EP 324: AI News That Matters - July 29th, 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.

Meta has a new model. Why is Apple now delaying its AI features? And OpenAI is now going after Google and perplexity with Search GPT. What is does that even mean? Well, we're going to be answering those questions and a lot more today on Everyday AI. What's going on, y'? All? My name is Jordan Wilson and I'm the host of Everyday AI. And this thing is for you. This is your daily live stream, podcast and free daily newsletter helping all of us not just better understand AI, but how we can leverage it to take advantage and to grow our companies and careers. So each and every Monday we bring you the AI news that matters our. All right, so if you're brand new here, each and every Monday we take about 30ish minutes to just tell you what's going on in the world of AI. So before we get into the top AI stories for the week, I have to shout out our new Thanks a Million giveaway campaign. So it's going to be in our newsletter today. It's super simple. We are getting close to a milestone here at Everyday AI with nearly a million downloads and we're going to be doing a huge, huge giveaway. So you need to make sure, if you haven't already, go to our website, sign up for the free daily newsletter at your everyday AI.com and get into that campaign before it's too late. Trust me, you're going to want to enter. All right, so with that, let's get into the AI news that matters for the week of July 29, 2024. Like I said, we do this almost every single Monday so you don't have to waste hours a day or hours a week trying to keep up. We keep you caught up each and every Monday. All right, let's dive in with the AI news that matters for this week. All right, well first Meta, they released its next open source AI model, llama 3.1. So Meta has released llama 3.1 featuring the new groundbreaking 405 billion parameter model making a significant advancements in large language model technology. This update from Meta includes major improvements to the existing 8b and 70b parameter models and introduced a new 405b model designed for a wide range of applications. Yeah, so essentially before meta had just two different versions. When they announced Llama 3 back in April, they had a small 8B and we'll call the medium 70B, and they hadn't yet released the large one, 405B. So just on Tuesday they not only finally released the large one, the 405B, but they also updated the other two models as well. So kind of two different announcements into one. So both introducing this larger 405B model and then also updating the small and the medium, and that is the 3.1 or going from the 3 to the 3.1. So the 405B model has been highly anticipated and is positioned as a direct competitor to industry leaders like GPT4 Omni and Claude 35 Sonnet in the MMLU benchmark, which is the kind of the gold standard for large language models. Llama's 31405B model scored pretty much the exact same as the GPT4 Omni model, so showcasing its competitive edge there. So pretty impressive benchmarks. And one of the more exciting aspects of llama 31 obviously is the fact that it is open source, setting it apart from those proprietary and closed source models like OpenAI's GPT4O on GPT4 Omni, like Anthropics, Claude 35 Sawnet, like Google, Gemini, 1.5, Flash, whatever you want to throw out there. Most every single model that we talk about on the show for the most part is usually a closed proprietary model. But this open accessibility from Meta allows researchers, developers and businesses to freely access, tweak and build upon the model, potentially accelerating AI innovation and democratizing advanced AI capabilities. So if you want to get your hands on the new model, it is available if you go to Meta AI platform and you have to just log in via a Facebook or Instagram account or you can download it from Hugging Face, from Meta's own website, et cetera. So Meta plans to incorporate multimodal capabilities in the future, as well as agent like functionalities and possibly a premium version of the model in future updates. So we did not get that the multimodal and the agentic capabilities here in 3.1. It did seem like Meta was working toward that, hit some snags and instead just released the 3.1 model as is no multimodal. So right now it is just inputting text and then outputting text and photo. So it can't output photo, you know, nothing, you know, groundbreaking there with Meta's AI image generator. But I do assume it's going to get better in the very near future also. I mean this is pretty big because this release signals Meta's intent to lead in large language model development and challenges the status Quo by making cutting edge AI technology more accessible to a broader audience by making it open source. So yeah, pretty completely different approach here from Meta versus all the other tech conglomerates versus the other, you know, trillion dollar companies making their own models. Meta is just going open source here and they're saying they are so confident in their models. Yeah, go ahead, fork it. Here's, here's everything. Try to, you know, find bugs, find flaws and we will fix it and we will build upon it. Also make sure you check out our YouTube channel and we'll probably throw in the newsletter today as well because we did even show you how you can download and install LLAMA 3.1 locally on your own computer. Yeah, that's the biggest that you know, if you're a brand, brand new person in large language models or generative AI, you might not even under fully understand this open source concept. So number one, it allows you to build upon it, tweak it, you know, use it to build actually other models and distill other models. So great capabilities there, but you can also download it and run it offline. So it is the most, you know, technically open source models are the most secure way to run large language models. Right? Because otherwise you