Tech Brew Ride Home - Fri. 12/13 – Gemini As The New Android/OS?

Episode Date: December 13, 2024

Ok, what if AI models are really the new OS for… everything. But also: smartglasses and a Vision Pro competitor. I’ll explain. Could NotebookLM become a fully fledged product? Seemingly the leader in the clubhouse in the race to create the next AI paradigm beyond the transformer model. And, of course, the Weekend Longreads Suggestions. Links: I saw Google’s plan to put Android on your face (The Verge) Google’s NotebookLM AI podcast hosts can now talk to you, too (The Verge) Google says its breakthrough quantum chip can’t break modern cryptography (The Verge) Liquid Set to Raise $250 Million to Build AI Inspired by Tiny Worm Brains (Bloomberg) Weekend Longreads Suggestions: The Cult of Claude (NYTimes) How WhatsApp ate the world (Rest Of World) 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 Friday, December 13th, 2024. I'm Brian McCullough today. Okay, what if AI models are really the new OS for everything, but also smart glasses and a Vision pro competitor, I'll explain. Could Notebook

L.M become a fully fledged product, seemingly the leader in the clubhouse and the race to create the next AI paradigm beyond the transformer model? And of course, the weekend long-reach suggestions. Here's what you miss today in the world of tech. Something, something new product category enabled by AI, Google yesterday unveiled Android XR, a mixed reality OS for headsets and smart glasses with Gemini built in, and plans a 2025 debut for this with a piece of hardware from Samsung called Project Mujan, which is a headset. This is Victoria's hands-on report of all of that, quoting the verge. It's an ordinary Tuesday. I'm wearing what look like ordinary glasses and room surrounded by Google and Samsung representatives. One of them steps in front of me and starts speaking

in Spanish. I don't speak Spanish. Hovering in mid-air, I can see her words being translated into English subtitles. Reading them, I can see she's describing what I'm seeing in real time. I mumble an expletive. Everyone laughs. This is my first experience with Android XR, a new mixed reality OS designed for headsets and smart glasses like the prototypes I'm wearing. It's Google's big bet to power a new generation of augmented reality devices that embody our wildest dreams of what smart glasses can be. Google wants everyone to know the time is finally right for XR, and it's pointing to Gemini as its North Star. Adding Gemini enables multimodal AI and natural language, things it says will make interactions with your environment richer. In a demo, Google had me

prompt Gemini to name the title of a yellow book, sitting behind me on a shelf. I briefly glanced at it earlier but hadn't taken a photo. Gemini took a second and then offered up an answer. I whipped around to check. It was correct. On top of that, the platform will work with any mobile and tablet app from the Play Store out of the box. Today's launch is aimed at developers so they can start building out experiences. The average person won't be able to buy anything running Android XR right now, but in 2025, Samsung will be launching its long-remered XR headset, dubbed Project Mujan, Korean for infinity, the headset will be the first consumer product to ship with Android XR. Technically, it's running the same software as the glasses I tried, but Project

Mujan will also be capable of VR and immersive content, stuff that wouldn't be suited to a pair of smart glasses. It's essentially a showcase for everything that could be possible, hence why Google is going with XR, a catch-all term that stands for extended reality and encompasses AR, VR, and mixed reality. Trying on Mujan, it seems a bit like the Vision Pro, but unlike that, the light seal is optional, so you can choose to let the world bleed in. It's lightweight and doesn't pinch my face too tightly.

My ponytail easily slats through the top, and later I'm thankful that I don't have to redo my hair. At first, the resolution doesn't feel quite as sharp as the Vision Pro until the headset automatically calibrates to my pupillary distance. It's at this point when I start feeling deja vu. I'm walked through pinching to select items and how to tap the side to bring up the app launcher. There's an eye calibration process that feels awfully similar to the Vision Pros. If I want, I can retreat into an immersive mode to watch YouTube and Google TV on a distant mountain.

