Tech Brew Ride Home - Tue. 10/22 – The Next Gen Snapdragon

Episode Date: October 22, 2024

Qualcomm has unveiled its big, next gen Snapdragon SoC which will someday be in all the phones. Except iPhones, of course. Is Netflix pulling back on its gaming strategy? Why a new marketplace from Epic might actually point the way to the metaverse. And how are various people trying to get AI to have a better personality? Links: Snapdragon 8 Elite deep dive: A return to custom CPUs and much more (Android Authority) Scoop: Netflix shuts down 'AAA' Team Blue gaming studio, amid gaming shake-up (Game File) Netflix has closed its AAA gaming studio (Engadget) Hulu and Disney+ No Longer Support Signups and Payment Using App Store (MacRumors) Epic’s ambitious digital asset shop is now open (The Verge) Biden administration proposes new rules governing data transfers to adversarial nations (The Record) How AI groups are infusing their chatbots with personality (Financial Times) 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 Tuesday, October 22nd, 2024. I'm Brian McCullough today. Qualcomm has unveiled its big next-gen-snapdragon system on a chip, which will someday be in all the phones, except iPhones, of course, is Netflix pulling back on its gaming strategy, why a new marketplace from Epic might actually point the way to the metaverse, and how are various people trying to get AI to have a better personality? Here's what you miss today in the world of tech. Qualcomm has unveiled the 3-nometer Snapdragon 8 Elite. mobile system on a chip with two prime and six performance CPU cores, claiming 45% faster CPU and NPU over the 8 Gen 3, quoting Android Authority. The Orion CPU is this year's headline ticket.

It marks Qualcomm's return to CPU development after years of buying parts from Arms Cortex series. The CPUs are still based on the ARM architecture, but a fully custom design frees Qualcomm to configure the inner workings of Orion exactly as it sees fit, with the additional development expenses, of course. Qualcomm is keen to stress that the CPUs inside the Snapdragon 8 Elite aren't just a slim-down version of the Phoenix CPU cores inside its laptop-grade Snapdragon X series. No, they're a ground-up design specifically for mobile applications. For starters, the CPU core setup has been greatly simplified compared to the previous years. The quad-tier arrangement of big, middle, and little cores is replaced by a simple two-tier design of two prime cores and six performance cores.

This is a much more Apple-like approach, which also doesn't use any small cores for background tasks. The clock speeds are also notably different, reaching a colossal 4.32 gigahertz on the two fastest cores, virtually a whole 1 gigahertz faster than its predecessor. This will surely only be for short bursts, but these laptop-class clock speeds should ensure snappy responsiveness. The great unknown is just how hot and power-efficient this setup is, but TSM's cutting-edge 3-nometer manufacturing process will undoubtedly help Qualcomm reach its 45% better efficiency claim, end quote. While the clock speed is just one aspect of CPU performance, Qualcomm remains tight-lipped about the internal architecture of its Orion cores, making direct comparisons with Apple or Arm Cortex difficult.

The prime and performance cores share the same architecture but are optimized differently for single-and-maltreaded workloads. The revamped Adrino GPU brings impressive improvements offering either 40% faster performance or 40% better power efficiency, plus a 35% boost in ray tracing capabilities. The GPU now features 12 megabytes of dedicated memory and 40% reduced memory bandwidth, contributing to potentially longer gaming sessions, though actual battery life gains will vary by device and usage. On the AI front, the hexagon NPU has been significantly upgraded. It now supports up to 70 tokens per second, on certain LLM models and is compatible with multimodal Gemini Nano.

With additional scalar and vector processors, now 8 and 6 respectively, and improved memory throughput, the NPU delivers up to 45% better performance and power efficiency depending on the workload. Qualcomm has also enhanced the integration between the image signal processor and NPU through architecture changes that optimize memory sharing for computer vision and AI tasks. This improvement allows for more efficiency, and simultaneous execution of multiple AI models and tasks on the NPU. And quoting Android Authority again, Qualcomm's Snapdragon 8 Elite undoubtedly marks a key turning point in the series history.

The return to custom CPU development promises major performance and efficiency gains that could propel the elite well ahead of mobile rivals. Paired with seemingly huge GPU, AI, and imaging improvements, there's something for everyone's next-gen phone, although this tech looks set to come at a much higher price. Thankfully, we won't have to wait too long to get our hands on the Snapdragon 8 Elite. ASIS, Honor, Iqo, One Plus, Apo, Real Me, Samsung, Vivo, and Jaume are all listed as planning announcements, some of which may appear before the end of the year.

