Tech Brew Ride Home - Wed. 05/10 – Google I/O

Episode Date: May 10, 2023

I’m writing this this morning before Google’s I/O, so I’m assuming that’s going to be the entirety of this episode. But who knows. Maybe I’ll be able to squeeze something non-Google stuff at the end. Join me in the time machine, won’t you, and find out. In the meantime, here’s what Sundar had to share with us today. Sponsors: Leadership.OregonState.edu/cic Grammarly.com/go Links: Google Photos to gain a new ‘Magic Editor’ feature powered by generative AI (TechCrunch) Google launches PaLM 2, its next-gen large language model (TechCrunch) Google rebrands AI tools for Docs and Gmail as Duet AI — its answer to Microsoft’s Copilot (The Verge) The AI takeover of Google Search starts now (The Verge) Review: Pixel 7a affordably delivers on every promise of the Pixel series (9to5Google) Pixel Tablet Hands-On: Google's Return of the Tablet and More (CNET) Google Pixel Fold hands-on: Finally, a real rival for Samsung’s foldables (Engadget) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the Tech meme right home for Wednesday, May 10th, 2023. I'm Brian McCullough today. I'm writing this this morning before Google's I.O. kicks off. So I'm assuming this whole episode is going to be the entirety of that event. But who knows? Maybe I'll be able to squeeze something non-Google in at the end. Join me in the time machine, won't you, and find out. In the meantime, here's what Sundar had to share with us today. I'm going to do this a bit differently today instead of just vomiting all the announces in one segment. I'm I'm going to try to break them up a bit thematically to hopefully make it a bit easier to follow, but also because this is the whole show, it turns out.

So I can't really do that other thing. Google kicked off I.O. this morning by announcing some new little AI-powered consumer-facing tools that were fun. For example, immersive view for routes in Google Maps with traffic simulations, parking, and more, starting in 15 cities, including New York City, London and Paris. So if you want to plan out a quick walk or bike ride through the west side of Manhattan, say you can literally zoom through it a la Superman to fully simulate what it will be like. Seems like another placeholding product that would be really powerful and useful in some AR or XR future. And Google Photos got a magic editor feature for more complex edits than Magic Eraser can achieve, like repositioning the subject of a photo, first coming to, select pixel devices. Here's the example they used on stage to give you a sense of how this works.

Picture a kid sitting on a bench holding balloons, but the picture isn't centered. The kid is on the edge of the bench off to the side, and some of the balloons are actually outside the frame of the picture. No worries. Just pull the kid over to the center of the bench on your device, and AI generates the balloons that got cropped out, and voila, it's a perfectly centered photo that includes content you never actually captured. Gmail is getting a cool, Help Me Write feature. Here's the example they used again on stage. You want a refund for the flight that got canceled over the weekend. You can just ask Gmail to write that request for a refund for you. It does. It does a really convincing and conventional email to the customer service email,

outlining what happened, what you want. It's like autocomplete, but for a whole email, not just a sentence. But even if what it generates is not convincing enough, you think, or long enough, or maybe vitriolic enough, you can say, make it, more persuasive or make it more impactful. They have actual pull-down menus for this. Make it longer. What have I been saying about knobs and dials? Then it was on to the language models. Say hello to Palm 2, what Google called their most capable large language model yet, with improved multilingual, reasoning and coding capabilities. They say it's multimodal. And is available in four different sizes. the smallest size can be run on a mobile device, and it is apparently already powering 25 Google products.

Google said Palm 2 will power its updated Bard Chat tool, which is now available to developers through Google's Palm API, Firebase, and CoLab. Also, the wait list for Bard is gone, so you can have at it, quoting TechCrunch. Google, similar to OpenAI, didn't provide a lot of the technical details about how it trained this next-gen model, including parameter counts. Palm 2 is a 540 billion parameter model for what it's worth. The only technical details Google provided are that Palm 2 was built on top of Google's latest jacks and TPUV4 infrastructure. Google says the new model is better at common sense reasoning, mathematics, and logic. Indeed, as was noted on stage, the company trained the model on a large amount of math and science texts, as well as mathematical expressions. It's no secret that large language models with their focus on language have

