EP 250: AI News That Matters - April 15th, 2024

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.

Will Google's AI announcements finally make them a little more relevant in the generative AI game? What's going on with Adobe and also OpenAI in a data leak? All right, there's a lot going on in the world of AI news. We're going to get to all of that today and more. What's going on, y'? All? My name is Jordan Wilson. I'm the host of Everyday AI. Thanks for joining us. This is for you. We are a daily live stream, podcast and free daily newsletter helping everyday people like you and me not just learn generative AI and what's going on there, but how we can use all of these, these news developments, new tools, new tips, new softwares, how we can actually leverage this to grow our companies and to grow our careers. So on most Mondays, we come to you live and bring you the AI news that matters. So that's what we're doing today. Well, a little bit, if you're joining on the podcast, you probably won't see or, you know, feel anything different. But for our live stream audience, you might notice I have these kind of vacation vibes going on. That's because, yes, I'm actually on vacation. But it is so important to stay up with the AI news that taking a little, little break on my Sunday afternoon to make sure that you have the news that you need for the rest of the week. So, hey, if you're joining us live Monday, I might technically be on a roller coaster or something, but make sure you still leave your comment. Let me know what is the news that matters to you, what questions, what is actually going to impact your company or your career? And hey, there's, who knows, probably the fact that I'm recording this on a Sunday, there's probably going to be some huge breaking news. But don't worry, we'll still have it in our newsletter. So make sure if you haven't, please go to your everydayai.com and sign up for that free daily newsletter every single Monday through Friday it comes out so you can keep up to date. And not just keep up, but how. But you can get ahead. And I tell people it's like a free generative AI university on there. We have access to now like 250 backlog episodes of great content around generative AI from some global thought leaders. All right, but let's just get straight into it. What is the AI news that matters for you for the week of April 15th. So let's start with the, I guess the big piece of news here. So Google Google from its Google Cloud Next conference introduced a lot of new generative AI features. So we're going to go over them kind of in bullet point. We've already talked about them multiple times in the newsletter already, but let's just dive straight into it. So Google has unveiled a lot of new AI announcements from its Cloud Next 2024 presentation. So Google Workspace users will soon have access to these new AI powered tools to enhance their productivity. Let me just hit pause there and preface this, we hope, right? So you know, Google said this will be a general release, so it should in theory be rolling out to everyone if you listen to the show at all. You know that I normally have a little bit of a bone to pick with Google because, you know, as great as their, you know, new generative AI features seem or sound, it has been so far very difficult. Even if you are on a paid enterprise account or on a paid advanced account, it's been very difficult so far if you are a Google Workspace user to take advantage of all of these generative AI features from Google. Right. So as an example here at Everyday AI we, you know, our team uses Google Workspace and we pay, I don't know, 15, 20amonth, something like that to use all these features. But it's, we still can't connect our data. So you know, this has been six months in the making. So hopefully this kind of, this signaling from Google that this is going to be a general, general release. That means that, you know, for a lot of people that have been waiting months or more than a year to use all of these Google AI features at their, you know, small, medium sized business, connect all their data if they wish to, we'll have that option. So we'll see just how to get that out of the way. Right. We don't just spoon feed you marketing talk, we like to tell it to you how it is. All right, so let's go over some of these newer AI features. So the Help me Write feature, which a lot of people have had technically beta access to for many months, should be rolling out in a general general release. And the Help me Write will feature inside Google will now allow users to trigger generative AI with their voice, making it even more convenient to start writing or to continue writing. So whether you're writing an email inside Google, Google Docs, et cetera, Gmail users will now have the option to polish Draft. Hey, look at that. Google using our polish from Prime Polish, right? So Gmail users will have the option to polish draft using Gemini to enhance the quality of their email drafts. Google Sheets as well getting some updates. So they've introduced alerts for cell changes to keep everyone updated and emphasize new templates to stream to streamline spreadsheet creation. So if you're working with your team, Google Docs will now support tabs about time. All right. Allowing users to work on multiple documents within the same file. Efficient Google Chat now integrating Gemini to summarize messages and translate conversations into 69 