Tech Brew Ride Home - Tue. 11/12 – Apple Into The Smarthome

Episode Date: November 12, 2024

Looks like Apple wants to get serious about smarthome tech. The EU gets Meta to change, but has another requested change from Apple. Signal has some new features. Netflix’s ad-tier continues to win. And a deeper dive into how AI can leap this scaling wall it seems to have run into. Here’s what you missed today in the world of tech. Sponsors: Shopify.com/ride Links: Kuo: Apple to release its own smart home camera in 2026, with AI features (9to5Mac) In Europe, Instagram Ads Are About to Get Less Personal (WSJ) Apple Gets EU Warning to Stop Geo-Blocking on App, ITunes Stores (Bloomberg) Nvidia’s new app replacement for GeForce Experience is now out of beta (The Verge) Signal calls now work a lot more like Meet and Zoom (The Verge) Netflix Says Ad Tier Reaches 70 Million Users Globally (Variety) OpenAI and others seek new path to smarter AI as current methods hit limitations (Reuters) 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 TechMeme right home for Tuesday, November 12th, 2024. I'm Brian McCullough today. Looks like Apple wants to get serious about smart home tech. The EU gets meta to change, but has another requested change from Apple. Signal has some new features. Netflix's ad tier continues to win and a deeper dive into how AI can leap this scaling while it seems to have run into.

Here's what you miss today in the world of tech. Ming Chi Kuo says that Apple has scheduled production of a smart home. IP camera for 2026 and is aiming for more than 2 million in annual shipments. Quoting, see what I did there, the smart home IP camera scheduled for production in 26 is designed to integrate seamlessly with other Apple hardware products via wireless connectivity. Current global smart home IP camera shipments are in the range of 30 to 40 million units per year, with Apple's long-term goal of annual shipments of 10 plus million units for this product line. This strategic move demonstrates Apple's continued exploration of growth opportunities in the home market.

I believe the user experience will be significantly enhanced by Apple's great ecosystem and deep integration with Apple Intelligence and Siri. Following the success of the Apple Watch, AirPods are being repositioned to focus on health management with future models integrating more health management features. This strategic positioning is expected to drive AirPods shipments from approximately 48 million units in 2023 to 53 to 53 to 55 million in 2024. 58 to 62 million in 2025 and 65 to 68 million in 2026. Goer Tech is positioned to secure the NPI for the new 2026 AirPods model and serve as the primary supplier, end quote. And quoting 9 to 5 Mac. We have heard for a while that Apple is plotting a more serious attempt to crack the smart home space with intermittent rumors about Apple designing its own accessories. However, it's unclear at this stage what unique functionality Apple could offer through a first-party camera rather than continue

continuing its integrations with other manufacturers through the home app. Perhaps a powerful embedded Apple Silicon chip will enable enriched Apple intelligence features that other cameras cannot offer. The first big Apple Smart Home Play will be a new Smart Display device that is likely going on sale sometime next year. The new Smart Display will combine a HomePod-like speaker with a small, squareish touchscreen for showing a visual user interface. That device will be Apple's competitor to products like the Echo Show with a likely

the kitchen top focus for showing timers and recipes. A couple years after that, Apple is said to be making a more advanced display that will be hoisted on an articulated robotic arm so the display can tilt and pan to follow the user around a room, end quote. Meta plans to ask Facebook and Instagram users in Europe if they want to get less personalized ads and is cutting subscription prices by 40% a concession to the EU regulators. Quoting the journal, The less personalized format will show European users what Meta calls contextual ads based on content that a user sees during a given browsing session, rather than a user's broader activity history, like the ads most users now see peppered into their feeds and stories on Meta's apps.

Some of the new ads, which will also be targeted based on age, gender, and location will cover the whole screen and be unskippable for a few seconds. Meta's new ad option comes amid pressure from European Union regulators who say users should have access to a free version of the company's apps with less personalized ads, end quote. Meta has historically avoided offering users a free option to prevent their platform activity from being used for ad targeting since doing so could undermine its primary revenue stream, precisely targeted advertising based on user behavioral data, such as their sharing patterns and video viewing habits. The company maintains that personalized advertising enhances the experience

for both users and advertisers. The financial impact of this change remains uncertain and will largely depend on user adoption of the new ad settings. In communications with EU regulators, META has indicated it anticipates negative business effects from offering less personalized advertising. The broader European region, including EU nations, represents 23% of META's revenue, according to recent financial statements. This advertising policy shift comes amid growing tensions between META and EU regulators.

