The AI Daily Brief: Artificial Intelligence News and Analysis - Google’s Announcements Show 2 Key 2024 AI Trends

Episode Date: January 24, 2024

Google Chrome announces 3 AI features for Chrome as well as an innovative new Video generation tool Lumiere. Also SAP is restructuring 8000 jobs around AI. Interested in the EDU Beta? Learn more and sign up here: https://bit.ly/aibeta ABOUT THE AI BREAKDOWN The AI Breakdown helps you understand the most important news and discussions in AI.  Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown Join the community: bit.ly/aibreakdown Learn more: http://breakdown.network/

Today on the AI breakdown, we're looking at how Google's AI announcements reflect the two totally different AI trends happening this year. Before that on the brief, SAP restructures 8,000 jobs to better compete in AI. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.network for more information about our Discord, our YouTube, and our newsletter. Welcome back to the AI breakdown brief, all the AI headline news you need in around five minutes. Apologies for the ingested voice, but alas, AI news does not respect any sickness. We kick off today with yet another big business rethinking its whole structure in the face of the AI revolution. This week,

it is European enterprise software giant SAP, who's restructuring about 7% of their 108,000 jobs for a total of 8,000 jobs impacted as part of a larger shift to focus on artificial intelligence. So what What does restructuring mean in this? Is this just a euphemism for layoffs, or does it mean something different? SAP is a little bit circumspect about that. They said in a statement, the majority of the approximately 8,000 affected positions is expected to be covered by voluntary leave programs and internal reskilling measures. So it sounds to me like a combination of layoffs, although given that they're calling them voluntary leave programs, they're going to try to basically get people to agree to leave on their own, and some number of them at least are

going to be retrained and re-skilled for new AI-related roles. The firm is planning on spending $2.2 billion on the transition this year, and they said that once reinvestments are made in both reskilling as well as acquisitions, they expect to end this year at a headcount similar to current levels. In other words, although these 8,000 people may have their jobs changed or even eliminated, in aggregate, there won't be 8,000 jobs less at the end of the year. They'll just be different. CEO Christian Klein said in a statement, SAP is opening the next chapter. With the planned transformation program, we are intensifying the shift of investments to strategic growth areas, above all, business AI. Now, this is not SAP's first forays into the

AI world. Last summer, for example, it announced that it had invested in three different generative AI companies and pledged to invest more than a billion dollars to fund AI enterprise tech companies. In this sort of transition, they join, well, basically everyone who is trying to court investors by talking about how much they're going to spend on the AI transformation. For example, last July, WIPRO, one of India's biggest software service providers, said that it would spend a billion on improving its AI capacities as well. That one billion included retraining for its entire staff of 250,000 people. Now, SAP also on Tuesday reported some good news from an earnings perspective, forecasting a jump in revenue of 24 to 27% for its cloud business. And between these two things,

the market likes what it's hearing. SAP is up 7% on the day and 11% over the last week. Now, another big company that is in the news related to its AI initiatives is Apple. The Financial Times is reading some tea leaves and writes Apple boosts plan to bring generative AI to iPhones. Basically, the Financial Times is running down things like recent acquisitions, job postings, academic papers that Apple is publishing, and argues that, quote, Apple is quietly increasing its capabilities in AI, making a series of acquisitions, Acquisition staff hires in hardware updates that are designed to bring AI to its next generation of iPhones. The FT notes that a recent research report from Morgan Stanley identified that almost half of Apple's AI job postings

now include the term deep learning, which suggests that they're more focused on generative AI. And Wed Bush analyst Daniel Ives said, they're getting ready to do some significant M&A. I'd be shocked if they don't do a sizable AI deal this year because there's an AI arms race going on and Apple is not going to be on the outside looking in. Now, of course, it's an open secret as to what Apple is trying to. to do with AI. FT writes, Apple's goal appears to be operating generative AI through mobile devices, which would allow AI chatbots and apps to run on the phone's own hardware and software, rather than be powered by cloud services and data centers. That technological challenge requires

reduction in the size of the large language models that power AI as well as higher performance processors. This seems to be the be-all and the end-all of Apple's AI efforts. On the one hand, increase the performance power of the hardware itself, and at the same time, improve the performance of ever smaller LLMs. Luckily, there are lots and lots of projects, many of them open source, that are working on exactly that sort of shrinkage with some fairly positive results right now. Now, the other big bit of Apple news today was that Bloomberg reports that insiders are suggesting that its plans to release a self-driving car are being pushed back as well as being a little bit more limited in scope. Apple had initially wanted to have a fully autonomous car, what they

call a level four or level five automation, which would be able to drive everywhere and might not even have pedals or a steering wheel, but that their ambitions have now been downgraded to level two, which has things like lane centering and adaptive cruise control, as well as steering and brake acceleration support. Now, Apple's car plans have been a very ill-fated project in some ways inside the company. It's seen lots and lots of delays and lots of staff turnover, but at the same time, many think it represents one of the biggest opportunities for Apple. These latest reports suggest that instead of coming to the market by 2026, we're now looking at 2028 at the earliest. Jumping to another big tech giant, Microsoft has followed Apple to become just the second company