languages. As well as Google Meet should be offering automatic transcription about time and translation features with the ability to take notes during meetings as an optional add on for $10 per user per month. So, you know, there's a lot more but probably one of the biggest ones saving for the last year. So the highlight I would say of their announcement was Google Vids, which is a AI powered tool that helps users create videos by generating storyboards and compiling rough drafts using various media elements. So there was actually a lot more happening at Google that we, you know, already talked about in the newsletter. But let me just tell you a couple more. So obviously the general availability of Gemini 1.5 Pro, so that's Google's large language model with 1 million token context, which is huge. You know, Google also talked about agents, you know, pretty a lot of talk actually around agents and the ability to kind of build, you know, agents that can connect with your data and perform certain tasks. Right. So right now I think when people think of generative AI, they think of it, oh, it's just kind of, you know, helps me write content or, you know, photos or videos. And that's definitely not the case. Right. I've been talking about that here on the everyday AI show now for a year, you know, talking about how agents I think are going to be much more impactful much sooner than people realize. So that I believe that is reflected here right in Google's really just heavy marketing. Their new agents also support for Nvidia's Blackwell GPU system which I think should be pretty huge. Also the, the Vertex AI model garden which you know, so in Google's AI, kind of their vertex, I won't call it a sandbox per se, but it's, it's where all their different AI models are connected. The ability to switch between, I think it was dozens of different models. So obviously Google's Gemini Anthropics, Claude Llama from Meta Mistral etc, So a lot of open Open Models as well as Claude from Anthropic as well as their Vertex AI agent builder. So a lot going on at this Google Cloud Next conference. Again, we'll be detailing it even more in the newsletter. All right, next, but not least. So Open AI has fired two prominent researchers over alleged information leaks. So according to a report from the information, OpenAI has dismissed two researchers and hopefully I get the names right. I'm probably not going to, but Leopold, Ashton Brenner and Pavel Ismailov definitely didn't get those right. But you know, has fired Leopold and Pavel for suspected information leaks. Again, that is according to reports from the Information. So Leopold is known for his work on AI safety and was reportedly aligned with OpenAI's chief scientist, Illa Shuttavar. So this incident marked one of the first public staffing changes since CEO Sam Altman's return to the board in March following an inquiry by OpenAI's nonprofit board that cleared him of previous allegations going all the way back to his November firing and rehiring. So, yeah, there's been a lot of kind of messy behind the scenes stuff going on in the last couple of months from OpenAI, but it had kind of been quiet after, you know, Sam Altman was, you know, kind of rehired back on as CEO. You know, he kind of made the, the media rounds, you know, kind of really, I think did a really good job kind of controlling that narrative. So it's been kind of a little silent in the, you know, open AI drama now until these recent firings. And so reportedly the internal investigation revealed that both of these individuals were part of OpenAI's safety safety team, adding a layer of complexity to the company's internal dynamics. So, you know, right now we don't really know what data was leaked, but I mean, you have to be paying attention to this and it, and it matters, right? Because OpenAI has nearly 200 million users, right. And it is the fastest growing, you know, app or software ever, you know, in human history. I mean, people talk about threads from, from Meta. I don't know if that counts per se. Right, because Meta, all these new users that they had signed up for threads were essentially existing users from their other platforms. So I would still argue that OpenAI and ChatGPT is, is the largest or the fastest growing software ever. So anytime a company that is still young, it's still, you know, startup mode. Technically, I think it's, you know, somewhere around 500 employees. I first of all, I don't think OpenAI gets enough credit, firstly for what they've accomplished with such a small team. But also the fact that we haven't had a data leak in the past two, you know, two years since the world has been talking about OpenAI. So again, we don't know exactly what was leaked here. This is all just according to a report from, from the information. So we'll be linking that if you do want to read that. But it could have, you know, honestly, just huge rippling impacts across the, across the industry. Because here's why I think there's, you know, still hundreds of thousands of large companies here in the United States that are kind of, you know, one foot in generative AI, kind of one foot in the generative AI pool and one foot out. And I think that this is going to cause a little bit of hesitation for some of these companies to kind of go all in and into the, into the generative AI pool, so to speak, which, you know, that's on them to, to make their, their own calls on that. But I