are still connecting to a third party service from a Google, from a Microsoft, from an Amazon, from OpenAI, whoever it is you're using, where you know, on a local model that you can download and run on your own machine, you can do it without Internet, right? Once you download it, you actually don't even need the Internet to use it, which is, which is pretty cool. All right, our next piece of AI news. And this might be one of the bigger stories of the week. Apple has delayed its AI features in iOS 18 until October. So according to reports from Bloomberg, Apple has announced a delay in the rollout of its AI features for iOS 18, which were initially expected to launch with the iPhone 16 in September. So now the AI features will be included in the 18.1 update, slated for release in October 2024. So this is all according to Bloomberg, but these delays allow Apple to address potential bugs through its developer and beta testing programs. So this move is seen as a strategic decision to ensure the AI features are both accurate and secure, avoiding the pitfalls experienced by rivals Google and Microsoft as well others. So at the WWDC in June, Apple showcased the potential of its AI capabilities, which have been met with, with mixed reactions. I mean, well, they didn't even really call it AI. They called it Apple Intelligence, which I thought was kind of cheeky, but you know, Apple is trying to, in typical Apple fashion, redefine what artificial intelligence is, yet they are not even announcing it or releasing it on time. This delay though, underscores Apple's commitment to quality and security, aiming to avoid the public and critical backlash faced by competitors when they first released their AI tech. Yeah, if you go back and look, I mean especially Google, Google did not do a good job rolling out Gemini. They had a kind of sneaky marketing campaign that really painted Gemini in a light that wasn't really true. So Apple is trying to avoid that and trying to seemingly under promise and over deliver. So the iOS 18 developer version is expected to be released in the coming days, offering a preview of the new features to developers and beta testers. This decision may impact the excitement around the iPhone 16 launch, but is intended to enhance user experience in the long run. So ultimately what this boils down to is there was concerns about stability, right? Is this going to work? Is it going to break Apple? You know what, we've said it here all along, they're never the first person at the party, but they're always the coolest kid. But you got to make sure if you're the last one to enter the large language model or generative AI race, at least when it comes to your front facing your consumer product, you got to make sure you get it right. So just some concerns about stability here and user experience from Apple. Also reports did not indicate, and we're not sure if this is because of the OpenAI integration, right, because there's actually two different pieces here that Apple is running under the hood and they didn't even really call this out by name in their keynote, which I actually thought was a shame. But the way that Apple is actually doing their Apple intelligence is there's two different models. So they actually have a, a local Edge AI model, right. So there it's their own proprietary 3 billion parameter JAX XLA small language model. Okay, So a lot of what's going to be happening when it comes to these generative AI capabilities on the upcoming iPhone is going to actually be be processed locally by this JAX, this 3 billion JAX XLA small language model. And then whatever that model cannot handle will be passed off to this direct integration with GPT4O. So we are not really sure where the concerns are right now with this new Apple intelligence roll off. But regardless, it's not ready. Right? Apple said, hey, we already waited a year or a year and a half longer than most of our competitors. We'll, we'll delay it another month. So yeah, we thought we were, you know, about a month away from starting to see this in Apple's next iPhones, but it doesn't seem like it's going to be ready in time for the new iPhone launch. So 18.1, keep your eyes out on that iOS development because according to reports that is now where we're going to be seeing. And hey thanks thanks to the live stream audience from for tuning in. What do you think? Do you think that we're actually going to be seeing this new Apple Intelligence in 18.1 or are they going to kick the can a little bit? Might we be waiting a couple more months? And yeah, I'm curious if you think it is their own small language model or is it this integration with GPT4? Oh from OpenAI. So should be pretty, pretty interesting to see here how this all plays out. All right, our next Big AI News that matters. Well, two big banks, Morgan Stanley and JP Morgan Chase bank have gone all in on large language models. So Morgan Stanley has successfully launched its second generative AI tool for financial advisors, emphasizing the benefits of custom built solutions over promoting pre made options. So Morgan Stanley's new AI tool called AI at Morgan Stanley debrief summarizes video meetings and generates follow up email drafts, enhancing efficiency for their financial advisors. The tool was developed in partnership with OpenAI following the launch of their AI Knowledge Assistant tool in September 2023 which helps advisors quickly find information from Morgan Stanley research. All right, and that is not the only big bank going now all in on large language models. Chase bank also reportedly is in on the action according to a new article from the Financial Times. So Chase has now introduced a new generative AI product internally known as LLM Suite, designed to assist JP JP Morgan Chase employees in its asset and wealth management divisions. So the LLM suite is described as a chatgpt like product that can perform tasks typically handled by research analysts