struggled with handling math and math questions without resorting to third-party plugins. Google, however, argues that Palm 2 can easily solve math puzzles, reason through problems, and even provide diagrams. Palm 2 also now features improved support for writing and debugging code. The model was trained on 20 programming languages, including popular ones like JavaScript and Python, but also the likes of Prolog, Verilog, and Fortran. Palm 2 forms the basis of Cody, Google's specialized model for coding and debugging, which it is also launching today as part of its code completion and generation service, among other things. Google Today also highlighted that Palm 2 was trained on a corpus that features over 100 languages

making it, in Google's words, Excel at multilingual tasks, including more nuanced phrasing than previous models, end quote. Google said Bard is available in English, in 180 countries and territories promising AI image generation with Adobe and integration with services like Instacart, end quote. Google also rebranded AI tools. for its productivity apps as duet AI. Also teased something called Sidekick, which will read, summarize, and answer questions on documents across its apps. Quoting the verge, sidekick, which Google described at I.O. as the future of collaboration with AI, though with no date on when it might be available to users, is in appearance like a slide panel that's able to analyze the document you're looking at.

Google promises that it will not only be able to answer queries about content, but also suggest its own. For example, if you're writing a story, it might suggest that you generate some images to illustrate it. In another example, the user is writing an email about an upcoming potluck. Not only does Sidekick analyze the contents of the email chain, but it also looks at linked documents, including, in this case, a list of dishes being served, and is able to suggest a good accompaniment, end quote. They unveiled features to help identify AI images, or at least images generated by AI, including an about-this image feature and a markup that will allow, images to be labeled as AI generated. Google Cloud announced new A3 VMs powered by NVIDIA's H100 GPUs, purpose built for complex machine learning models and LLMs running generative AI applications. And they announced a competitor to GitHub's co-pilot, a chat tool

for questions about coding and AI-assisted coding and its no-code app sheet product. But I think you get the idea of what that's all about. So moving on. So AI across all of Google's productivity apps, is now duet, AI, and then the dev tools got AI, cloud stuff got AI, super. What about AI in search? Google demoed AI features coming to search, including an AI snapshot, which will provide a summary of search results with links that corroborate the information you've searched for. The traditional 10 blue links aren't gone exactly, they're just getting shoved further down the page, quoting the verge. To demonstrate, Liz Reed, Google's VP of Search, flips open her laptop and starts typing into the Google search box. Why is sourdough bread still so popular,

she writes, and hits Enter? Google's normal search results slowed almost immediately. Above them, a rectangular orange section pulses and glows and shows the phrase, generative AI, is experimental. A few seconds later, the glowing is replaced by an AI-generated summary, a few paragraphs detailing how good sourdough tastes, the upsides of its prebiotic abilities, and more. To the right, There are three links to sites with information that Reed says corroborates what's in the summary. Google calls this the AI snapshot. All of it is by Google's large language models. All of it sourced from the open web.

Reed then mouses up to the top right of the box and clicks an icon. Google's designers called the Bearclaw, which looks like a hamburger menu with a vertical line to the left. The Bearclaw opens a new view. The AI snapshot is now split sentence by sentence with links underneath to the sources of the information for that specific sentence. This read points out, again, is corroboration. And she says it's key to the way Google's AI implementation is different. We want the LLM when it says something to tell us as part of its goal, what are some sources to read more about that, end quote. A few seconds later, Reid clicks back and starts another search. This time, she searches for the best Bluetooth speakers for the beach.

Again, standard search results appear almost immediately, and again, AI results are generated a few seconds later. This time, there's a short summary at the top detailing what you should care about in such a speaker. Battery life, water resistance, sound quality, etc. links to three buying guides sit off to the right and below our shopping links for a half a dozen good options, each with an AI-generated summary next to it. I ask Reid to follow up with the phrase under $100, and she does so. The snapshot regenerates with new summaries and new picks. This is the new look of Google's search results page. It's AI first. It's colorful. It's nothing like you're used to. It's powered by some of Google's most advanced LLM work to date, including a new general purpose

model called Palm 2 and the multitask unified model, or MUM, I guess, that Google uses to understand multiple types of media. In the demos I saw, it's often extremely impressive, and it changes the way you'll experience search, especially on mobile, where that AI snapshot often eats up the entire first page of your results. There are some caveats to get access to these AI snapshots. you'll have to opt in to a new feature called search generative experience or SGE for short, part of an also new feature called search labs.

