does look a little more real life. So now it is a driving through a city in the daytime where the first two scenes of this looked more like CGI, things like that. And then obviously they share the prompt here, the text prompt that they said created this, which was, you know, a fast track, a fast tracking shot through a

bustling dystopian sprawl with bright neon signs, flying cars and mist, blah, blah, blah. So that's all on the screen there and we'll make sure to share the link to this in the newsletter as well. So VO, I think, was pretty impressive. So, you know, I'm curious for our live stream audience. Were you impressed with VO and with Project Astra? I personally was, right? Again, I have to take things with a grain of salt with Google. But it seemed like there was a lot of hype, right?

Especially with Open AI Sora. It kind of shook, if I'm being. honest, right? It shook the internet, right? I mean, it was on cable TV, right? It, Hollywood was reacting to it. You know, there's the story where, you know, Tyler Perry had, you know, eight or nine figure expansion to his studio that he reportedly put on hold because of after he saw SORA, right? So I, I do think that there was a lot of eyes and ears and in news coverage, even on Open AI SORA. So I don't think that VO is really as good quality-wise. It doesn't look like it, but it's still pretty good, right?

I would say, you know, again, SORA is in its own category, but it looks like right now, at least what Google decided to show, was pretty well ahead of every other, you know, at least publicly available text to video or photo to video product out there, such as runway and Pika Labs. So, you know, this VO product actually looks pretty good. So I'm surprised. Again, Google just dumped so much.

So I think so many things are being overlooked here. Right. So it's, I think that's important to point out is, you know, there was, there is and reasonably right. And I think rightfully so a lot of hype and a lot of attention on SORA. But Vio literally just flew under the radar like that futuristic car in that scene. Another good thing that I like is gems. So I think gems in the same way that I think that custom GPTs can kind of change how people interact with large language models.

Similarly, I think Google gems can do that as well. Again, all we got was kind of a little demo video and a bunch of marketing. But again, I think this is the future of how people interact with large language models. If I'm being honest, and this is another show for another day, I think the future is actually small language models and working with many of them that are highly tuned for specific purposes. However, I think in the interim, you know, something like GPTs or GEMs is how people are ultimately going to get the most

out of large language models, right? So if you don't know, you know, GPTs and presumably gems, it kind of looks like, you know, Google's gems are just GPTs, which have already been announced and they've been out for, you know, six months. So here's one area that it looks like Google was actually playing catch-up on compared to Open AI. But essentially, it allows you to kind of, it's not custom instructions necessarily, but it allows you to configure a customized version of the model. So for different tasks. So let's say you wanted something that was a creative copywriter. You know, if you start a new chat, you know,

in Gemini, you might have to kind of quote unquote coach it to get it there. And then maybe you want something that helps you summarize long PDFs in a very formal tone in bullet points, right, like as an example. So normally, you would have to start a new chat and kind of go through a process to get them there. And instead, this is a way to customize Google Gemini and kind of save it. I do believe also you can give it access to certain files and information in your Google Drive. So also, you know, I call that mini-rag, right, retrieval.

augmented generation. It's not true rag, but I think when you can upload, you know, files, it is kind of this, you know, miniature rag, how you can bring in your company's data or, you know, personal data that you want to work with into the model so you can improve, you know, the capabilities and the output out of the gates. So I do think gems, another huge announcement that I think I'm pretty impressed by. Again, if it all comes to fruition as it was marketed, which is another story with Google sometimes. All right. So let's talk the bad. Yeah.