such as writing idea generation and document summarization. According to the internal memo seen by the Financial Times, the tool provides access to third party models and is intended to enhance general productivity. Approximately 50,000 employees, or about 15% of J.P. morgan staff will now have access to LLM Suite, making it one of the largest uses of large language models on Wall Street. So pretty impressive news there from both Morgan Stanley and JP Morgan Chase and it will be also interesting to see how their respective stocks respond here this Monday morning after these announcements. Also worth noting, JP Morgan Chase one of was one of the first, you know, highly visible big name companies very early on that did an all out ban of chat GPT. Right. And at the time I said this is not a good move from JP Morgan Chase. And I said within a couple of months or a couple of quarters, they're going to realize that by doing this they are losing billions or tens of billions of dollars in revenue. And I, at the time I said I expected them to do a 180, right. Because if you are one of those companies that does an all out ban of generative AI, the weird thing is is that might actually push you to do the right thing and to responsibly and ethically and to, you know, really just adapt to a large language model the correct way. And you know, sure enough, this isn't the first generative AI offering from JP Morgan Chase, but this new report here from the Financial Times, like I just said, would be one of the largest implementations of a large language model, at least that we know of on Wall street with 50,000 JP Morgan Chase employees having access to this. Yeah, pretty, pretty big news there, both from Morgan Stanley and JPMorgan Chase. Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on gen AI. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for ChatGPT training for thousands or just need help building your front end AI strategy, or you can partner with us too. Just like some of the biggest companies in the world do. Go to your everyday AI.com partner to get in contact with our team or you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on gen. All right, we have two more big stories here, both of them from OpenAI. So OpenAI has unveiled Search GPT, challenging Google's search dominance. So OpenAI has introduced Search GPT, an advanced AI powered search engine that aims to revolutionize online information retrieval and challenge Google's long standing dominance. All right, so what's important to know? This isn't available yet. They're just really teasing it. So SearchGPT is currently in a very limited beta release accessible to only 10,000 users as OpenAI fine tunes the engine and gathers diverse feedback. So OpenAI's new search GPT is designed to provide more conversational and context aware search experiences going beyond traditional keyword based searches. So unlike traditional search engines like Google or Bing, Search GPT will presumably work a little bit more like perplexity in understanding the nuances of user queries, remembering previous interactions and offering detailed sourced back responses. So users can join a wait list right now on OpenAI's website to potentially gain access to the beta version before everyone else. We obviously shared about this in the newsletter when we, when this first broke mid week last week. So you know what, you got to make sure to read the newsletter on a daily basis. It's quick, it's digestible. You know, I'm wondering, any of our live stream audience out here, do you also read the newsletter? It's, it's written by, by me, a human. So you know, I try to do a good job at cutting through all the marketing fluff and telling you what's what. So a little bit more about Search GPT. So it's obviously built on OpenAI's GPT4 language model and it's also capable of real time web crawling and multimodal processing, integrating text, images and videos for comprehensive search results. Reports are showing that it is still powered by Microsoft Bing because OpenAI has in ChatGPT has had real time kind of web access now for a long time thanks to the browse with Bing capabilities and functions inside of ChatGPT. So the AI powered engine collaborates with content creators, potentially offering access to premium content and ensuring fair compensation for publishers. That piece is important because OpenAI has been signing a lot of exclusive, exclusive content agreements, right? Because we talked about this in actually on the podcast and the newsletter multiple times last week as an example. Reddit has now officially started blocking different search engines from its, from accessing its content. You know, because Reddit has a, has partnership deals with Google's, Google's Gemini and OpenAI's ChatGPT. So you know, this kind of new search GPT we might see, right? Which I know sounds weird. So you know, right now there's certain airlines, at least here in the US that aren't on certain airline search engines, right? So as an example, you might use a airline search engine like Kayak, but I'm pretty sure that Southwest as an example isn't on Kayak. So we might see something in the future that, you know, you might go to as an example, Search GPT, if you need certain content from certain publishers, and you might have to go elsewhere if you need content from other publishers, depending on how of how all of these content agreements actually shake out and the lawsuits, obviously. So there's obviously a lot of ethical considerations, including safeguards against misinformation, respect for intellectual property and protection against biased content. And those are all reportedly being worked into SearchGPT's design. Potential challenges. Yeah, there's a lot. Include privacy concerns, the impact on web traffic, and just digital business models in general. Right. Because here's the reality. All this means less clicks to publishers. Right. You know, probably