Tech Brew Ride Home - Thu. 09/28 – Meta Connect

Episode Date: September 28, 2023

The Meta Connect keynote was yesterday and they announced a bunch of stuff including AI chatbots, a new Quest 3, and the continuation of those Ray-Ban smart glasses that are actually getting kind of interesting. Looks like OpenAI is serious about working with Jony Ive. A big new open-source LLM available to download now. And the new Raspberry Pi 5. Sponsors: ShopBeam.com/ride and code RIDE for up to 40% off TryFum.com and use code ride for 10% off Links: Facebook and Instagram will soon get a slew of AI-powered creator tools (The Verge) Meta is putting AI chatbots everywhere (The Verge) Meta Rolls Out Higher-Priced Quest 3 Headset, Just Ahead of Apple’s Vision Pro (Bloomberg) Meta’s smart glasses can take calls, play music, and livestream from your face (The Verge) OpenAI and Jony Ive in talks to raise $1bn from SoftBank for AI device venture (FT) Mistral AI makes its first large language model free for everyone (TechCrunch) Elon Musk Wins US Space Force Contract for Starshield (Bloomberg) The Raspberry Pi 5 is finally here (The Verge) 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 TechMexam. right home for Thursday, September 28th, 2023. I'm Brian McCullough today.

The Meta Connect keynote was yesterday, and they announced a whole bunch of stuff, including AI chatbots, a new Quest 3, and the continuation of those Rayban smart glasses that are actually getting kind of interesting. Looks like OpenAI is serious about working with Johnny Ive.

A big new open source LLM is available to download now, and the new Raspberry Pi 5. Here's what you miss today in the world of tech. Meta held their Connect event yesterday, announcing a whole bunch of things, so much, in fact, then I'm going to break it up a bit. They announced AI-generated chat stickers for Facebook, Instagram, WhatsApp, and Messenger, and unveiled new AI-powered image editing tools for Instagram.

So again, this is what I was talking about recently when we were talking about AI coming to social. Coding the Verge. AI-powered photo editing on Instagram includes two new features, restyle and backdrop. With restyle users input a text prompt. Meta's examples include watercolor or magazine collage, and the tool updates the existing image based on those. directions. The backdrop feature also utilizes a text prompt by the user to add new AI-generated backgrounds to images, surround me with puppies, for example. For both editing features, Meta says it will

identify when images are created using AI tools so audiences can discern whether what they're seeing is synthetic or human-generated. The company says it's experimenting with other labeling features, including visible and invisible markers, end quote. They also showed off AI assistance for WhatsApp, Messenger, and Instagram, which feature 28 AI characters based on celebrities like Mr. Beast, Charlie Demelio, and Snoop Dog, quoting the Verge. For anyone who has used OpenAI's chat GPT or other chatbots like Anthropics Claude, Meta's AI will immediately feel familiar. Meta sees it as a general purpose assistant for everything from planning a trip with friends

in a group chat to answering questions you'd normally ask a search engine. On that latter piece, Meta is announcing a partnership with Microsoft's Bing to provide real-time web results, which sets meta-AI apart from a lot of the other free AIs out there that don't have super recent information. Another big aspect of the meta-a-I is its ability to generate images like Mid-Journey or Open AIs Dahl E via the prompt slash imagine. In my brief demo, it produced compelling high-res photos in a few seconds. Like all of meta-a-i's features being announced this week, this image generation tool is totally free to use. Alongside Meta's assistant, the company is beginning to rollout an initial roster of 28 AI characters across its messaging apps. Many of them are based on

celebrities like Charlie Demelio, Dwayne Wade, Kendall Jenner, Mr. Beast, Snoop Dog, and Paris Hilton. Others are themed to specific use cases like a travel agent. An interesting twist is an aspect of these characters that All Dolly calls embodiments. As you chat with one of them, their profile image subtly animates based on the conversation. The effect is more immersive than the 2D chatbots I've interacted with to date. For now, Meta AI isn't trained on public user data across Instagram and Facebook, though it sounds like that is coming. It's easy to imagine asking it to Show Me Reels from the south of Italy, and that being a compelling use case that other chat pots can't replicate. We see a long roadmap for us to tie in some of our own social integrations as part of the assistant to make it even more useful, says Al Dali, Meta's spokesperson, end quote.