And hey, I agree. I agree with Adam here joining us from YouTube. It would be cool if we knew that it was real, real Google fakes everything. Yeah. I will say they're very good at marketing, right? Google is a marketing company that also has other products, right? Like, I know that sounds wild, but at least when it comes to their AI, that's what I think it is. It's about marketing, branding.

etc. Tara, Tara says clear and never a disappointment, great as usual. All right. So let's go talk about the bad. I think one of the bad things here is access, right? You know, as an example, all these Gemini updates stole the show, you know, especially with its long context window. But here's the thing. Google does not make it easy for the average person to get access to all of this, right? And also, Google is very, right? So if you're just using the front end of Google Gemini, right? So you just log on to Gemini and you're just using it like you are, Chat, GBT. Google is not very clear about what capabilities of Gemini you're actually getting. Another thing is they were, especially when it first came out, they were changing the name

constantly. It was it was Gemini Advance and, you know, then it was, I think, Gemini Pro. And then they started using Gemini Ultra versus Pro. And then they took that away and went back to advance, right? Like from a branding and even knowing what you're getting access to from the front end of Gemini, Google does not do a good job, if I'm being honest, right? I think OpenAI and Claude, you know, do a much better job. They say, oh, here is GBT4. You are selecting the model. here are G, you know, or GPD40, here are the capabilities, right? Within Anthropics, Claude, you go in to the front end and, you know, you're like, okay, I'm using Claude 3 sonnet.

I'm using Claude three haiku. I'm using Claude three opus. And you know what you're getting. That's the thing. The access to all of this, you know, Google makes a big deal and they just say, oh, you know, Gemini, you know, 1.5 pro, 1.5 ultra, whatever, right? And one million tokens, two million tokens. That's huge, right?

But here's the thing. You don't have access to that. The everyday person doesn't. You got to be not a little bit of a dork, but you got to be a little bit of a dork to access all of that. Right. So, you know, sharing this here on the screen for my live stream audience. But if you want access and, you know, they even kind of announced a newer, faster,

lightweight model called Gemini 1.5 Flash, right, which is supposed to be lighter weight, faster than the kind of 1.5 Ultra or 1.5 Pro, whatever, they're calling it today. But you have to actually go into Google AI Studio and you have to create a new Google Cloud project to create an API key in its pay as you go, right? So I do understand that, right, because they probably can't just release something with one million tokens or two million tokens into the kind of consumer or the front end version of Google Gemini, because, yeah, from a billing standpoint, that's expensive, right? And if you don't know anything about tokens, first of all, the fact that Google is even in its

AI studio is offering a million tokens, very impressive. It is. So, you know, as an example, let me just very quickly talk about, you know, what that means. Because it is, it is big. It's, it's impressive. But essentially, you know, if we, compare it to chat dbt. So right now on the front end of chat chbt, you have a 32,000 token memory. So what that means, that means after about 26,000 words or so, chat gbt is going to start forgetting things, right? So it doesn't have a very good memory.

But this does, right? So Google Gemini in the AI studio does. A one million token memory is huge. Also, another great thing and something aside from that memory, that context window that Google is leading the pack in. Another thing that they're leading the pack in is multimodal. So you can input video. And I did some testing on it last night.

So you can upload videos and ask questions of your videos, which is amazing. And that's the future. That is the future of large language models, right? But here's the thing. Everyday AI is for everyday people. That is not going inside Google AI Studio or, you know, vertex, you know, Google AI's, you know, vertex AI, which is more of kind of like a developer playground, right?

So you kind of have to be a little technical to be able to use all of these features right now. So on the front end of Gemini, which is how I would believe, you know, 90% or more of our audience would, you know, access Gemini that way, the same way that, you know, there's, there are capabilities with, you know, GPT4 that you can only get if you're using the playground or using their API. But for the most part, what we get in the chat GPT version is pretty similar. And that's not the same case with Google Gemini. So yes, you always hear these things like, oh, you should use Gemini because of their,

you know, their million token context window. okay, well, yeah, you've got to be a little bit of a dork, you know, and use Google AI Studio or use Vertex AI. And that is, at least for now, pay as you go, even if you are on a monthly paid plan, right? Because the context window, yes, is very expensive. So that's something. The access to all of these features right now is bad. And it's not just that, right?

the vague release dates also related to access, right? So let's talk a little bit about that. Hey, Gemini 1.5 Pro with this 2 million context window. When they said release date, all they said is private preview and there's no specific date provided. Ask photos in Google. That looked awesome. No specific date provided or mentioned at the I.O. Conference.

