Tech Brew Ride Home - Tue. 12/26 – Apple’s AI Strategy Comes Into View

Episode Date: December 26, 2023

Apple’s whole AI strategy has been coming into focus these last couple of weeks. Two stories on this, including Apple’s first announced LLM. OpenAI is fundraising again. Hackers are targeting gaming studios again. And as President Biden didn’t deliver the pardon, what is this patent case against the Apple Watch? I’ll explain. Links: Apple quietly released an open source multimodal LLM in October (VentureBeat) Apple Explores A.I. Deals With News Publishers (NYTimes) OpenAI Is in Talks to Raise New Funding at Valuation of $100 Billion or More (Bloomberg) Ubisoft says it's investigating reports of a new security breach (BleepingComputer) GTA 5 source code reportedly leaked online a year after RockStar hack (BleepingComputer) Puzzle game ‘Royal Match’ dethrones ‘Candy Crush’ from top of app store (Financial Times) Apple Watch battle heralds further patent wrangles (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, December 26, 2023. I'm Brian McCalla today. Apple's whole AI strategy has been coming into focus these last couple of weeks. Two stories on this, including Apple's first announced LLM. OpenAI is fundraising again. Hackers are targeting gaming studios again, and President Biden didn't deliver the pardon. So what is this patent case against the Apple watch? I'll explain. Here's what you missed today in the world of tech. In a recent interview on CNBC, Apple's Johnny Shrugi and John Ternis were asked if, they were worried Apple was falling behind in AI. Both of them just chuckled and said, Not too worried. Yeah, you had to know Apple wasn't sleeping on AI, and increasingly,

their belated entry into the artificial intelligence race is coming into focus. Researchers from Apple and Columbia University have released Ferret, an open-source, multimodal LLM that can recognize and describe any shape in an image. They did so back in October, and nobody noticed until now, quoting Venture Beat. The company released two new research paper, introducing new techniques for 3D avatars and efficient language model inference. The advancements were hailed as potentially enabling more immersive visual experiences and allowing complex AI systems to run on consumer devices such as the iPhone and iPad. Many in the AI community who belatedly noticed the ferret release celebrated Apple's unexpected

entry into the open source LLM landscape, especially since Apple has traditionally been known as a walled garden. This morning, Bart DeWitt, who runs a European nonprofit focused on open source AI in medicine posted on X, I somehow missed this, he wrote. Apple joined the open source AI community in October. Ferrett's introduction is a testament to Apple's commitment to impactful AI research, solidifying its place as a leader in the multimodal AI space. P.S., I'm looking forward to the day when local large language models, or LLLMs, run on my iPhone as an integrated service of a redesigned iOS, end quote. So Apple phones already have those custom neurochips in them, that excel at running machine learning right. If they can run this LLM locally on device, that would

satisfy their security and privacy bona fides. More on that in the next segment. What's interesting is how they did this. The name of the paper is efficient large language model inference with limited memory. As Lao Kai said on threads, quote, they have proposed an inference cost model that coordinates with the behavior of flash memory usage guiding optimization in two key areas, reducing the amount of data transferred from flash memory and reading data in larger, more continuous blocks. These methods collectively enable running models up to twice the size of the available DRAM with a 4 to 5x and 20 to 25x increase in inference speed compared to native loading approaches in CPU and GPU, respectively, end quote.

And by open sourcing this, is Apple opening the door to developers around the world building out this technique so that AI can live on hardware, even small hardware, even on smartphones, not in a data center? That would make sense. Microsoft and Amazon and Google all make their money in the cloud. Apple makes its money selling hardware. Would be wild if this is a strategic move to shift AI to local hardware and aviate data centers. Long way away from that, but still, as CSPen said on threads, quote, do you see where Apple is going if they can get mixture of experts to run at 2x in the available DRAM and get 20 to 25x increases? an inference, you could run MixTroll at top speed on a MacBook paralleling GPT3.5 performance on a

consumer laptop, end quote. Interesting that Apple is embracing open source too, but if they integrate LLMs at the OS level, then why not? As Bart DeWitt said on Twitter, quote, with the advancement of edge AI technology becoming a reality, Microsoft reported annual spending of $50 billion on data centers needs to be reexamined, end quote. So sort of shot and chaser to that whole Apple has its own LLM story. Sources also say Apple has discussed multi-year deals with NBC News, Condi Nast, IAC, and other media organizations to train its generative AI systems on those various companies' news articles. Now, we've talked about this recently.