The One Plus 13, for example, will launch before the end of October. It's reported that next year's Samsung Galaxy S-25 series will also exclusively feature the Snapdragon 8 Elite, given the sweeping changes and performance claims, it's not so surprising that the Exxinus Snapdragon Divide might be too big for Samsung to bridge this year, end quote. Noting this as interesting, Netflix has shut down AAA Game Studio Team Blue, which hired industry veterans and seemed set to pursue a big budget multi-device strategy. So does this signal a gaming strategy shake up for Netflix? Quoting Game File. The company's Southern California Game Studio, one of a handful of internal studios assembled by Netflix in recent years as part of its expansion

into video games has been shut down. GameFile has learned. The studio, also known as Blue, seemed poised to break the mold of what Netflix is doing in gaming. Early hints about the team's work suggested it was pursuing a big-budget multi-device strategy signifying the ambitious edge of Netflix's initial mobile-focused expansion, end quote. And quoting in gadget. In 2022, Netflix brought on former Overwatch boss Chakosani to head up a new SoCal-based AAA studio. A year later, former Halo exec, Joseph Staten, was hired on as creative director, followed by God of War art director, Raphael Grissetti. At the time, Staten said he was working on a multi-platform AAA game with all-new IP. Those executives are no longer

with the company and Netflix confirmed to GameFile that Team Blue has been shut down, and Gadgett reached out to Staten and Grissetti for comment. Netflix got into gaming in 2017 with its retro-inspired Stranger Things mobile game. In the following years, its strategy was primarily centered around obtaining mobile publishing rights to respected indie titles like Into the Breach and Terra Nill. It also built a number of mobile games in-house. In 2020, Netflix had nearly 90 such titles in-house and partner studios.

Those can be accessed through Netflix's app and play directly on your device, so they fit in pretty well with its overall streaming strategy. The Team Blue Studio Venture didn't quite line up as neatly with that model, though, which made it a bit of a surprise. As in Gadgetz, Igor Bonafachic noted at the time, quote, funding the development of a multi-platform AAA game is significantly more ambitious, and it should be noted, risky, end quote. Quoting Henry Gilbert on Twitter, every huge media company wants to make a AAA game for a couple of years,

until they see how much it actually costs, then they just go whoopsie in the form of layoffs, end quote. Sticking with streaming for a second more, but returning to the whole App Store debate, Disney no longer lets new and returning customers sign up for and buy subscriptions to Hulu or Disney Plus through Apple's App Store, thereby avoiding Apple's cut. Quoting Mac Rumors. The change was noted on Reddit over the weekend, and there are details on the Disney Plus and Hulu websites. Both websites say that new and returning subscribers cannot sign up for and pay through Apple, but existing Apple build subscribers are not affected. Hulu and Disney Plus customers can still watch content on Apple. devices by signing into their Hulu and Disney Plus accounts in the iOS apps. Disney is now directing

customers to its websites in order to see all plans and sign up for a subscription directly. For digital purchases like subscription services, Apple collects a 15 to 30% fee from each transaction. By removing the option to sign up for streaming services using the in-app purchase system, Disney will not have to pay a fee to Apple, and it will be classified as a reader app. Disney's decision to stop supporting in-app purchases correlates with price increases that went into effect on October 17th. Disney Plus plan pricing went up by $2, while Hulu with ads is $2 more and Hulu without ads is $1 more. Eliminating the in-app purchase fees paid to Apple seems to be part of an effort to boost streaming revenue paired with price hikes and also a recent crackdown on

password sharing. Disney Plus and Hulu will no longer be eligible for Apple's video partner program going forward, and subscribers may notice a change in TVOS and the Apple TV app. The video partner program provides integration with Apple technologies like Siri, Airplay, Universal Search, zero sign-on, and more for streaming video apps that support an app purchase, end quote. This is interesting. Epic has launched its Fab Marketplace, a unified place to get digital assets like environments and animations, offering an 88% revenue share to creators. Quoting the Verge. The launch of the marketplace is an important step in Epic's plans to make it easier for developers to acquire assets for developing games, and maybe some.

someday to let them make content for an interoperable metaverse. The company first revealed its plans for Fab, which will combine offerings from the Unreal Engine Marketplace, SketchFab Store, Quixel, and Art Station Marketplace into one place in March 2023. When Epic initially planned to launch Fab in 2023, the company later delayed the launch to 2024. Epic is offering an 88% revenue share for goods sold on Fab, matching the revenue share it offers developers on the Epic Games Store. of the launch, Epic has allowed creators to prepare their Unreal Engine Marketplace and SketchFab products to be available on Fab. According to Epic, ArtStation Marketplace will continue operating as it does today, but the company plans to let ArtStation users bring over their content to Fab in 2025.

Epic is also sharing some plans about updates to come for Fab. Perhaps most notable is that the company is reiterating that it plans to add support for offering Roblox and Minecraft assets. Epic also plans to let people access its hyper-realistic metahumans on Fab. end quote. Here's why this is interesting. Increasingly, if we do get a metaverse soon, people are starting to think that we might get it in a bottom-up way, sort of how social media is now increasingly just millions of people creating video and other types of content, but largely video, that at least in terms of time spent consuming, competes with Hollywood. Again, not necessarily competing with Hollywood in terms of quality, but in time spent.