I can open apps, resize them, and place them at various points around the room. I've done this all before. This just happens to be Google-flavored. I want to ask, how do you expect to stand out? I don't get the chance to answer before I'm told, Gemini. the Mujan headset, I can say, take me to Y-YP Entertainment in Seoul, and it will automatically open Google Maps and show me that building. If my windows get cluttered, I can ask it to reorganize them. I don't have to lift a finger while wearing the prototype glasses. I watch and listen as

Gemini summarizes a long rambling text message. To the main point, can you buy lemon, ginger, and olive oil from the store? I was able to naturally switch from speaking in English to asking in Japanese what the weather is in New York and get the answer spoken and written in Japanese. It's not just interactions with Gemini that linger in my mind either. It's how experiences can be built on top of them. I asked Gemini how to get somewhere and saw turn-by-turn directions. When I looked down, the text morphed into a zoomable map of my surroundings. It's very easy to imagine myself using something like that in real life. But as cool as all that is, headsets can be a hard sell to the average person. Personally, I'm more enamored with the glasses down.

demo, but those have no concrete timeline. Google made the prototypes, but it's focusing on working with other partners to bring hardware to market. There are still cultural cues that have to be established with either form factor. Outside of Gemini, there has to be a robust ecosystem of apps and experiences for the average person, not just early adopters. It's not going to be a singular product. It's Android, says Shiram Izadi, Google's VP of AR and XR, noting that Google has a three-pronged strategy for Android XR. Laying the groundwork with devs is one. element. Gemini's conversational experience is another, and the third is the idea that no one device is the future of XR. Head sets, for example, may just be episodic devices you use for entertainment.

Glasses could supplement phones and smartwatches for discrete notifications and looking up information. The way I see it, these devices don't replace one another. You'll use these devices throughout your day, and if there's consistency with Gemini and generative AI experiences across these form factors, people will get more comfortable with wearing computers on their faces. That's the on-ramp to get more immersive devices, says Inzati, end quote. So a Vision Pro-like device coming soon from Samsung, but hints that all of this can be baked into anything, lighter things as well. Mark Gurman also had a hands-on with Project Mujan, and he says it was lighter and more comfortable to wear than Apple's Vision Pro. And Android XR is highly reminiscent of Vision

OS, but he would say that, wouldn't he? Google has rolled out Notebook LM plus, an enterprise version of Notebook LM with added controls for access and data management, a bit more on that in a little bit. But hey, that little novelty experiment, maybe it could become some sort of a real product, no, but also what more could it be? Because Google also announced interactive audio overviews in Notebook LM, an experimental feature that lets users talk to the AI hosts of the overviews, those AI, I guess, podcast that they create. Quoting the Verge, the ability to actually talk with Notebook LM seems like a potentially useful way to learn more about what you've collected in the app.

But Google cautions that it's an experimental feature and that hosts may also pause awkwardly before responding or occasionally introduce inaccuracies, so it may not be a totally polished experience to start, in addition to the interactive. audio overviews. Google is introducing a new interface for Notebook LM that organizes things into three areas, a sources panel for your information, a chat panel to talk with an AI chatbot about the sources, and a studio panel that lets you make things like audio overviews and study guides. I think it looks nice. Google is announcing a Notebook LM subscription to Notebook LM Plus. The subscription will give you five times more audio overviews, notebooks, and sources per notebook, let you customize the style and tone of your notebook responses, let you make shared team notebooks, let you offer additional privacy and

security, Google says. The subscription is available today for businesses, schools, and universities, and organizations and enterprise customers. It will be added to Google One AI premium in early 2025. Google is also launching Agents Space, a platform for custom AI agents for enterprises. Agent Space can provide conversational assistance, answer complex questions, make proactive suggestions and take actions based on your company's unique information, Google says. It also has connectors for apps like Microsoft SharePoint, Jira, and Service Now, end quote. Let's keep going with Google because remember that reported breakthrough in quantum computing from this week? What's the biggest fear of quantum computing actually coming to reality that we've discussed in the