The company recently led the creation. of an open letter expressing concern that EU regulations could hamper regional advancement in artificial intelligence. A key point of contention centers on whether meta must obtain explicit user permission for data collection, both for advertising purposes and AI development. The company put its AI chatbot launch in the EU on hold, after regional privacy authorities suggested meta might need user consent to utilize public Instagram and Facebook posts from European adults for model training. Meanwhile, the EU has notified Apple of potentially prohibited geoblocking practices on the App Store and other apps and says it must align with EU anti-geoblocking rules. Quoting Bloomberg,

Apple's App Store, iTunes Store, and other media services unlawfully discriminate against European customers based on their place of residence, according to a European Commission statement on Tuesday. The geolocating investigation was conducted together with a network of national consumer authorities and found Apple media services only allow users to use payment cards issued in their countries. They registered their Apple accounts, according to the statement. The App Store also blocks users from downloading apps offered in other countries the investigation found, end quote. The warning comes as Apple confronts its first ever penalty under the Digital Markets Act over restrictions preventing app developers from directing users toward more affordable options,

according to recent Bloomberg news coverage. This sanction follows several months after. After Apple received a $1.8 billion fine for comparable violations under the EU's standard competition regulations. Under the terms outlined in the statement, Apple has been given a 30-day window to address the findings and submit proposed solutions to resolve matters. NVIDIA has launched its NVIDIA app for Windows designed to overhaul and replace the G-Force experience app after a beta earlier in this year. Quoting the verge. The NVIDIA app includes a redesign and improved overlay with support for 4K 120 frames per second AVI video capture, NVIDIA's AI-powered RTX game filters, and an improved customizable statistic overlay.

Just like GeoForce experience before it, the NVIDIA app includes optimal playable settings for games, and it even includes a number of options that were previously only available in the NVIDIA control panel app. You can now configure displays, enable G-Sync, and tune GPU performance all inside the NVIDIA app without having to head into the separate control panel app, end quote. NVIDIA continues to migrate functionality from its control panel application into its primary Nvidia app. In future updates, we'll continue to add remaining NVIDIA control panel options like surround and multiple display mode settings, along with new features and further enhancements,

they say in a blog post. The company has recently integrated display configuration options, RTS Video HDR Adjustment Controls, and RTS Video Super Resolution compatibility into this updated application. The revamped Nvidia app will be available as an optional component during G-Force driver installation, and users can access it without requiring an Nvidia account login. Signal has rolled out new call features, including the ability to share a link to a call, emoji reactions, a raise-hand button, and a dedicated call history tab, quoting the verge. Previously, you could only initiate a call from a Signal group chat,

so this new feature should make it much easier to hop on the phone with a friend or coworker by sending them a link inside or outside of Signal. You can also reuse call links, which could be handy for recurring work meetings or a weekly chat you have with your family. Within calls, Signal is getting some upgrades to, including a raise hand button to let people know that you have something to say, an emoji reaction, so you can spam your emoji feelings. And the app now has a tab just for keeping track of your calls, which could be useful if you want to look back on your call history. The new call features are available on the newest version of Signal on Android, iOS, and desktop, end quote. Netflix says its ad tier now has 70 million users globally up from 40 million in May of this year,

and 22 million in January of 2024. And, quoting variety, Netflix continues, quote, to see steady progress across all country's member bases, said Amy Reinhard, president of advertising at Netflix in a statement, now over 50% of new Netflix signups are for the ads plan in ad-supported countries, end quote. While Netflix's global numbers are impressive, most marketers don't make global ad buys, choosing to aim different promotional messages in specific geographic regions, and media buyers in the U.S. have indicated that advertisers want to see the Netflix ad tier generate more impressions overall. But the growth in ad tier subscribers has been sustained, which could spark new interest from Madison Avenue.

In the meantime, Reinhard says Netflix has sold out all available commercial inventory in its coming Christmas Day NFL broadcasts the first time the streamer will show live games from a major sports league. Advertisers in the games will include Fanduel and Verizon. Fandual will serve as an exclusive sportsbook betting partner during a pregame program that will offer a sponsored segment that will have talent in the show offering analysis and predictions tied to Fandul's odds for the games. Verizon will sponsor a segment that will run before kickoff. She also offered details of sponsorship for the next season of Squid Game. The executive noted that Netflix had partnered with multiple advertisers across our 12 ad-supported countries, including Kia and Korea. As part of that pact, Kia will get to spotlight its new Sportage, a crossover SUV with a three-part set of link commercials. Kia will also host a pop-up event at Kia Unplugged Ground, a complex in Seoul, end quote.

So my usual joke here is, you know, blowing up the cable bundle just to reconstitute it. But let me ask this a different way. Was the original Netflix offering of no ads but pay us a monthly subscription wrong in retrospect? Or was it just right for its time? I.E., when Netflix was going up against the cable bundle, the subscription seemed like a deal, right? Going ad-free seemed like freedom. But now that they're competing with other subscription streamers and possibly subscription overload is ads the way to go. I guess I'm asking something that might be too basic and obvious,

but did we just run an elaborate experiment just to prove that ads supported is the best business model for video content, and we've just spent a decade and a half proving that out for no good reason. Finally today, remember our recent musings about whether or not the current models of AI are hitting a wall in terms of scale improvements, and maybe we need to find a new technique? Well, Ilya Saskervir, formerly of OpenAI, but now at a new AI startup, as a founder says, quote, we're back in the age of wonder and discovery as AI companies focus on pre-training, inference improvements and finding the, quote, next thing. Quoting Reuters, artificial intelligence companies like OpenAI are seeking to overcome unexpected delays