Project Astra. there was no date given. Right. So I think a lot of these kind of headlining features, one thing that I think was kind of bad about that announcement is they didn't do a good job of saying when we were getting all of these features or if they were going to be publicly available, right? Which again, I think goes to Google is just, I think, they are more marketing right now

than they are like meat, if that. makes sense, right? There's not a lot of meat on the bones right now. It's just a lot of marketing. So, yeah, obviously, if you watched Google's I.O, like I did, it seems very impressive, right? But when will we get access to all of this? Or will we? Or will it only be available for developers, right? Or might it be another year or two? So that's the other thing, right? So yes, you know, Open AI, when they had their announcement, they didn't roll out. everything, but one of the main things was their model last week, GPT-40. And they rolled that out immediately for paid users.

Like literally, I had it within an hour, right? And they said, hey, the rest of these features, kind of this live Omni and the desktop app, will be rolling out in the coming weeks. Okay. So although they didn't give a specific date, GPT-40 was available almost immediately to paid users. And then they at least said, hey, all these other features are going to be rolling out in the coming weeks. Microsoft, similarly, at their, you know, Build Conference,

I think they were a little bit better. You know, they gave a date. Hey, a lot of these, you know, co-pilot plus PCs, June 18th. A lot of these features, June 18th. They gave a hard date. So I don't know why Google has always struggled with this. But actually, as we get into the WTF, We're going to get in there because I think the branding, the marketing, the marketing, and the marketing focused execution and discoverability were just head scratching, right? Legitimately had scratching. All right. So let's go over some of the WTF stuff here, right? And I know this has just become a joke, but Google said AI more than 140 times in its Google I.O.

keynote. And they kind of made a joke. They made a joke about it, right? So Google CEO at the end said, oh, I know we, you know, there's a meme last year at our conference about how many times we said AI and, you know, we actually had AI count how many times we said AI this time and they put it up on the screen and everyone laughed and clapped. All right.

Well, I get it. But also, I think companies need to start focusing less on all this buzzwords, right? I mean, same thing. like we've actually covered this on the show before, looking at earnings calls and looking at how, how much more of the terms AI, large language models, generative AI are being announced. But again, it was just so, so scripted, right?

Google's announcement, especially compared to, you know, in nine days, we essentially, you know, it's kind of like the, you know, the NBA finals week of AI, right? This week. So in nine days, we had big announcements and big events from OpenAI, from Microsoft, and from Google. And out of those three, Google's just, I mean, it almost seemed like it was done by AI. It was overly polished.

It was very robotic. It was almost like overly sanitized, if that makes sense. Which, again, I think one thing about AI, it's trust, right? And I don't know. I don't have a ton of trust with Google right now in their AI products. It seems overproduced. It seems like marketing. It seems robotic.

It doesn't seem friendly. It doesn't seem inviting. Again, maybe that's just my personal opinion. And, you know, hey, I mean, also, am I a little biased? Sure, because, you know, even this podcast, this live stream, it's unscripted. Right. Yeah, I sometimes have slides and notes, but it's unscripted. It's unedited, right?

Yeah, make a lot of mistakes sometimes. But I think, especially when it comes to AI, I think you need that ability to see. seem human. You need that relatability, right? We saw it even in Open AIs announcement. They did it live. They demoed it live, right? And there was some hiccups. There was some things that went wrong, which I think, though, is relatable. And people need to understand, you know, large language models, AI, it's a lot of it's degenerative. And it's normal to have things go off track. So when Google had this overly polished, overly marketed, very robotic presentation, I don't know. To me, rub me the wrong way, didn't like it. And they just were saying AI like every third word. Also, pretty interesting.

And this kind of went a little viral. So someone named Scott Jensen, who, you know, was a former employee of more than a decade here at Google. I'm going to share this. And this came out kind of just right after Google's I.O. So what he said in his LinkedIn post that went super viral said, I just left Google last month. The quote unquote AI projects I was working on were poorly motivated and driven by this panic as long as it had AI in it. It would be great. And then saying it is stone cold panic that they are getting left behind and it's not being driven by user need. And that's kind of at least the takeaway that I got from Google's IO announcements, right? They're seeing things that Open AI is released.