Not all searches will spark an AI answer. The AI only appears when Google's algorithms think it's more useful than standard results, and sensitive subjects like health and finances are currently set to avoid AI interference altogether, end quote. Google also plans to add a section called perspectives to search results, with information from individual users. So, you know, stuff featuring YouTube videos, Reddit comments, and more. Then it was on to Android. Google has updated its Find My Device Network,

including proactive alerts about unknown trackers may be traveling with you. And that will also support Apple's air tags and other air tagging devices. Remember that industry standard these companies announced, was it this week? Yeah, this is that. Also, generative AI wallpapers for pixel phones and magic compose, which lets you reply to texts with auto-suggested responses for Android's default messages app. If we're onto Android, that must mean. We're also onto hardware. Google announced the Pixel 7A with a 6.1 inch screen, face unlock, and a 64-mixel camera available tomorrow for $499. As ever with these phones, that's quite the deal. quoting 9 to 5 Google to tell you why.

Fresh out of the box, the Pixel 7A makes just as sharp of a first impression as its flagship sibling. Google's now signature camera bar design has returned, and it doesn't stick out as far here as it does on some other recent pixels. The aluminum alloy frame and glass exterior are available in four colorways, of which our team was able to test two, see blue and clearly white. The bluer shade is far more subtle than its name may suggest to the point of being an off-white. It's the tint of the aluminum alloy that pulls the color scheme together. The coloration also continues into the USBC port, a detail I appreciated. On that note, the 4,385-mph hour battery has been more than enough to get through an entire day of my typical use. You'll probably get even better battery life than that at first because Google has set the display to 60 hertz by default.

However, in another first for affordable pixels, the 7A can be set to have a 90-hertz refresh rate or smooth, display, which helps make supported apps look and feel slick. I'm honestly shocked at how Google was able to pack so many nice to have features into the Pixel 7A while maintaining an affordable price tag. Nothing in the build feels like it's obviously lacking in comparison to the rest of the phone. By far, the biggest improvement of the Pixel 7A over generations past is the upgraded camera sensor. Technically, the Pixel 7A features a camera with the highest megapixel count to be featured in a phone from Google beating out even the Pixel 7 Pro. 64 megapixels versus 50 megapixels. Of course, that raw pixel count doesn't mean much. For one,

Google camera bins your photos down 4X to around 16 megapixels to enable better detail. The more important spec for quality and detail is the sensor size, with the pixel 7 and 7A being evenly matched with one by 1.3 inch sensors. In practice, this new sensor paired with the excellent as ever, Google camera enhancements and the Lightning Fast Tensor G2 can take exceptional photos in a flash, No pun intended. The Google Pixel 7A is all the phone you really need. It's premium design, near-flagship level specs, and exceptional camera quality all add up to far more real-world value than the $499 asking price. In the time that I've used the Pixel 7A, I haven't once missed my

Pixel 7 Pro, and that's saying something, end quote. Now, also on hardware, Google also announced the pixel tablet with a tensor G2 processor, an 11-inch display, megapixel front-facing camera and a speaker hub, available in June for $499.99, but I'm going to allide over that for now for a few reasons, first for time, of course. But also, have Android tablets ever really been a thing? I mean, Google's still trying, but sorry Google fans. I think we can alight over this, though I will include a link in the show notes to a full description from CNET. The third reason to move on is to talk about the star of the show. Google announced the $1,799 pixel fold with a 120-hirts 7.6-inch main screen and a 5.8-inch exterior screen,

Tenser G2 chip and 12 gigabytes of RAM. Pre-orders begin now, and they will arrive in June, quoting a gadget. Google's foldable phone features a different design than most. It has a wider 5.8-inch exterior screen with a 17.4 by 9 aspect ratio. This makes a huge difference, Giving you a more usable screen on the outside allows the pixel fold to function like a more full-feature device before you ever open it up. In comparison, the Galaxy Z Fold4's small cover screen feels like it's designed to mainly do simple tasks like looking up directions or responding to text. And in your hand, the pixel folds wider dimensions make it feel different too. It's shaped more like a passport than Samsung's skinnier baton-like handset. On top of that, the pixel

of folds heft is somewhat deceptive. Weighing 10 ounces or 283 grams, it's actually heavier than the Galaxy Z-Fold 4 at 263 grams. However, because of its super slim design, that measures less than 6 millimeters thick when open or 12.1 millimeters when closed, it doesn't feel nearly as dense. Then, on the outside, Google created what it claims is the most durable hinge on any foldable yet. But to me, the clever thing is that by pushing all the moving parts as close to the outside of the device as possible, there's less of a gap between.