Matt also debuted the $500 Quest 3, thereby upping the price for their run-of-the-mill VR headset a whopping $200. It does feature a 30% higher resolution than the Quest 2, double the processing power, new lenses, a thinner design, and new controllers, however. Quoting Bloomberg. The device looks similar to the previous version, but has three sensors on the front. The left and right modules are cameras, while the center sensor is a new component for determining where objects and walls are located within a user's room. This allows the device to automatically place virtual guardrails so a user doesn't say walk into a table. It can also be used for gaming, such as one title that allows players to shoot items on a wall. After struggling to make money in VR headsets, Meta is seeking to generate a profit from the Quest 3. It's raising the price of the

device by 67% for a model with 128 gigabytes of storage. The Menlo Park, California-based company is also offering a $650 version with $512 gigabytes. of space. The Quest 3 goes on sale October 10th. Boosting prices in the face of sluggish sales is a gamble for meta. It's failed to attract consumers to its upscale model, the Quest Pro, which costs $1,000, but all the products in its lineup will cost a fraction of Apple's new Vision Pro, which is due in early 2024, end quote. Addy Robertson at the Verge got a hands-on with the Quest 3 and says it's an iterative update, but its mixed reality, pass-through video is less grainy than the Quest Pro, and it offers color, unlike the Quest 2.

Now, and I kind of can't believe I'm going to say this, but maybe the most interesting thing they announced yesterday was the $299-plus Ray-Ban meta-smart glasses, featuring a 12-machixel camera, speakers, five mics, a Snapdragon AR-1 Gen 1 chip, and more, shipping October 17th, quoting the verge. These glasses have two primary purposes. The first is to replace your headphones.

The smart glasses have a similar personal audio system, like Amazon's Echo Frames and the Bose Tempo series, all of which play music but endeavor to make sure only you can hear it. With the new generation of glasses, Meta also upgraded the microphone system in a big way. The specs have five mics, including one in the nosebridge, which should make both your calls and voice commands much clearer. The stories only had one mic, and it kind of fell apart in louder windy conditions. The other job of the glasses is as a camera. The smart glasses have small camera lenses on each right temple, just like the stories, but these cameras take 12 megapixel photos and 1080p videos, both big upgrades from the previous generation.

You can store roughly 500 photos and 130-second videos, that's the maximum length of glasses allow, before you fill up the 32 gigabytes of internal storage, and everything sinks through the MetaVue app. The app also lets you quickly share anything you capture to Meta's many, many sharing platforms. In addition to taking photos and videos on the camera, you can also now start a live stream to Facebook or Instagram with just a couple taps on the stem of the glasses. When you're recording, a white light around the lens pulses to indicate you are recording. Adding live streaming exacerbates a lot of the products already serious privacy questions. It's also one of meta's attempts to answer the big question surrounding all smart glasses. What are you supposed to do with them?

Meta reportedly struggled to keep users interested in wearing its stories, with more than 90% of buyers eventually giving up on the their devices. Last time, Meta's big pitch was messaging. This time, it's pitching the smart glasses as more of a creator tool so you can cook or play drums or do any number of two-handed activities while still recording. One report earlier this year said that the glasses would even let viewers talk back to you, but Meta didn't mention that capability, end quote. More confirmation of this rumor, sources have told the financial times that OpenAI and Johnny Ives love from are indeed in advanced talks with Masio Chishon, two. raise more than a billion dollars from SoftBank in order to build what is being called the iPhone

of artificial intelligence. Quote, Altman and I have held brainstorming sessions at the designer's San Francisco studio about what a new consumer product centered on OpenAI's technology would look like, the people said. They hope to create a more natural and intuitive user experience for interacting with AI in the way that the iPhone's innovations in touchscreen computing unleashed the mass market potential of the mobile internet. The process of identifying a design or device, remains at an early stage with many different ideas on the table, they said. Sahn, SoftBank's founder and chief executive, has also been involved in some of the discussions, pitching a central role for Arm, the chip designer in which the Japanese conglomerate holds a 90% stake,

as well as offering financial backing. Sahn, Altman, and Ive have discussed creating a company that would draw on talent and technology from their three groups, the people said, with SoftBank investing more than a billion dollars in the venture. Discussions are said to be, quote, serious, but no deal has been agreed, they cautioned, and it could be several months before a venture is formally announced. Any resulting hardware product is likely to take years to bring to market, end quote. Paris-based Mistral AI has made its first LLM, Mistral 7B, that it claims outperforms open models with up to 13 billion parameters,

free to use right now without restrictions. Quoting TechCrunch. The most popular language models out there may be accessed via API, but open models, as far as that term can be taken seriously, are gaining ground. Mistral, a French AI startup that raised a huge seed round in June, has just taken the wraps off its first model, which it claims outperforms others of its size, and it's totally free to use without restrictions. The Mistral 7b model is available today for download by various means, including a 13.4 gigabyte torrent with a few hundred cedars already.