past? Quoting the verge. Experts believe that one day quantum computers could make today's systems of encryption utterly obsolete, but Google tells the verge its new breakthrough, Willowchip, is nowhere near ready for that. The Willow chip is not capable of breaking modern cryptography, Google Quantum AI Director and C-O-O-Chiro-Telts, Charina Chow, tells the verge, a so-called crypto-aniletically relevant quantum computer, or CROQ, crock, could, quote, jeopardize civilian and military communications, undermine supervisory and control systems for critical infrastructure, and defeat security protocols for most internet-based financial transactions, the White House warned in 2022, ordering that U.S. agencies must transition to new systems to mitigate

that risk by 2035. But Willow is not a crock, according to Google. While the company does claim it can solve a computing challenge in five minutes that would take the world's fastest supercomputer 10 septillion years, Google has only produced 105 physical cubits worth of that computing power and suggest they would need millions to literally crack the codes. Estimates are that we're at least 10 years out from breaking RSA, and that around 4 million physical qubits would be required to do this, Chow writes. She says Willow doesn't change that timeline at all. And though Chinese researchers have repeatedly claimed to discover new ways to break RSA encryption with much smaller quantum computers, ones with just a few hundreds or thousands of cubits,

security experts have repeatedly been skeptical. Google is one of the many companies' to defend against the potential threat of broken encryption with post-quantum cryptography or PQC, ever since the Edward Snowden leaks revealed that spy agencies like the NSA were quietly funding code-breaking quantum computer research. A few years back, we wrote about how the National Institute of Standards and Technology created a competition to develop quantum-safe cryptography standards back in 2016. This August, NIST released three finalized algorithms and its standards for integrating them into products and systems and plans to select one or two more by the end of the year. The Rand Corporation, a think tank famous for advising on U.S. national security in the past,

suggested in 2023 in an editorial that the moment in RSA-breaking quantum computer exists, it'll trigger a worldwide rush to defend against it. As soon as the existence of the crock becomes public knowledge or is even considered plausible, and the threat becomes concrete, most vulnerable organizations will immediately move to upgrade all their communication systems to post-quantum cryptography, end quote. From the, is there something beyond the transformer model file, Liquid AI, which is building AI systems powered by liquid neural networks, raised a $250 million series A led by AMD at a $2.3 billion valuation following a $46.6 million seed. Liquid neural networks. What's this now?

Quoting Bloomberg. Founded in 2023 by a group of AI researchers at the Massachusetts Institute of Technology, the startup is building a novel technology called Liquid Foundation Models that instead uses mathematical techniques discovered by studying the architecture of a worm's brain. By focusing on an organism, specifically a small nematode worm called, I'm not going to pronounce that, that's just one millimeter long with far fewer neurons than the human brain. Liquid AI said it's able to build an AI system. that is more flexible. Its systems also require less data and computing power than the conventional transformer-based models that underpin chatbots and other popular AI tools, the company said.

Two years after ChatGPT's release kicked off a frenzy around generative AI, there's a growing debate about whether top developers are hitting a wall in building more advanced artificial intelligence systems that justify the tremendous cost. A key part of Liquid AI's pitch is effectively that its technology can do more with less. The startup said its models are better suited to run on devices, reducing the need for costly data centers. Liquid also claims a text-based version of its AI system outperforms some leading alternatives of similar size on various benchmarks. Liquid AI has quickly attracted attention from academia and business leaders. Of all the companies trying the non-transformer architectures, I believe Liquid AI is the frontrunner. Shopify chief

technology officer Mikhail Parakin wrote on X this summer. Perplexity CEO Aravinds Srinavos also posted about the promise of the model, end quote. Time for the weekend long-reach suggestions. First up, Kevin Ruse outlines something that I've seen personally as well. Sure, the masses use chat GPT, but Anthropics Claude has become the chatbot of choice for AI industry insiders who say its responses feel more creative and empathetic and less gradingly robotic. Quoting the times, to the people who love it, Claude just feels different. more creative and empathetic, less gradingly robotic. Its outputs, they say, are like the responses a smart attentive human would give and less like the generic prose generated by other chatbots.

