Tech Brew Ride Home - Thu. 03/27 – AI Studio Ghibli

Starting point is 00:00:00 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 Thursday, March 27th, 2025. I'm Brian McCullough today. Everyone is using that new chat GPT image generator to make Miyazaki memes.

Starting point is 00:00:44 Google is taking Android development private. Are we seeing more signs of an AI data center pullback? Now even you can do extra gig work for Instacart. And quantum computers seem to have cracked truly random, random number generators. Here's what you miss today in the world of tech. I guess a lot of people had the... same impulse I did after OpenAI released that new chat GPT image generator yesterday, social media has been flooded with AI memes in the style of Studio Ghibli, highlighting copyright concerns,

Starting point is 00:01:20 quoting TechCrunch. In the last 24 hours, we've seen AI-generated images representing Studio Ghibli versions of Elon Musk, the Lord of the Rings, and President Donald Trump. OpenAI CEO Sam Altman even seems to have made his new profile picture a Studio Ghibli-style image, presumably made with GPT4O's native image generator. Users seem to be uploading existing images and pictures into chat GPT and asking the chatbot to recreate it in the new styles. OpenAIs and Google's latest tools make it easier than ever to recreate the styles of copyrighted works simply by typing a text prompt.

Starting point is 00:01:53 Together, these new AI image features seem to reignite concerns at the core of several lawsuits against generative AI model developers. If these companies are training on copyrighted works, are they violating copyright law? According to Evan Brown, an intellectual property lawyer at the law firm Neil and McDevitt. Products like GPD40's native image generator operate in a legal gray area today. Style is not explicitly protected by copyright, according to Brown, meaning open AI does not appear to be breaking the law simply by generating images that look like Studio Ghibli movies. However, Brown says it's plausible that Open AI achieve this likeness by training its model

Starting point is 00:02:28 on millions of frames from Ghibli's films. Even if that was the case, several courts are still deciding whether training AI models on copyrighted works falls under fair use protections. I think this raises the same question that we've been asking ourselves for a couple years now, said Brown in an interview, what are the copyright infringement implications of going out, crawling the web, and copying into these databases, end quote. Quoting variety, Heyo Miyazaki, the co-founder of Studio Ghibli, has previously expressed strong disapproval of AI-generated animation. In a 2016 meeting where he was shown in AI animation demo, Miyazaki said, I am utterly disgusted. If you really want to make creepy stuff, you can go ahead and do it.

Starting point is 00:03:08 I would never wish to incorporate this technology into my work at all. He also said, I strongly feel that this is an insult to life itself. More recently, Hollywood actors and other creatives have voiced concerns about efforts by Open AI and other artificial intelligence companies to, quote, weaken or eliminate protections on copyrighted works for training AI systems. In comments filed with the Trump administration's Office of Science and Technology Policy earlier this month. More than 400 filmmakers, actors, musicians, and others objected to what they said was lobbying by OpenAI and Google, quote, for a special government exemption so they can freely exploit America's creative and knowledge industries, end quote.

Starting point is 00:03:45 Nobody tell Wes Anderson about my Jack Dorsey pick from yesterday. Google says, going forward, it plans to develop Android fully in private to streamline the development process, but will continue to publish the source code for Android for new releases. quoting Android Authority. The Android Open Source Project, or AOSP, for short, is an operating system that Google releases under the Apache 2.0 license. Apache 2.0 is a software license that allows anyone to use, distribute, or modify, and distribute operating systems based on AOSP without the need to pay any licensing fees or release source code. This permissive licensing structure has facilitated the widespread adoption of AOSP, leading to the creation of customized forks like Samsung's OneUI. Like many other open source projects, AOSP accepts code contributions from third-party developers.

Starting point is 00:04:42 However, Google conducts most AOSP development itself, as it treats the Android project as a full-scale product development operation to ensure the vitality of Android as a platform and as an open-source project. Therefore, Google has the final say on what code can be merged into AOSP and when new version source code is released. The company develops AOSP components privately to allow developers and OEMs to use a single version of Android without tracking unfinished future work just to keep up. To balance AOSP's open nature with its product development strategy, Google maintains two primary Android branches, the public

Starting point is 00:05:16 AOSP branch and its internal development branch. The AOSP branch is accessible to anyone, while Google's internal branch is restricted to companies with a Google Mobile Services licensing agreement. Because Google develops large portions of Android in its internal branch, the public AOSP branch often and lags far behind what's available privately. This difference is apparent when comparing feature and API availability between a clean AOSP build and Google's latest Android 16 beta, which was built from its internal branch. While the shift to trunk-based development reduced this discrepancy, it persists and continues to pose challenges for Google. This discrepancy forces Google to spend time and effort merging patches between the public AOSP branch and its internal branch.