The company has also started a GitHub repository and Discord channel for collaboration and troubleshooting. Most importantly, the model was released under the Apache 2.0 license, a highly permissive scheme that has no restrictions on use or reproduction beyond attribution. That means the model could be used by a hobbyist, a multi-billion dollar corporation, or the Pentagon alike, as long as they have a system capable of running it locally or are willing to pay for the requisite cloud resources. Mistral 7B is a further refinement of other small large language models like Lama 2, offering similar capabilities, according to some standard benchmarks, at a considerably smaller compute cost. Foundation models like GPT4 can do much more, but are far more

expensive and difficult to run, leading them to be made available solely through APIs or remote access. Our ambition is to become the leading supporter of the open generative AI community and bring open models to state-of-the-art performance, wrote Mistrel's team in a blog post accompanying the model's release. Mistral 7B's performance demonstrates what small models can do with enough conviction. This is the result of three months of intense work in which we assembled the Mistral AI team, rebuilt a top performance MLOP's stack, and designed a most sophisticated data processing pipeline from scratch, end quote. SpaceX has landed its first U.S. Space Force contract, valued at up to $70 million, offering the military custom satellite communications via the

company's Star Shield Service. Quoting Bloomberg. The previously undisclosed task order adds to SpaceX's growing portfolio of Pentagon business. That includes its competition against United Launch Alliance, a joint venture of Lockheed Martin and Boeing to send up national security payloads, as well as a June Pentagon contract of undisclosed value to provide Starlink satellite communications to the Ukraine military and a Falcon 9 launch of 13 satellites this month for the Pentagon's Space Development Agency. SpaceX's one-year contract for Starshield was awarded September 1st, according to Air Force spokeswoman and Stefanik. The contract with a $70 million ceiling, quote,

provides for Star Shield end-to-end service via the Starlink Constellation, user terminals, ancillary equipment, network management, and other related services, she said. By September 30th, about $15 million will be obligated to SpaceX with funding that supports 54 mission partners across the Army, Navy, Air Force, and Coast Guard, she said. The Star Shield contract, quote, is for a service, but how SpaceX or any other company provides that service is up to them. Lieutenant Colonel Omar Villa Real, a Space Force spokesman said in an email, I am unable to get into specifics, but requirements were received from the Army, Navy, Navy, Air Force,

Coast Guard, and other outside agencies, and combined, he said, end quote. Finally, today, the Raspberry Pi 5 has launched with a 64-bit quad-core arm cortex A76 processor, up to 8 gigabytes of RAM, two 4K-hdmi ports with HDR. It's available in October. starting at 60 bucks for the 4-gibite model. Quoting the verge. Powering the brain of the Raspberry Pi 5 is a 64-bit quad-core arm cortex A76 processor that runs at 2.4 gigahertz, allowing for two to three times the performance boost when compared to the four-year-old Raspberry Pi 4. The device also comes with an 800-MHz video core 7 graphics chip that the Raspberry Pi Foundation says offers a substantial uplift in graphics performance. I got to try out the device.

for myself, and while I didn't have time to do much tinkering with it, I found that it boots up pretty quickly, while also loading web pages fast when compared to my older Raspberry Pi 3 Model B Plus. It does get pretty hot, but luckily Raspberry Pi sent over an active cooling component that I could mount directly on the board. Additionally, the Raspberry Pi 5 features a component made by the Raspberry Pi Foundation for the first time, the Southbridge, also known as a part of the motherboard that helps it communicate with peripherals. With the RP1 Southbridge, the Raspberry Pi Foundation says the microcomputer, quote, delivers a step change in peripheral performance and functionality, enabling faster transfer speeds to external UAS drivers and other peripherals.

It also opens up two four-lane, 1.5 GBPS MIPI transceivers that let you connect up to two cameras or displays. There's also a new single lane PCI Express 2.0 interface for the first time, offering support for, quote, high bandwidth peripherals. However, the Raspberry Pi Foundation notes that you'll still need a separate adapter such as an M2 hat or hardware attached on top for you to take advantage of it. In terms of ports, you can expect dual 4KP60 HDMI display outputs with support for HDR, a microSD slot to USB 3.0 ports, 2 USB 2.0 ports, gigabit Ethernet, and a 5-Val connection via USBC. Some other nice to haves include support for Bluetooth 5.0 and Bluetooth low energy and peak SD card performance that's doubled with the SDR 104 high-speed mode.

Together, all these upgrades make the Raspberry Pi-5 even more versatile, whether you're using it as an ultra-budget desktop PC, a media server, or even a DIY security system. The Raspberry Pi-5 will come with a couple of different RAM options at launch, costing $60 for the 4-gigabyte version and $80 for 8 gigabytes. That makes it slightly more expensive than the Raspberry Pi-4, which is priced at $55 for for 4 gigabytes of RAM and $75 for 8 gigabytes. The Raspberry Pi 5 will be available to purchase before the end of October, end quote. Nothing for you today. Talk to you tomorrow.