to hit a $3 trillion market cap. Bloomberg calls it the latest example of how optimism over artificial intelligence has fueled a seemingly unstoppable advance in the software giant. Now, part of what makes this important is that whenever we have a new technology, one of the big questions that the market ultimately comes to is will it actually make money for all the people investing in it? Or are they just investing in hype? One of the things that Microsoft is showing is that investing in AI actually impacts other segments of their business, including some of their most important segments like cloud computing and actually drives revenue. According to data from Bloomberg Intelligence, Microsoft revenue is expected to rise nearly 15% in 2024, which is faster than the overall

tech sector. Now, Microsoft, of course, has just opened up its co-pilot tools in Microsoft 365, that's its famous programs like Word, Excel, PowerPoint, etc. And whereas they used to be only available to enterprises with 300 or more licenses at a time, now anyone can pay for them. While some of those tools may be less sexy than the very cutting edge of other AI tools that you read about on Twitter or X, you've got to think that a heck of a lot more people are going to use their first AI inside of Word than because they saw something on product hunt. Now, speaking of a bellwether of how people are using AI, a new survey from Canva and Sago shows that 45% of job seekers report using AI to improve their resume and cover letter. What I think is interesting about this survey is that

it really feels like a leading indicator of where the world is headed to me. People transitioning jobs are at a moment where they have a lot of time to reflect around how they improve their prospects for the future. The fact that nearly half of them are reaching for these new generative AI tools, once again, suggests that they're not just hype being peddled upon us by media, but things that are actually useful right now in a day-to-day kind of way. Last story today, a new report from the Center for Non-Proliferation Studies in California shows some alarming developments around North Korea's AI program. Now, North Korea has obviously been targeted by sanctions basically forever,

and so there have been some limitations on the hardware that it's been able to access. And yet, this report suggests that they are making progress in a number of different areas of AI development. The report says, North Korea's recent endeavors in AI-ML developments signify a strategic investment to bolster its digital economy. For instance, North Korea's pursuit of a wargaming simulation program using machine learning reveals intentions to better comprehend operational environments against potential adversaries. So just when you thought you had enough to worry about with AI, now you have North Korea as well. However, that is going to do it for today's AI breakdown brief.

Next up, the main AI breakdown. Hello, friends, briefly before we dive into the main part of the episode, I'm excited to share that the AI education breakdown beta that I have been running for the last couple months is returning for February. This is an experiment in a different way to increase people's capacities to actually use artificial intelligence tools. Every day we drop new tutorials, case studies, and challenges that make a huge variety of different tools accessible and get you actually experimenting with AI and using these tools in minutes, not hours, and certainly not days or weeks. What's more, because we've been running this now for a couple months, we've got a library

of over 50 videos and challenges that you can do starting right away. On top of all that, there is an incredibly supportive community of people from all walks of life, from professionals to creators, to CEOs, to small business owners, who are all learning these tools at the same time, sharing their creations, answering each other's questions, and generally just being incredibly passionate and excited. Now, this is a paid experience. It's 20. $20 a month, and the reason for that is twofold. One, I want you guys to critique this on the basis of having actually paid for it and give me feedback that way.

And two, I really want folks who are super motivated and have their own skin in the game when it comes to this community that we're building. I'm so excited to welcome you guys to be a part of this. To sign up and learn more, go to bit.ly slash AI Beta. That's bit.l.ly slash AI Beta. registration ends on Thursday at midnight East Coast time. Appreciate you all. Now on with the show.

Welcome back to the AI breakdown. You might have recently heard me do an episode about the four battles of the AI Stack War. It was a concept that was coined by my friends over at Leighton Space, and I think that one could expand it even further and use this idea of the various sub-battles to help understand and describe the industry. Broadly speaking, when I look at where the field is right now, I see companies running in two very different directions. On the one hand, there is, of course, the never-ending pursuit of the state-of-the-art, more advanced models, be they ever-bigger models, or more performant

models at smaller sizes. Whatever it is, the idea is that some number of companies or some parts of companies are focused on pushing the boundaries of what is actually possible. On the other hand, probably the most used word on this show right now is integration. And the other big trend is much more of a consumer-facing trend, a product trend, which is AI coming in to the products and workflows that consumers actually use. This is an adoption trend, of course. It's making AI accessible and useful in a day-to-day way. Now, obviously, in some cases, these two trends converge, but what was interesting about the news coming out of Google today is that there was really two totally different sets that represented each end of the spectrum, which I guess makes sense given just the sheer size

of Google and the importance of the AI effort to it. Remember in Siddharpichai's 2024 goal note, he put AI-related goals as the number one objective of the company. So let's talk about the integration side first. Google Chrome has announced that it has three new integrated generative AI features that are either there or coming soon. The first is something that I think will be familiar to all sorts of deep browsers, which is the problem of having too many tabs open. I'm sure many of you can relate to that point at which you can't even see what each tab is because there's so little space for letters to describe it up top. Well, now Chrome will have a feature where you can right-click on a tab and select organize similar tabs. Chrome will try to bring together all of the tabs that are