Starting point is 00:05:59 Due to how different the branches are, merged conflicts often arise. But beginning next week, all Android development will occur within Google's internal branch, and the source code for changes will only be released when Google publishes a new branch containing those changes. As this is already the practice for most Android component changes, Google is simply consolidating its development efforts into a single branch. Google confirmed to Android Authority that it is committed to publishing Android source code, so this change doesn't mean that Android is becoming closed source. The company will continue to publishes a source code for new Android releases,

Starting point is 00:06:32 so when Google releases Android 16 later this year, we'll get the source code for the update. In addition, Google will continue to publish the source code for Android's Linux kernel fork as it is licensed under GPLV2, which mandates source code releases and is separate from AOSP. What will change is the frequency of public source code releases for specific Android components. Some components like the build system, update engine, Bluetooth stack, virtualization framework, and Linux configuration are currently AOSP first, meaning their development. fully in public. Most Android components like the core OS framework are primarily developed internally, although some features such as the unlocked-only storage API are still developed

Starting point is 00:07:12 within aOSP, end quote. From the do we have overcapacity in the AI space file, some worrying signs. According to TD Cohen, Microsoft canceled new data center projects in the U.S. in Europe that would have amounted to around two gigawatts of capacity, quoting Bloomberg. Microsoft retrenchment in the last six months included lease cancellations and deferrals, the TD Cohen analyst said in their latest research note dated Wednesday. Alphabet's Google had stepped in to grab some leases Microsoft abandoned in Europe, the analyst wrote, while meta-platforms had scooped up some of the freed capacity in Europe. Microsoft says it remains on track to spend about $80 billion building out AI data centers in its fiscal year that ends in June. Executives have said the pace of growth should slow in the company's next fiscal year. Earlier this week, Alibaba Group holding Chairman Joe Tsai warned of a potential bubble in data center construction, saying new projects may exceed demand for AI services. We continue to believe the lease cancellations and deferrals of capacity points to data center oversupply relative to its current demand forecast.

Starting point is 00:08:21 TD Cohen analysts Michael Elias, Cooper Bellinger, and Gregory Williams wrote, end quote. And then there's this. Sources say many new Chinese AI data centers are currently sitting unused due to weak demand. and deep-seek-driven shifts. Local reports say around 80% of new computing resources are idle, quoting MIT Technology Review. Just months ago, a boom in data center construction was at its height, fueled by both government and private investors. However, many newly built facilities are now sitting empty. According to people on the ground who spoke to MIT Technology Review, including contractors, an executive at a GPU server company and project managers,

Starting point is 00:09:00 most of the companies running these data centers are struggling to stay afloat. Renting out GPUs to companies that need them for training AI models, the main business model for the new wave of data centers, was once seen as a sure bet. But with the rise of deep seek and a sudden change in the economics around AI, the industry is faltering. The growing pain China's AI industry is going through is largely a result of inexperienced players, corporations and local governments jumping on the hype train, building facilities that aren't optimal for today's need, says Jimmy Goodrich, senior advisor for technology at the Rand Corporation. Upshot is that projects are failing, energy is being wasted, and data centers have become distressed assets whose investors are keen to unload them at below market rates. The situation may eventually prompt government intervention, he says. The Chinese government is likely to step in, take over, and hand them off to more capable

Starting point is 00:09:49 operators, end quote. But good news on the silicon front, sources say that after taking five years to build its first U.S. plant, TSM expects new factories to take two years. A 3-nanometer plant is set for 2028 and a 2-nanometer before 2030. So, U.S.-made chips for your iPhone are coming down the pike, older ones at least, quoting 9-5 Mac. Apple first announced its plan for Made in America chips back in 2022, with the news hailed as one of the success stories of the U.S. Chips Act. The initiative will see a series of TSM chipmaking plants built in Arizona, with some of the production reserved for Apple chips for older devices. The project has been beset with delays and uncertainties.

Starting point is 00:10:40 Mass production at the first plant was due to start last year and was then pushed back to this year. Questions were raised about the promise U.S. jobs, with many employees recruited from Taiwan and the company even being accused of anti-American discrimination. The first TSMC plant is only capable of making larger processed chips used in older Apple devices. This includes the A-16 chip first developed for the iPhone 14 Pro, which is no longer made, and the base model iPhone 15 models, which are still available from Apple but likely sell in small volumes. If it takes years to bring a new U.S. plant online, that would mean U.S. made Apple chips would forever be relegated to significantly older devices.