one of the, one of the easiest examples is looking at these big companies like Quora, like Reddit, like, I mean, probably the biggest one is Stack Overflow. Stack Overflow in particular saw their traffic go down, I believe, in the first year or so after chat GPT and after all these large language models, they saw their traffic go, go down initially by I believe, 30 to 40%. So as all of these online publishers, news organizations are seeing a huge drop in the number of users that are ultimately going to their website, what that means is a couple of things. You're going to see more and more content partnerships like we've seen. Right. So even big organizations like Axel Springer, like the Financial Times, like, like Reddit, there's so many big media groups and traditional news organizations that have entered into partnerships with OpenAI, with Google, etc. So you're either going to see that or you're going to see a lot more lawsuits. As always, make sure stay tuned here. But also keep your eye out on the New York Times versus OpenAI lawsuit. I've been saying all along that that is the biggest domino to fall on. Just not, if I'm being honest. It's not just the future of how we use the Internet, it's the future of how we obtain information. Yeah, let me say that again. That lawsuit, the New York Times versus OpenAI. So the New York Times is suing Open AI for allegedly taking, quote, unquote, taking or scraping millions of different articles that were copyrighted. And you know, according to the New York Times, using that to train its model. So however that lawsuit pans out is really going to change how we use the Internet in the future. Right. One of the demands that the New York Times is making is that the GPT model be destroyed. Yes, that is in the actual lawsuit. Comical, maybe? In the lawsuit, yes. Plausible, probably not. However, you really have to take a look at how that lawsuit plays out. It's going to not only impact, like I said, how we browse the Internet, how we get information, but it's also probably going to impact largely large language models. All right, I'm glad that Jennifer and Fred also read the newsletter and yeah, Michael, I might have called this a while ago. All right, so our last piece of AI news for the day. This one's pretty big. So new reports are showing that Open AI faces a potential 5 billion with a B5 billion dollar loss in 2024. So according to a new reports from the Information, OpenAI might incur a loss of up to $5 billion this year based on an analysis by the Information. So despite being valued at $80 billion, the company's high operational costs are driving significant financial losses. Again, this is just according to this report from the Information, but internal financial data and insights from individuals involved in the business suggests that OpenAI will need to secure additional funding within the next 12 months. Reports also show that OpenAI's operational costs are staggering. So OpenAI is reportedly spending about $7 billion on training and inference for its large language models and around 1.5 billion billion on staffing, which is a ton considering I believe the last headcount at OpenAI was like less than 600 employees. So a lot of people don't understand. You know, when you look at Microsoft and Google, those are companies with hundreds of thousands of employees. When you look at OpenAI, they're technically still a startup. So that is actually a huge staffing costs when you talk about 1.5 billion for at least according to the last headcount that we saw was just under 600 employees. I'm sure it's a little more by now, but still. So the rapid growth of OpenAI's business is contrasted by its substantial expenses, highlighting the challenges of scaling AI technologies. So the company's financial situation could impact its ability to to maintain its current trajectory without new investments. And this development raises questions about the sustainability of high growth AI ventures and what is their financial strategy. So obviously we talked about potential lawsuits from the New York Times, but I think this is also reflective now from stiff competition from Anthropics, Claude from Meta coming out here with this new open source llama 3.1 that is at least right now showing comparable scores to OpenAI's GPT4 Omni. Obviously Llama 3.1 right now is not multimodal. As you know, Chat GPT has many more multimodal capabilities right now. But I mean, here's what I think is going on y' all open AI has been. They've kind of taken a slightly different approach here. In the last couple of months, we've seen a lot of talk on the new desktop model, right? So this new GPT4O, people are always like, oh, well, it's free. Why would I have a paid account? Well, that's because there are so many new features that should in theory be dropping any day. So last week, there was actually some reports over the weekend that showed OpenAI was pushing some small developments and working on some small developments over the weekend, which generally happens right before a big release. So we probably should be seeing either this kind of new desktop capability. So for Mac, if you have the paid version of ChatGPT plus, essentially on the Mac, you will be able to click one button and the ChatGPT will be able to see your screen. It will be able to understand what's going on in near real time. And also these new more neural kind of voices, right, with much less latency than the current versions. So think of like a way smarter, right? So ChatGPT has had the kind of the voice capabilities now for many months, which is pretty good, right? Actually, if you compare it to Siri or Alexa, I would actually say the voice assistant is much better. Not only is it about the same speed, but it is way more accurate. So we should be seeing the newer version of kind of the ChatGPT's voices or text to speech, much more neural, lower latency, etc. But also this new desktop version, which