between the two halves of the phone, resulting in an almost completely flat device when closed. As for its flexible main screen, we're looking at a big 7.6 inch 2208 by 1840 OLED display with bright, vivid colors. Now, yes, there is a crease, and I can't wait for the day when device makers can banish them for good, but at least on the pixel fold, it looks slightly less pronounced compared to some other devices. That said, the bigger differences are that Google's foldable features slightly larger bezels and a wider landscape orientation by default, which is better for watching videos without rotating the phone. And instead of an under-display sensor like Samsung uses, Google has opted for a more traditional interior camera located on the top right

corner. Aside from its design, where things get really interesting is the pixel folds software and multitasking features. Even though I've only had a few minutes with it thus far, I love how snappy and responsive it is. Switching from the front to the interior screen, is super smooth. And I must admit the way Google syncs the movement of birds' wings on the wallpaper with the position of the screen is a really nice touch. It even seems like Google has learned from Samsung a bit, like with the inclusion of a disappearing taskbar that you can summon by swiping up from the bottom of the screen. Google also says more than 50 first-party apps have already been optimized to run on devices with large screens like the pixel fold. You can see this in apps like

YouTube, which automatically switches to a special UI when the phone is bent in half. And while Samsung offers similar tweaks in apps on the Z-Fold line. It's really nice to see continued development for adaptive UIs as they are going to play a huge part in making the next generation of foldable devices really shine. Google even showed off a new interpreter mode that uses both the inside and the outside screen at the same time, which works with the phone's speech recognition so that people can see two different languages get translated in real time. Moving on to photos, despite the challenge of finding room for camera sensors and a phone this thin, you still get essentially the same great experience that we saw on the Pixel 7 Pro, including that 5x telephoto zoom. That's really nice to see

compared with the ZFold 4, which doesn't have nearly as good sensors and lenses as the S-22 Ultra. And just like Samsung, Google included some nifty modes that allow you to shoot photos when the phone is half open or use the rear camera on the front screen at the same time to capture super-detailed selfies. Sure, these use cases are a bit niche, but they're really nice for taking group shots without needing to carry a little mobile tripod around. Finally, rounding out the pixel fold specs are a list of very premium components. We're talking a Tensor G2 chip, 12 gigabytes of RAM, up to 512 gigabytes of storage, and 120 hertz refresh rates on both screens.

You also get a side-mounted fingerprint sensor built into its power button, support for wireless charging and IPX-8 water resistance, which remains a real rarity among foldable phones. Unfortunately, the pixel fold also supports a very premium price of $1,199. I was really hoping to see the costs of big foldable phones come down a bit now that they've been around for multiple generations. But considering that's basically the same cost as a Galaxy Z Fold4, it's hard to be too mad. Long-term durability is another concern. It took Samsung's several generations to create a foldable phone that can really survive daily use,

and even now things aren't perfect. And while I'm sure Google has learned a lot from Samsung's mistakes, only time will tell how the first foldable pixel will hold up. Now, if you'll allow me to throw one last comparison into the mix, in a lot of ways, I feel like the pixel fold is an even more refined version of what I wanted the Surface Duo to be. It's got a similar shape, thanks to its sleek design, but with an extra display on the outside and way better software. As a longtime user of big foldable phones, I can't wait to test this thing out for real, sometime later the summer, end quote. It was actually useful that I.O. was today because I had a corporate speaking gig in New Jersey this morning. Thanks Dan for putting that together.

And hello, anyone, any new listeners that came in because they saw my speech this morning, guess what the topic was? The history of AI tech and where I think we are at this moment in the AI revolution. So anyone out there want me to talk to your team or your department or your whole company about a similar thing about AI? I now have a whole deck and speech outline that I can do for you, get in touch if interested. And also, again, I'm very serious. Any devs out there working in AI? If you want a quick side project building me an AI resume generator, get in touch.

I'll either pay you a flat fee or I'll share the revenue with you so it could be earning you and me. Money for years to come. Brian at Techmeme.com. If interested, talk to you tomorrow.