think that you have to look at this as part of a broader picture of, of data security. Right? I, I mean, look at. Over the past couple of years, there's been huge confirmed data leaks from, you know, big Fortune 100 companies. You know, it's like, I, I feel you hear about these almost every month. So is it alarming to reportedly have some, some data leaks that lead to firings? Not necessarily, but I think it is different when you're talking about a company like OpenAI. And I think that people for the most part haven't really been sure, like, oh, should I upload data? Right? Like, you know, I know there is a big Home Depot, you know, data leak a couple of years ago, but it's like, all right, well, you know, aside from if you have an account on home depot.com, which I don't know how many people do, you know, it's like, okay, you know, maybe your credit card information was in there somewhere. Not. But aside from that, it's like people aren't necessarily uploading a lot of their information to Home Depot or to a lot of these other companies, you know, that have gone through these kind of data leaks or kind of these scandals in terms of, you know, keeping consumer and client information private. So I think that's a little different because people haven't really known how they should be using, you know, large language models like ChatGPT, like Anthropics, Claude, you know, so all these kind of generative AI tools that allow you to upload information. I think a lot of people have been uploading their information, maybe in a lot of cases when they shouldn't. Right. So I should, as I normally do, put the preface out there that unless you're on an enterprise plan, you should not be uploading confidential, sensitive, proprietary information into these large language models. But I think so many people are. Right? So that's why I think you do have to pay attention to this report. And I'm sure in the coming weeks and months, we're going to see a little bit more information come out of it. It, but you have to be paying attention to it. All right, next, speaking of Open AI, so YouTube's CEO kind of fired some warning shot at Openais for its reported training practices for its AI image, or, sorry, it's AI video model Sora. All right, so YouTube CEO Neil Mohan has mentioned that if OpenAI had used content from YouTube to train its new AI video program Sora, it would be considered a clear violation, in quotes, of YouTube's terms of service. So this is following the story from last month where OpenAI's CTO, Miratai, was unable to confirm the type of content used to train Sora in an interview with, I believe it was the Wall Street Journal where. And this was raising a lot of transparent. A lot of transparency issues, right? So we actually played it a short clip here on the show a couple of weeks ago where she kind of was asked a somewhat simple question about, hey, how is this, you know, this new model from OpenAI called Sora, which produces amazing results so, you know, text to video? And she was asked just kind of point blank, hey, how was this trained? Was it trained on, you know, YouTube videos? Is it trained on social media? And she kind of hesitated almost in a, like, why did she hesitate? Like that kind of way. And I think that this kind of implied to everyone that sora's trained on the open Internet. But guess what? The open Internet obviously trains or, sorry, the open Internet obviously contains a lot of copyrighted materials, right? So this is where we get into this, this, this new kind of report featuring this quote that we just heard or this kind of warning shot that we just heard from YouTube CEO Neil Mohan. So Mohan has emphasized that creators who upload their content to YouTube expect their work to be used in accordance with the platform's terms of service, and any unauthorized scraping or downloading of YouTube's content would breach those terms. So although Mohan did not explicitly confirm whether OpenAI did actually use YouTube content to develop Sora, he did highlighted that such an action would pose a significant issue so obviously OpenAI has faced a lot of scrutiny over the training of its new AI video model Sora amidst, you know, all these concerns about copyright infringement and data sourcing. Right. So especially, you know, in the couple of weeks since this, this interview with Mira Mira Tai, like a lot of people have just been saying, okay, well, so if she stumbled on a very basic question like that, you have to kind of assume that, you know, that OpenAI just used the open Internet, everything out there. So, so copyrighted materials and everything, again, that's the assumption that, you know, kind of everyone is, is under now. So we'll see where this, this goes and maybe we'll save this for a hot take Tuesday. But I would really keep an eye on this kind of, you know, YouTube, which is obviously owned by Google, which is a big competitor now to OpenAI. Right, because OpenAI has, you know, partnership with Microsoft, which is probably Google's chief competitor. So here you kind of have, you know, Google versus Microsoft in the lens of YouTube versus OpenAI. So I do see, you know, I've been saying this for a long time, look at the kind of the OpenAI versus New York Times lawsuit. Whenever that may get settled, I doubt it's actually going to go to a trial, but whenever that gets settled, that's the first big domino to fall. So maybe in theory this could be one of the