Remember, OpenAI just signed a deal with Axel Springer, but that's the point. Weirdly, Apple finds itself in basically the same position as OpenAI. They have to pay up to get data to train on. Meta doesn't have this problem. They have all the data from Facebook, from Instagram, from their users. Google has all the web that their search engine has crawled and they have YouTube. But suddenly, Apple's vaunted obsession with user privacy, which has been such an advantage for them for years, means they can't access the data their billions of users produce every day to train on their forthcoming AI. So Apple has to pay up to others. Of course, they can afford it, but still, quoting the New York Times,

The technology giant has floated multi-year deals worth at least $50 million to license the archives of news articles, said the people with knowledge of the talks who spoke on the condition of anonymity to discuss sensitive negotiations. The news organizations contacted by Apple include Condé Nass, publisher of Vogue, and The New Yorker, NBC News, and IAC, which owns People, The Daily Beast, and Better Homes and Gardens. Some of the publishers contacted by Apple were lukewarm on the overture. After years of on-again, off-again commercial deals with tech companies like Meta, the owner of Facebook. Publishersers have grown wary of jumping into business with Silicon Valley, end quote. Apparently, the publishers were nervous about the extensive scope of Apple's proposed terms. The core of the proposal involved a wide-ranging license for the publisher's entire archive of content. This raised alarm bells about the potential legal liabilities publishers could face due to Apple's utilization of their materials. A significant point of contention was Apple's opaque strategy regarding the integration of AI into the news sector, given Apple's considerable user base for news on its platforms, this lack of clarity there was seen as a potential competitive hazard. But despite those concerns,

according to the Times, a sense of optimism remained among some executives about the possibility of a collaboration with Apple. They noted Apple's proactive approach in seeking permissions was a stark contrast to the practices of some other AI-driven companies. The other AI outlets have, in their opinion, been, you know, training their generative AI models on their content without consent for a while now. The Times says Apple's leadership has engaged in internal debates for years now about gathering data for developing generative AI products. Apple has tons of data on its users, or it could have, but it doesn't want to make use of that. Apple's cautious stance on data collection, particularly from the internet, is longstanding as well. Take the example of Apple's acquisition

of Topsy, a social analytics firm in 2013. Following the acquisition, Apple instructed Topsy to halt its data aggregation from Twitter. This directive was in line with Apple's policy against collecting data that could potentially include information from Apple customers who also use social media platforms like Twitter. Surprising absolutely no one. Open AI is reportedly in early talks to raise funding at a more than $100 billion valuation, and that includes possibly raising between $8 and $10 billion from Abu Dhabi-based G4-2 for a chip venture. So it looks like Sam Altman's grand ambitions are back in play, quoting Bloomberg. The startup has discussed raising between $8 and $10 billion

from G42. It's unclear whether the chips venture and wider company funding efforts are related, Open AI chief executive officer Sam Altman had been seeking capital for the chipmaking project codenamed Tigris. The goal is to produce semiconductors that can compete with those from Nvidia, which currently dominates the AI chip market Bloomberg News reported last month. In October, G42 announced a partnership with OpenAI, quote, to deliver cutting-edge AI solutions to the UAE and regional markets, end quote. No financial details were provided. The firm founded in 2018 is led by Sheikh Tanun bin Zayed al-Nal-Nan. Ayan, the UAE's National Security Advisor and chair of the Abu Dhabi Investment Authority.

OpenAI is set to complete a separate tender offer in early January, which would allow employees to sell their shares at evaluation of $86 billion, Bloomberg previously reported. That is being led by Thrive Capital and saw more demand from investors than there was availability. People familiar with the matter have said, end quote. I've asked this before, and I'll ask it again, why do hackers love going after Game Studios? Ubersoft is investigating reports of a new security breach after security research collective VX Underground shared screenshots of the company's internal tools, quoting bleeping computer. As part of this alleged breach, the threat actor claimed they gained access to the Ubisoft SharePoint server, Microsoft Teams, Confluence, and MongoDB Atlas' Atlas' access to some of these services. MongoDB Atlas recently disclosed a breach, but based on their disclosure, it does not appear that this incident is related.

The threat actors told VX Underground that they attempted to steal data from Rainbow Six siege users, but were detected and lost access before they could do so. Ubisoft was previously breached by the Greger Ransomware gang in 2020, who released portions of the Ubisoft Watchdog's Games source code. The company suffered a second breach in 2022 that disrupted its game systems and services, end quote. Seriously, is this a situation of, as the old saying goes, you rob a bank because that's where the money is? Do gaming studios tend to pay ransoms, maybe, or is it that gaming studios have weirdly bad operational security? Because here's another one.

GTA5 source code appears to have been leaked on Discord, Telegram, a dark website, and others. A little over a year after hackers breached Rockstar. Bleeping computer again. Links to download the source code were shared on numerous channels, including Discord, a dark website, and a telegram channel that the hackers previously used to leak stolen rockstar data. In a post to a Grand Theft Auto leak channel on Telegram, the channel owner known as Phil posted links to the stolen source code sharing a screenshot of one of the folders. The channel owner also paid homage to Lapsis hacker Ariane Cortage, who previously leaked pre-release videos of Grand Theft Auto 6 under the name T-pot tuber hacker. Cortage was recently sentenced to an indefinite hospital stay by a UK judge for hacking into Rockstar and Uber.

hashtag free Ariang Cortage. He started all of this and ensured the leak would become public. I have immense respect for him. Miss you, buddy, end quote, said the Discord channel owner, Phil. What was the biggest mobile gaming app of 2023? Royal Match, which dethroned longtime App Store King Candy Crush from the top of the App Store. Royal Match held the top spot for the biggest mobile game by monthly revenue globally since July. quoting the Financial Times.