is kind of like an agent, right? The fact that you'll be able to just click a button on your computer, CHAT GPT will be able to see and you'll be able to talk to it, right? You can just be like, hey, what's like, why am I getting this wrong? Rewrite this code. Or you could just click on, you know, you got a bunch of tabs open, you can just narrate what you're doing and then say, hey, write me this, this outline for this project that I'm working on. I'm doing all of this research and here it all is. Can you write this for me? Right? So it's pretty exciting what OpenAI is working on, but at the same time, like I said, they've taken a little bit different approach because I think early on, right, even when you just look at the original ChatGPT, it just kind of came out of nowhere. And even when you look at the first paid version of ChatGPT, it wasn't really announced, it just came, right? So now They've taken. I won't say they're quite in the Google kind of line here because Google is notorious for talking about all of these new AI features, all of these large language models features and not really releasing any of them anytime soon or even giving dates. Right. So Google at their developer, their IO conference a couple months ago, they talked about their new project Astra which was this kind of live AI assistant. We haven't really seen much of anything since. We heard about their new Gems feature which is similar to ChatGPT being able to develop GPTs so smaller versions of the model, you can kind of train it a little bit, upload your own data. So we heard about Google Gemini's Gems, they never came, haven't really heard anything about it since. So unfortunately it looks like OpenAI is kind of going more down this Google route. Right? So now we have both the desktop app, we have the kind of the new voice features and now we have this search GPT. All of these things that have been announced, but we don't have them yet. So I do expect we should see some movement at least on the, the new text to speech capabilities with the new voice capabilities as well as the desktop app. I do expect to see that this week, if not next however search GPT, who knows when it might be coming out and are these things related, right? Some of these delays and then this new reports here from the information talking about a potential $5 billion loss. Y', all, I've been saying this all along to pay $20 a month for a chat GPT, it is a steal, right? Pretty sure they are losing money on every single subscription. I mean same thing with Claude, right? Even though I wish Claude as, as great as you know, Sonnet is and the new artifacts feature, it doesn't have access to the Internet so I don't use it as much as I like. But still the fact that you can get these things like Claude and open AI's Chat GPT and Perplexity for $20 a month, it is presumably costing these companies so much more. So if you are not using these every day, you are missing out. And hey, hopefully these companies figure out the whole, you know, turning a profit. But I think it is a long term play and I don't think necessarily these companies are worri about turning a profit even though like we said, reportedly, you know, OpenAI was valued at $80 billion. So that's a lot for today. So let's go ahead and quickly recap what we have going on. So here are the five I believe five stories we talked about today. So Meta just announced well about 6 days ago now their 3.1 updates to llama as well as for the first time releasing its large model 405B. All right, so they updated their 8B and 70B as well as released their 405B. Next Apple is reportedly delaying their Apple Intelligence features for another month, so we may not see it until the iOS 18.1 release in October. Then we had Morgan Stanley and Chase bank going all in on large language models. Each of them respectively releasing new AI and large language model features for their employees. Chase Bank One pretty exciting there with 50,000 employees reportedly getting access. Next OpenAI kind of teased at least I won't say they released Search GPT but they teased Search GPT which should challenge Google and Perplexity. And then last but not least, a new report from the information showed that OpenAI is facing a potential 5 billion dollar loss in 2024. All right, that's it y'. All. I hope this was helpful. Make sure if you haven't already go to your everydayai.com Sign up for the free daily newsletter as well as you gotta get into our Thanks a Million giveaway. I'm telling y', all, I don't think I actually put the link on our live stream yet. I'll drop it in here once I'm done. But you got to go in there, you got to sign up. We are giving away not only a free year of chat GPT to the winner or you know, if, if your thing is is clogged or perplexity, that's fine. So you get a year premium subscription to your large language model of choice. But we are going to be dripping out in the coming weeks some different prizes, things that you literally cannot buy. So if you haven't signed up yet, it takes about 30 seconds and then to climb the leaderboard. All you got to do is go in there and refer some of your friends. So if you find everyday AI helpful, if you're listening on the podcast, people reach out to me all the time, which I really appreciate. I always put you know, our email and my LinkedIn in the show notes if you're listening on the podcast. But you know, people are always like hey, how can I help? This is so great. I can't believe this is free. Well, this is one of those things. So please, if you haven't already, go sign up for the Thanks a Million giveaway contest, go tell your friends and make sure to join us tomorrow and every day for more Everyday AI Thanks, y'. All.

And that's a wrap for today's edition of Everyday. AI, thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going for a little more AI magic. Visit your everyday AI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.