next, you know, kind of this YouTube versus OpenAI. And you know, I know OpenAI has kind of made the case for like, hey, what is actually copyrighted? Like, right? Like is copyrighted material what it meant decades ago? So I think we're going to be hearing a lot about the concept of copy copyright, copyright law. What does it mean if you share something online, even if you are, you know, protected by a, you know, YouTube or a meta, when you upload this content there, what does it mean and what does it mean or will it mean when you know, so many or so much of the content that we are going to be uploading will be AI generated, Right. I can see a time in the very near future, whether it's months or years from now, when so much of the, you know, information that is even uploaded to a YouTube or something like that is from a, you know, program such as Sora or maybe Runway or Mid Journey, right? Like as these other companies are starting to add kind of this video capability after everyone saw the literal jaw dropping capabilities of sora. So I do think that so much of what's going to be playing out online, you know, at least YouTube, social media and who knows, maybe, you know, eventually we'll see it in, you know, mainstream media, right, when you have your B roll, right, when someone's talking on the screen. Maybe like right now, right. Maybe in the future when I'm talking about something we're going to see real time AI generated video that's splashed up there. So it's like, okay, well who owns it, what was it trained on and who's ultimately getting paid for it, right? So definitely something to keep an eye on. Hey, speaking of that, Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Genai. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for ChatGPT training for thousands, or just need help building your front end AI strategy, you can partner with us too. Just like some of the biggest companies in the world do. Go to your everydayai.com partner to get in contact with our team or you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on gen. Related story Right, so Adobe has been accused of training its ethical AI image model on AI images. Yeah, kind of inception here, but let's break it down. So there's been a lot of ethical concerns arise over Adobe's AI image generator, Firefly. So Adobe's AI image generator, Firefly is known for its ethical training on license stock images. But it has recently faced controversy as it was revealed in a recent report that some images were sourced from competitor Mid Journey and potentially without proper licensing. But then also it's like, how do you, how can you license images from a company like Midjourney that is technically training, in theory, off copyrighted materials? Right. So you get this whole cycle of like, where does the new content start, the AI image generating begin? Right. So it's, it's, it's a little messy, but according to the report, approximately 5% of the images used to train Firefly were from questionable sources such as other AI image generators. But Adobe assures us that all non human Pictures are still copyright safe. That's the other thing, right? So with, so with so many of these AI image generators, like, you technically don't own the copyright, right? So if I go in there and if I, you know, create a bunch of images with a mid journey or you know, Dall E from OpenAI, I don't technically own the copyright on those. They're, they're technically kind of copywritten by no one because they're, you know, it's, it's, there's no technical original work being made there, or at least that's the argument that people are making. So it does get a little confusing. But Adobe has claimed that every image submitted to Adobe stock, which is where they are training their Firefly model on, including those generated with AI, undergo a strict moderation process to ensure legal compliance. So despite the training data controversy, Adobe maintains that images created with Firefly are safe to use without copyright infringement. Also worth noting, Adobe is reportedly working on an AI video generator and is rumored to be compensating artists per minute for video clips showcasing a potential shift toward more artist friendly practices. So hey, speaking of that, let's just go ahead and wrap up our AI news that matters with exactly that. So Adobe, Adobe is offering incentives for AI training data collection. So Adobe is investing in its generative AI platform, Firefly, like we just talked about, by offering up to 120 to photographers and other artists for submitting short, including videographers, but for submitting short video clips of everyday actions and emotions for AI training. This is according to a Bloomberg report. So the submitted videos should showcase everyday actions, emotions, basic anatomy, anatomy and people using objects like smartphones, fitness gear, etc. While avoiding copyrighted material, nudity or offensive content. So according to the Bloomberg report, contributors can earn between $2.60 to $7.25 per minute of submitted video, providing an opportunity to earn money from existing video content. So again, you know, our last kind of story led straight into this one, right? So first we talk about how Adobe is, you know, kind of facing a little bit of heat for reportedly using AI images to train its AI model, Firefly. Right. And you have to hand it to Adobe because they've, I think they've been probably one of the least