So imagine a social media-like world where people have their own environments, their own avatars, their own games, but largely their, I don't know, their own hangs, their own experiences, but in a bespoke sort of handcrafted way, like how social media is now, this seems to be the puck moving in that direction. The Biden administration has unveiled proposed rules for restricting the transfer of U.S. citizens' sensitive data, including biometric and genomic info, to so-called adversarial nations, quoting the record. The Biden administration announced on Monday new proposed rules for regulating the transfer of certain data to adversarial countries such as China and Russia, creating specific requirements for how sensitive personal and federal information can be shared, if at all. The proposed regulations follow the release of a February executive order designed to block four. foreign adversaries from exploiting easily obtained American financial, biometric, precise geolocation, health, genomic, and other data to carry out cyber attacks or spy on Americans. Under the proposed rules, data transfers to companies and individuals in six countries, China, Russia, Iran, North Korea, Venezuela, and Cuba, will be prohibited when specific preset

volume thresholds are exceeded, according to a detailed fact sheet released by the administration and comments from senior administration officials. Specifically, U.S. companies will be restricted from transferring more than 100 Americans genomic data across any 12-month period to targeted countries. Data transfers of more than 1,000 Americans' geolocation data and biometric identifiers, more than 10,000 Americans' health and financial data, and more than 100,000 Americans' personal identifiers will also be barred. Personal identifiers include names linked to device IDs, social security numbers, and driver's license numbers. Data belonging to even a single active duty member of the military or federal personnel will be prohibited from being transferred,

as will data broker sales, where the seller has reason to believe the information they are peddling, will make its way to any of the six countries. U.S.-based data brokers of all sizes and types are primary targets for the rule, a senior administration official suggested during a Monday press call. Both third-party data brokers and companies selling data they have collected will be entirely prohibited from data transactions tied to the six designated countries. Data broker sales to those countries pose a serious threat to national security, the official said, end quote. Okay. Now do a similar rule, but for corporations, even those in the

domestic U.S. Finally, today, a look at how OpenAI, Google, and Anthropic are using different approaches to improve so-called model behavior, an emerging field that is shaping systems' responses and characteristics. Quoting the FTE, leading artificial intelligence companies that are racing to develop the cutting-edge technology are tackling a very human challenge how to give AI models a personality. Their differing approaches to model behavior could prove crucial in which group dominates the burgeoning AI market as they attempt to make their models more responsive and useful to millions of people and businesses around the world. The groups are shaping their models to have characteristics such as being kind and fun while also enforcing rules to prevent harm and ensure nuanced interactions.

For example, Google wants its Gemini model to, quote, respond with a range of views only when asks for an opinion, while OpenAI's chat GPT has been instructed to, quote, an objective point of view. It is a slippery slope to let a model try to actively change a user's mind. Joanne Zhang, head of product model behavior at OpenAI, told the Financial Times, how we define objective is just a really hard problem on its own. The model should not have opinions, but it is an ongoing science as to how it manifests, she added. That approach contrasts with Anthropic, which says that models like human beings will struggle to be fully objective. I would rather be very clear that these models aren't neutral arbiters, said,

Amanda Askell, who leads character training at Anthropic. Instead, Claude has been designed to be honest about its beliefs, while being open to alternative views, she said. Anthropic has conducted specific character training since its Claude III model was released in March. This process occurs after initial training of the AI model, like human labeling, and is the part that, quote, turns it from a predictive text model into an AI assistant, the company said. At Anthropic, character training involves giving written rules and instructions to the model. This is followed up with the model conducting role play conversations with itself and ranking its responses with how well they match that rule. One example of Claude's training is,

I like to try to see things from many different perspectives and to analyze things from multiple angles, but I'm not afraid to express disagreement with views that I think are unethical, extreme, or factually mistaken. The outcome of the initial training is not a, quote, coherent, rich character. It is the average of what people find useful or like, said Askell. After that, decisions on how to fine-tune Claude's personality in the character training process is, quote, fairly editorial and philosophical, she added. Open AIs Jang said Chat Chepti's personality has also evolved over time. I first got into model behavior because I found Chat Chiept's personality very annoying, she said. It used to refuse commands be extremely touchy, overhaging, or preachy, so we tried to remove the annoying parts and teach some cheery aspects like it should be nice, polite, helpful, and friendly.

But then we realized that once we tried to train it that way, the model was maybe overly friendly, end quote. Zhang said creating this balance of behaviors remained in, quote, ongoing science and art, noting that in an ideal world, the model should behave exactly as the user would want it to. AI groups are also developing customizable agents that can store user information and create personalized responses. One question presented by Zhang was if a user told chat GPT they were a Christian, and then days later asked for inspiration. inspirational quotes, would the model provide Bible passages? While Claude does not remember user interactions, the company has considered how the model might intervene if a person is at risk. For example, whether it would challenge the user if they tell the chatbot they are not socializing with people due to being too attached to Claude. A good model does the balance of respecting human autonomy

and decision-making, not doing anything terribly harmful, but also thinking through what is actually good for people and not merely the immediate words of what they say that they want, said Askell. She added, that delicate bouncing act that all humans have to do is the thing I want models to do, end quote. Nothing more for you today. Talk to you tomorrow.