Launched in 2021, Royal Match is the debut title from Dream Games, a Turkish startup valued at $2.75 billion early last year. For more than a decade, King's Candy Crush saga has been one of the world's most consistently popular games on any platform hitting $20 billion in cumulative revenue this year. Now part of Microsoft after its $75 billion buyout of Activision Blizzard, Candy Crush has spent only six months outside the top 10 highest revenue mobile games since it was released in late 2012, according to Data AI. Consumer spending on Royal Match more than doubled in the year to October, increasing the game's annual gross revenue run rate before paying out

app store fees to $2 billion, said Sonor Idamir. Dream Games co-founder and chief executive. Royal Match is a Match 3 puzzle game, which would typically involve lining up tiles or icons to clear a grid. These have become the most popular casual gaming genre since they were popularized by bejeweled in the early 2000s. While it spawned, many imitators, Candy Crush Saga, came to dominate the Match 3 market, ranking number one by consumer spend on mobile app stores for nearly 127 consecutive months, according to Data AI, end quote. Royal Match has achieved remarkable player retention, apparently, with over 90% of its users continuing into their second year of gameplay. Its widespread appeal stems from its simple yet

engaging puzzle mechanics and an inviting storyline, distinguishing it from the typical fantasy battle or casino-style games that often lead in revenue on mobile. The game boasts around 55 million monthly active users and has managed to generate higher average spending from its players compared to Candy Crush, which has a larger base of about 160 million users, as reported by data AI. Dream games are optimistic about maintaining their momentum, unlike rivals such as Playwrix's gardenscapes and homescapes in 2020, these competitors briefly surpassed Candy Crush and sales only to fade back down again. Looking forward, Dream Games is preparing to build on its success by launching a new game Royal Kingdom next year, aiming to solidify its position in the mobile gaming market.

Finally today, President Biden declined to veto that USITC decision to ban Apple Watch series nine and Ultra 2 imports over a patent dispute. But what I was curious about is what this patent dispute is even all about. Turns out, this is a dramatic demonstration of the collision of different IP regimes when tech companies move into new markets, markets like health care. Massimo is a medical technology company, basically a lot of those patient monitoring devices and sensors you see pinging in the background in hospital rooms. They make those. They have a market cap of $6.24 billion, quoting the Financial Times. Massimo claimed Apple breached two of its patents on pulse oxymetry, a technology for monitoring the oxygen level in blood.

In the early days of smartphones, the convergence of mobile communications and computing produced a barrage of lawsuits between computing companies such as Apple and Google on the one hand and mobile technology companies including Nokia and Motorola on the other. Much the same is now happening on a broader front as mobile computing invades markets, including health care, where medical device companies have their own IP motes. Massimo has suffered setbacks in its campaign against Apple in other venues, including a case in federal court that resulted in a hung jury this year. but it had better luck before the International Trade Commission a body set up more than a century ago

to manage U.S. tariffs. The ITC has since morphed into a forum for deciding whether imports pose an unfair threat to domestic industry, a quaint-sounding anachronism in an age of multinational companies and global supply chains. The ITC has become a favorite venue for plaintiffs such as Massimo. Since a landmark 2006 patent case involving eBay, it has been far harder to persuade a federal court to grant an injunction barring the sale of products found to infringe on patent rights. By contrast, the ITC's only sanction is to issue exclusion orders that direct customs officers to block imports at the border. Analysts argue that the commission could wield this cudgel in less damaging ways. Others have proposed reforms, including strengthening the White House's ability to veto ITC bans

when a valid public interest is at stake. Large companies still have plenty of other weapons at their disposal as a parallel case involving a different medical sensor on the Apple Watch shows. The ITC last year issued another import ban against the device after a complaint from medical equipment maker AliveCorps. The case involved electrocardiogram technology, which measures electrical signals from the heart. That ban was put on hold after Apple challenged AliveCore's patents before a body known as the patent trial and appeal board. This was set up by the 2011 America Invents Act, the last major reform of the U.S. patent system to make it easier to challenge the validity of patents and tip the balance away from so-called patent trolls accused of abusing the system, end quote. So essentially,

the mechanism here is that since Apple has to import every device it sells from its overseas manufacturing, even though the patent issue is still outstanding, that's why they simply can't sell the watch right now. Quick, possibly dumb question. Why didn't Apple just buy Massimo? I mean, $6 billion market cap. Apple can afford that. And okay, they probably don't want to run a medical device company, though they could probably learn a lot from them if they do, as we assume, want to get more into health, right? But also, if that's Massimo's market cap, could the license Massimo had been asking for for the patents have been that insane? Again, Apple could have afforded possibly avoiding all of this mess. So it turns out if you long press on share or repost on threads,

you can quickly quote a post or access the native share sheet. Just kidding. But if you were on threads at all, and I mean at all, this past weekend, you'll get the joke. Talk to you tomorrow.