I mean, they obviously know they have inside information. They hire, you know, employees from all the other companies. So I'm sure they know what other companies are working on. And it did seem. It just seemed like a bunch of marketing. Let's just throw some slides and some demos up there. Let's not really have a release date, an actual roadmap. Let's just, you know, essentially wow people.

with marketing. Hopefully analysts respond positively. Hopefully our stock price goes up and, you know, we'll start delivering whenever, right? Just throw a bunch of AI in it, make a cool marketing presentation, a cool demo video. Hopefully it works, whether it's all real live on edited. I'm not sure. Is it one X speed? I don't know. Right. But that's just kind of the vibe that I got. And, you know, here you have a veteran Google employee who said that he worked on AI projects and he had been working at Google for more than a decade, said it. You know, it said it was just driven by panic. And, hey, as long as there's AI in there, it's going to be good.

So I don't know. I kind of had this same feeling. And, you know, Scott in his post, which will link to in our newsletter as well, said, you know, it's not, you know, he said it's not just Google. He said Apple is no different. All right. Here's another thing. Discoverability. This is one thing.

It really irks me with Google. First of all, they change names for things all the time, all the time, right? So I don't know, maybe something wasn't working very well. And then they, you know, needed to give it a quote-unquote facelift and they need to make it better, et cetera. But, you know, an example, Google gems, right? You want to go find more information on Google gems? Even Google search can't find it easily because when you search in Google gems, you can an auto correct that says Google Games, right?

And you've got to scroll quite a bit to find information about Google Gems, which I think is going to be one of its biggest, you know, its biggest announcements coming out of this is Google Gems. I think that's going to make Google Gemini actually usable for a lot of people. Right now, I don't think Google Gemini is a good model. I think it is in fourth place, right, behind GPT40, behind Windows co-pilot, which obviously uses GPT4, GPD40, and then behind Clawed. I don't think Google Gemini right now is a great large language model.

Yeah, it's got a great, you know, a very impressive context window. It has multi-modality input, but you got to have, you know, that tech background. The front facing Gemini, I don't think is very great. So I think Google gems, you know, the ability to quickly kind of make it better for your own use case and then reuse it via this gems is huge. But if you want to go find information on Google gems, you really got to do some digging. So I think even discoverability, right, when Google is constantly changing its branding, changing its marketing, changing names, and then you go look for something and you can't even

find it on Google's own search engine, pretty bad. And hey, for a laugh, go into Google Gemini and ask Google Gemini about all of these individual products. And yeah, you're going to laugh about it. Much better results in Chat Choo-a-T, by the way. All right. So that's it covered kind of what's good, what's bad, and what's WTF. So quick recap.

So what I think is good from Google's I.O. Announcement, Project Astra super impressive, VO, their AI video product, gems, kind of their version of GPT, I think all super impressive. Some bad things is access. It's not easy to access all of these great features. and also the vague release dates, right? You got to at least say, hey, buy fall or by, you know, this date or in the coming weeks like OpenAI. So I think that was pretty bad from the IO announcement.

And then the scratching, the WTF, I mean, the branding, just being overly market, like overly marketing, polished, you know, and then the discoverability, I just think things were WTF. So I hope this was helpful. Thank you for joining us. If you're on the podcast, make sure to check out your show notes. You can always reach out to us. But leave us a review and rating on Spotify and Apple. If this was helpful, we'd appreciate that.

Also, we'd appreciate it. If you haven't already, go to your everyday AI.com. Sign it for the free daily newsletter. Every single day, we recap the topic of the day as well as going on, going over every single thing that you need to keep up in the world of AI to grow your company, grow your career. Thank you for joining us. We hope to see you back tomorrow and every day for more everyday AI.

Thanks, y'all. Meet Firefly AI Assistant. Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time.

See it today at firefly.adobie.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.