and challenges in the pursuit of ever bigger large language models by developing training techniques that use more human-like ways for algorithms to think. A dozen AI scientists, researchers and investors told Reuters they believe that these techniques, which are behind OpenAI's recently released O1 model, could reshape the AI arms race and have implications for the types of resources that AI companies have an insatiable demand for, from energy to types of chips. OpenAI declined a comment for this story, but after the release of the viral chat GPT chatbot two years ago, technology companies whose valuations have benefited greatly from the AI boom

have publicly maintained that scaling up current models through adding more data and computing power will consistently lead to improved AI models. But now some of the most prominent AI scientists are speaking out on the limitations of this bigger is better philosophy. Elias Askever, co-founder of AI Labs, Safe Super Intelligence, SSI, and OpenAI, told Reuters recently that results from scaling up pre-training the phase of training an AI model that uses a vast amount of unlabeled data to understand language patterns and structures have plateaued. Suskever is widely credited as an early advocate of achieving massive leaps in generative AI advancement through the use of more data and computing power in pre-training,

which eventually created chat GPT. Susceiver left open AI earlier this year to found SSI. The 2010s were the age of scaling. Now we're back to the age of wonder and discovery once again. Everyone is looking for the next thing, Susceiver said. scaling the right thing matters now more than ever. So Skever declined to share more details on how his team is addressing the issue other than saying SSI is working on an alternative approach to scaling up pre-training. Behind the scenes, researchers at major AI labs have been running into delays and disappointing outcomes in the race to release a large language model that outperforms OpenAIs GPT4 model, which is nearly two years old, according to three sources familiar with private matters.

The so-called training runs for large models can cost tens of millions of dollars by simultaneously running hundreds of chips. They are more likely to have hardware-induced failure given how complicated the system is. Researchers may not know the eventual performance of the models until the end of the run, which can take months. Another problem is large language models gobble up huge amounts of data, and AI models have exhausted all the easily accessible data in the world. Power shortages have also hindered the training runs, as the process requires. vast amounts of energy. To overcome these challenges, researchers are exploring test time compute, a technique that enhances existing AI models during the so-called inference phase, or when the model is being used. For example, instead of immediately choosing a single answer, a model could generate

and evaluate multiple possibilities in real time, ultimately choosing the best path forward. This model allows models to dedicate more processing power to challenging tasks like math or coding problems or complex operations that demand human-like reasoning and decision. decision-making. It turned out that having a bot think for just 20 seconds in a hand of poker got the same boosting performance as scaling up the model by 100,000x and training it for 100,000 times longer, said Noam Brown, a researcher at OpenAI who worked on 01 at TED AI conference in San Francisco last month. OpenAI has embraced this technique in their newly released model known as O1, formerly known as Q-Asteris and Strawberry, which Reuters first reported in

July. The O-1 model can think through problems in a multi-step manner similar to human reasoning. It also involves using data and feedback curated from PhDs and industry experts. The secret sauce of the O-1 series is another set of training carried out on top of base models like GPT4, and the company says it plans to apply this technique with more and bigger base models. At the same time, researchers at other top AI labs from Anthropic, XAI, and Google DeepMind have also been working to develop their own versions of the technique, according to five people familiar with the efforts. We see a lot of low-hanging fruit that we can go pluck to make these models better very quickly, said Kevin Wheel, chief product officer at OpenAI at a tech conference in

October. By the time people do catch up, we're going to try to be three more steps ahead, end quote. Hey, hive mind, I've got a question for you. So I recorded this week's Rad History episode yesterday. It's coming out tomorrow. But unlike some of the other episodes, this one is not an interview. It's just me. So yesterday I sat in front of a camera and read a script off of a teleprompter. But you know, who needs to see my ugly mug, right? And also, since I'm not sticking my head in my recording booth, the sound quality is not as good as, say, you're hearing right now. What I'd love to do is record the audio in the booth for a video to get the best quality, but then upload it to an AI avatar of myself to do the talking head part of the video.

hive mind, what I want you to do is when you hear of an AI startup or company that does that that I just described, let me know. Yes, I know about Synthesia and 11 Labs and Descript, but most of the talking head avatar stuff is merely text to video and audio. I don't want text. When it reads the script in my voice, it sounds robotic. I want to upload the audio of me actually reading the script, but have the AI avatar speak it through its mouth.

It doesn't even have to be me as the avatar. Heck, I'd take a cartoon avatar at this point, but none of the tools I can find can do audio to video. It's all text to audio and video. So to be clear, what I'm looking for is a workflow like this. I write the script and record the audio in the booth. Then I upload the MP3 or wave file of the audio to a service. It sock puppets it into a talking head video using that audio so it doesn't sound robotic.

but the avatar on the video speaks my words in a reasonably not uncanny Valley way. I just want to eliminate the part where I speak into the camera myself, talking from a teleprompter. If you ever run across something like that, let me know. Talk to you tomorrow.