one of the leaders in terms of creating models in an ethical way. Right? Like we don't hear a whole lot from the, you know, from the Googles, from the Microsoft OpenAI, anthropic mid journey Runway, Pika Labs. Right? Like we don't hear a lot specifically on, okay, you're training on the Internet. But doesn't that include a lot of copyrighted materials? Right. It seems like a lot of the bigger companies aren't really talking about that or they're just kind of waiting to, you know, settle some of these cases. Right. And there's obviously been some huge multi million dollar annual partnerships for these big companies, you know, like, like Google, Anthropic, OpenAI, etc, are forming these multi million dollars annual partnerships with these big, you know, content agencies essentially. Right. To, to use all of their content to train their models. But what about all the models that were trained before these partnerships were even formed? Right. So Adobe has been going about it reportedly a different way, you know, really just training its models on all of the information and all of the content that it had access or it had the kind of legal rights to train its models on until this story, kind of these stories came out like one right after another, right. Where first Adobe was apparently using, you know, small percentage, but, you know, reportedly 5% of the total that it was training its model on were from Mid Journey. Right. Which is tricky because all of those were, you know, in theory, trained on copyrighted materials. Right. We've talked about that here in the show all the time. You can just say, you know, give me a picture of a superhero in Mid Journey and it's going to spit out copyrighted materials. Right? You're going to get an Iron man and, you know, Superman and all these things, even if you aren't asking for it. Right. Because the model has been trained on presumably a lot of copyrighted material. So that does really throw a wrench in, in Adobe's whole, you know, super ethical, like ethically trained campaign. I think it would have been better for them to either, hey, we're going to quote, unquote, suffer in quality, or we're going to be a little slower by not using any of these, you know, any of these AI images in general. Because once you do that, you really, I'm not saying you close the book on your case that, oh, we're going about this in the right ethical way, but I think you really soften your ability to really take a hard stance. That's saying, like, hey, we are the super ethical AI model. So I don't know if this was intentional and if this was accidental. It could be hard, right? Because it, it does appear, according to this, to this report, that some users are submitting AI generated content to Adobe's stock services. So, you know, really, it's going to be interesting how this plays out as well, as, you know, kind of interesting also that Adobe is paying, you know, paying creators to say, like, hey, go, go shoot some videos. We're working on an AI model, so, you know, we're gonna pay you. Which I like that concept. But also, hey, as a former, you know, way back in my former life, I did a lot of, you know, videography photography, right? So, you know, this was obviously like, you know, almost 15 plus years ago, but still, right. So they are paying a lot of these creators, a lot of these videographers, you know, dimes on the dollar to essentially say, like, hey, in the future, a lot of the clients that are maybe paying you for this right now, right, Are going to just not be paying you. Right. They're going to be using our models instead. So it is this kind of this weird state of creativity that we're in right now where, hey, I do like this, that Adobe is, you know, making this very ethical effort to pay creators for specific footage. But all they're really doing is saying, hey, creators, we're going to pay you for this. Because in two years or 18 months or five years, all the people that are paying you to create content right now may not be paying you, they may just be using our services as well. So it is kind of this sticky gray area that, that we're in right now in the creative space where, you know, either these companies are just training models off the open Internet or they're paying creators. But, you know, essentially that means that they're going to be getting paid less by their, their current clients, current businesses that, that they're working with in the future. That's definitely a reality. All right. There's always more. So if you haven't already, there's a lot of news there, there's more. So make sure, if you haven't already, please go to your everydayai.com sign up for that free daily newsletter. Yes, I took a little break on my Sunday afternoon here. Normally we do this show live Monday, so if anything is late and breaking, don't worry, we'll still have that in the newsletter. So make sure that you go subscribe to that. If this was helpful, we'd really appreciate you leaving us a rating. If you listen to us on Spotify or Apple Podcasts or if you would just, you know, if you're watching here on social media, if you could, you know, share this, repost it to your friends to keep them up to date, we'd appreciate that as all. And hey, we'd appreciate you tuning in for the rest of the week for more Everyday AI. Thanks, y'. All.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic. Visit your 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.