Starting point is 00:11:17 However, NK. Asia reports that TSM has promised investors that the sluggish pace seen with the development of its first U.S. plant will not be repeated. While that first plant took five years to build and open, the company says that future U.S. plants will take no more than two. That will see a 3-nanometer plant brought online in 2028 and a 2-nanometer one before 2030. The U.S. plants will still lag behind the cutting-edge ones in Taiwan, meaning they will never make chips for the latest Apple products, but will at least close the gap, end quote. Want to make Instacart work for you by kind of sort of working for Instacart?

Starting point is 00:11:56 The company is rolling out StoreView, paying shoppers to take videos of store shelves one aisle at a time to assess inventory, starting with select retailers. Quoting the verge. Instacart says store view will also help its predictive models estimate when a store will restock a particular product. If you place an order at night for early next day delivery, we can predict whether the item is likely to be back on the shelves when the store reopens and then verify that information in the morning, Instacart explains. In the future, Instacart says its smart caper carts will have the ability to keep track of stock two. The outward facing camera on the caper carts will soon give stores an understanding of inventory down to the store level, allowing Instacart to update stock levels.

Starting point is 00:12:37 as often as every hour of the day. Along with Storeview, Instacart is rolling out second store check in the coming months, which Instacart bills as a way to source out-of-stock items from another location of the retailer you select. If you order an item but a shopper can't find it in store, Instacart will automatically prompt a second shopper to check if it's available at another store. The second shopper will then deliver the item for no additional service or delivery fee, but it's not clear whether you can tip them, end quote. And quoting Business Insider,

Starting point is 00:13:06 You'll get paid to take photos of what's in stock or refill displays, a message sent to a shopper about a beta version of the program in November said, which was seen by Business Insider. One shopper in Pennsylvania, who didn't want to be identified for fear of retaliation from Instacart told B.I. that they completed one of the tasks, which involved taking a photo of a display of Dove Body Care Products, which Unilever makes. Unilever did not respond to a request for comment. B.I. verified the shopper's identity and employment by Instacart. The gig paid about $12 and took about 10 minutes to complete, according to the shopper. That's more than an Instacart pays to shop and deliver some orders, which can easily take an hour the shopper added, end quote.

Starting point is 00:13:50 Finally today, JP Morgan researchers say they have generated and certified truly random numbers using a quantum computer, a world first with potential security and trading uses. Quoting Bloomberg, researchers created the sequence using a quantum computer built by Honeywell's Quantinium, according to a paper published in the Scientific Journal Nature on Wednesday. J.P. Morgan researchers alongside Argon and Oak Ridge National Laboratories and the University of Texas at Austin then became the first to prove mathematically that they had produced genuine randomness. Most so-called random number generators, which are important for encrypting sensitive data, aren't actually random. They're predetermined sequences. Computers run on a set of programmed mathematical operations that will always return the same answer,

Starting point is 00:14:32 raising the risk that hackers with access to increasingly sophisticated computer power could crack encryption codes. Other companies have produced random number generation already. Quantinium sells its products commercially for use in data centers and smart meters. But the certification gives users verifiable proof that the numbers are truly random, meaning higher stakes industries like critical infrastructure and financial services, can start using applications. It would be good for anything that has an audit trail to show the numbers are definitely randomly generated.

Starting point is 00:15:01 For example, the lottery, said Konstantinos Karagianis, director of quantum computing services at consulting firm Protaviti, who did not participate in the experiment. In cryptography, the provability is the whole thing. It's provably secure or it's provably not secure. There's no gray area. Between May 23 and May 2024, cryptographers at J.P. Morgan wrote an algorithm for a quantum computer to generate random numbers, which they ran on quantum's machine. The U.S. Department of Energy supercomputers were then used to test whether the output was truly random. It's a breakthrough result, project, and head of global technology applied research at J.P. Morgan, Marco Pistoia, told Bloomberg in an interview,

Starting point is 00:15:41 the next step will be to understand where we can apply it. Applications could ultimately include more energy-efficient cryptocurrency, online gambling, and any other activity hinging on complete randomness, such as deciding which precincts to audit in elections, end quote. Nothing more for you today. Talk to you tomorrow.

Tech Brew Ride Home - Thu. 03/27 – AI Studio Ghibli

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.