As a result, Claude is quickly becoming a social sidekick for AI insiders, and maybe a preview of what's coming for the rest of us, as powerful synthetic characters become more enmeshed in our daily lives. More and more of my friends are using Claude for emotional processing and thinking through relationship challenges, said Jeffrey Laddish in AI safety researcher at Palisade Research, asked what made Claude different than other chatbots. Mr. Ladish said that Claude seemed more insightful and good at helping people spot patterns and blind spots. The analogy I use is a highly liked, respected traveler, said Dr. Amanda Askell, a researcher and philosopher at Anthropic who is in charge of fine-tuning Claude's character.

Claude is interacting with lots of different people around the world and has to do so without pandering and adopting the values of the person it's talking with. A problem with many AI models, Dr. Askell said, is that they tend to act synchophantic, telling users what they want to hear and rarely challenging them or pushing back on their ideas, even when those ideas are wrong or potentially harmful. With Claude, she said, the goal was to create an AI character that would be helpful with most requests, but would also challenge users when necessary. What is the kind of person you can disagree with, but you come away thinking, this is a good

person, she said. These are the sorts of traits we want Claude. to have. Claude is still miles behind chat GPT when it comes to mainstream awareness. It lacks features found in other chatbots, such as a voice chat mode and the ability to generate images or search the internet for up-to-date information. And some AI makers speculate that Claude's popularity is a passing fad or that it's only popular among AI hipsters who want to brag about the obscure chatbot they're into. But given how many things that start in San Francisco eventually spread to the rest of the world, Claude's warm embrace could also be a preview of things to come, end quote.

And from rest of world, a reminder to U.S. listeners, the rest of the whole world, the whole world, runs on WhatsApp. So a deep dive into how it got that way. Quote, WhatsApp is the world's most widely used messaging app. The company says it has two billion daily users, daily. Those users send more than 100 billion messages every day in 60 languages across 180 countries. Some 400 million of those users are in India. WhatsApp's biggest market. followed by another 120 million in Brazil. WhatsApp initially achieved that global dominance in large part by doing just one thing very well, enabling cheap, private, and reliable messaging on almost any phone, almost anywhere in the world. But in the decades since Meta acquired WhatsApp for an eye-watering $22 billion in 2014, the app has been transformed from a narrowly focused utilitarian tool into a sort of everything app. In countries like India, Brazil, Mexico, and Indonesia, WhatsApp is now also a place for scheduling doctors' appointments and conducting real estate deals and buying ceramic ducks. In Brazil, the beauty juggernaut L'Oreal now makes an average of 25% of its online direct-to-consumer

sales on WhatsApp. The shift has been driven, of course, by money. WhatsApp has never been much of a moneymaker, while Meta makes billions off mining people's personal data to sell more ads. WhatsApp is an encrypted app, whose founders once very publicly swore off advertising altogether. lately, however, WhatsApp has been aggressively luring big businesses to its suite of paid messaging products for businesses and openly flirting with the possibility of introducing ads in the not-too-distant future. True to Meta's appetite for voracious expansion, WhatsApp's goal is nothing short of getting every business in the world on the platform. Meta's head of product for business messaging told rest of world, end quote. This might be the latest I've put a show out in years where I didn't

wait for some sort of news to happen or something to cover. Hopefully this weekend will allow my voice to recover. This is why it's late. But also, I've learned my lesson from pre-announcing things that end up not happening, right? I did record an episode of the Newsworthy podcast this week, where I did a rundown of the top stories of 2024 in the world of tech. Figure you'd like to hear that, since I haven't had time to do something similar here. So if they get me the audio of that episode, I will post it as a bonus episode. episode this weekend. If not, I won't. Maybe next weekend. We shall see. Talk to you on Monday.