related, as well as suggesting names and even emojis for those groups. So, for example, if you are on the one hand researching a particular topic for work, but on the other, other looking at flights for your vacation, theoretically, Chrome is going to be smart enough to be able to know which of those is which, and organize them together. The second AI feature coming to Chrome is a personalization feature. This is an extension of a feature that had previously been released on Android and Pixel phones. It's basically a generative AI wallpaper that allows you to personalize the mood and feel of your browser. Now, beyond just this being neat for people who want a custom background on their browser, one thing that I also found notable was the way in which

they're handling the prompting, specifically the fact that they're creating a user experience that isn't just a blinking line. Instead, you're not only going to have a topic, but you're going to be able to add optional details that include style and mood. I think we're seeing this sort of thing more broadly, where to some extent we're moving away from a pure natural language input to a combination of natural language plus presets that work for the particular application. The last AI feature of Chrome is one that makes a ton of sense and is something that, frankly, a lot of browser extensions we're already doing, but will now be more natively integrated into the browser. This is, of course, help with writing. So imagine as you are going around the internet just how many

times you have to write, leaving a review on a website, replying for an invite to a party, making an inquiry about something you want to buy or rent, to say, of course, nothing of email and other forms of creative writing. Well, now, with this new feature, you'll be able to right-click in text box or field of any site that you visit in Chrome and select the Help Me Write option, Prompt it just like you would any LLM, and this feature will start writing. Similarly to what we discussed, there are preset toggles to modify it for length and tone, again showing how we're moving to hybrid interfaces that include both natural language input as well as presets. So, as I've said frequently, this isn't revolutionary, it's just functional.

A lot of people are going to use these types of AIs much more than more out-there things or more state-of-the-art things, because these just live where they're spending their time and where they're doing things. But as I mentioned, that was not the only Google AI announcement this week. Yesterday, Google also announced a new video creation AI that they're calling Lumiere. So they call this model a space-time diffusion model for realistic video generation. It includes text-to-video features, image-to-video features, video stylization, and in-painting. The in-painting may be the thing that people are most excited about, so we'll come back to that in a minute. Now, when it comes to how this model differs from previous video generation models,

Here's the way the team describes it in the research paper abstract. They write, We introduce Lumiere, a text-to-video diffusion model designed for synthesizing videos that portray realistic, diverse, and coherent motion, a pivotal challenge in video synthesis. To this end, we introduce a space-time unit architecture that generates the entire temporal duration of the video at once through a single pass in the model.

This is in contrast to existing video models, which synthesize distant keyframes followed by temporal super-resolution, an approach that inherently makes global temporal consistency difficult to achieve. By deploying both spatial and importantly temporal down and up sampling, and leveraging a pre-trained text-to-image diffusion model, our model learns to directly generate a full frame rate low-resolution video by processing it in multiple space-time scales. We demonstrate state-of-the-art text-to-video generation results

and show that our design easily facilitates a wide range of content creation tasks and video editing applications, including image-to-video, video in-painting, and stylized generation. Now, if you are saying what to all of that, don't worry about it. I think that the most important thing for the lay perspective is just that this is not just a repeat of other types of video generation models, but is actually a fundamentally different approach. Now, for those of you who are watching all of the images on the screen were generated with Lumiere. You can hover over and see the prompt that created them, such as a cute mouse typing on a keyboard.

They also have examples of image to video, stylized generation, where a video is created around a particular style reference image, Cinemographs where just a specific part of an image is animated but not the whole thing, and video in painting which allows creators to change particular parts of a given video. For example, they show a source video of a young woman in a hat and a spotted dress turning around in a field of flowers and then change that dress in various ways, to a golden strapless gown, a striped strapless dress, a purple dress, a black dress, and more. Now, a couple things to note about this. First of all, it looks super impressive, but this isn't something that you

you can actually try quite yet. So of course, ultimately, we don't really know how it stands up in practice. Now, one thing people did notice is that this wasn't the only Google video model that they've been discussing, and recently, too. Theoretically media tweets, Lumiere looks really impressive, but what is going on at Google right now? Didn't they just in late December announced Google Poet, which was another video model? Did Poet get rolled into Lumiere? Or is this a case where the left hand has no idea what the right hand is doing? Sure enough, video poet is another LLM for Zero Shot video generation that Google research has recently been discussing. So far as I can tell, Video Poet and Lumiere are two totally separate efforts. Indeed, comparing

the authors of these two papers, there doesn't seem to be any overlap. Now, on the one hand, from the outside in, this certainly does make one wonder about whether Google's efforts are too divided, but at the same time, I think that to some extent, it also shows just how significant video generation is to the state of the art when it comes to generative AI right now. This is an area that's so important that Google apparently has multiple teams racing at it, and deploying what are ultimately different strategies to get there. I'm pretty excited for the rapid advancement in this area. I think it's going to open up a ton of creativity.

And so for now, I will take all the efforts we can get. That, however, is going to do it for today's AI breakdown. Thanks, as always, for listening or watching. And until next time, peace.