Big Tech and Big AI
Google's Antitrust Case Focuses on the AI Future • Mozilla is Doomed without Google Money • Eye-Scanning Orbs Hit Our Shores • Record Everything, Let AI Sort It Out • Beyond Control AI Startups
I started out on Monday writing about Amazon’s stake in Anthropic, which I had written about a few months ago, and Amazon’s own financial disclosures seemed to confirm my ability to do some simple math: Amazon likely owns just over 22 percent of the AI startup (with Google owning another 14 percent). And that led me back down the path wondering about Microsoft’s stake in OpenAI, were it to be consummated into equity soon. And sure enough, just a few hours later, OpenAI announced they were dropping their bid to transition fully to a for-profit company, but also moving forward with shifting the for-profit arm to a PBC. The main hold up? The ownership stakes for the non-profit and yes, Microsoft.
More in the three posts below:
Big Tech’s Big Stakes in Big AI
As the valuations get bigger, the stakes go higher, quite literally...
OpenAI Attempts to Have Their For-Profit Cake and Eat It Too
OpenAI leaves the for-profit path, maybe, with a trail of questions...
Microsoft's OpenAI Equity Jigsaw Puzzle
With the non-profit maintaining a controlling stake and an FMV floor set by Elon Musk, OpenAI still has a hell of a challenge...
• Listening to "Kite" by U2 🎶
• Written on an M4 MacBook Air 💻
• Sent from London, England 🏴 I Think...
🤖 The A.I. Race Could Be Upended by a Judge’s Decision on Google
While the trial itself was about Google’s past web search dominance, it does feel like the remedy aspect has shifted pretty quickly to be focused on the future. That is, AI. On one side, Google would argue that AI — and ChatGPT, in particular — is naturally coming up as a competitive technology to Search (and argument which Mozilla also bolstered, below). On the other, various other competitors, including in the AI space, note that Google can’t be allowed to leverage their dominance in Search to “win” the AI arms race. Search aside — litigating in hindsight — this is the real key decision of the whole trial. Does the judge let the market decide how the AI race evolves, or does he restrict Google, even only slightly, to make sure they don’t have some sort of advantage given their already dominant products? Is there a way to put some future guardrails in place — for example, making sure Google doesn’t bake Gemini directly into Chrome (the dominant web browser being the other big element of the remedies right now)? One idea that has been thrown out a few times now seems like it would be absolutely wild scale-tilting: letting OpenAI buy Chrome. [NYT]
🦊 Firefox Could be Doomed Without Google Deal, Says Executive
This remains one of the bigger squares needed to be circled with regard to the remedies in the Google Search antitrust case: how can they limit Google’s search distribution deals without harming the businesses that rely on the other side of these payment equations? Apple is the one brought up most often given the sheer size of that payment, $20B+ a year, but no one is going to shed a tear for Apple if they lose that money. At the same time, Apple didn’t lose this case, and this is a meaningful amount of money — certainly of profit — to their business. But for Mozilla, the risk is existential. Firefox makes up 90% of their revenue, and 85% of that browser’s revenue came from the Google Search rev share agreement. You do the math. But beyond Mozilla CFO Eric Muhlheim testifying how bad the loss of this money would be for the company, he also helped Google’s argument in noting that a new AI player could come in to compete with Google naturally (i.e. without remedies from this case). Still, there would be a gap between that theoretical competitor coming up and Mozilla losing these payments that they probably couldn’t survive... [Verge]
👁️ Worldcoin, the Eye-Scanning Crypto Orb Thing, Comes to the US
On one hand, I appreciate the attempt to get ahead of a potential future problem in the form of AI content and technology everywhere. Perhaps even robots! On the other, a solution from the same person driving the “problem” — OpenAI’s Sam Altman — feels a bit self-dealing, if not self-perpetuating? On one hand, Face ID and the various other facial recognition systems from other phone makers are now commonplace, so perhaps a third-party system to take the concept further is okay, or at least inevitable. On the other, there’s both Black Mirror and Minority Report pointing to the obvious downsides here. And sure, those are fiction. But so is Her. You can’t really tout one and not be okay with the other being touted. Mainly, this whole thing being tied to a cryptocurrency just feels a bit ick. ‘Scan your eyeballs to our database for a chance to win prizes!’ That kind of ick. All in all, there’s just a very palpable, “what could possibly go wrong?” vibe here. At least the orbs themselves are smaller now? Also, I thought it was just “World” now? (It is. And that won’t be confusing, I’m sure.) [Verge]
🎙️ I Recorded Everything for Three Months. AI Replaced My Memory.
In other what-could-possibly go wrong surveillance news, I’ve been fascinated by these “record everything and let the AI sort it out” wearables for a while now. As Joanna Stern’s thoughts make clear, it’s still way too early for this kind of device to get anywhere near the mainstream. But within the enterprise, in work contexts, such as meetings, I could definitely seeing these gaining some traction, just as apps that already do this kind of thing have. But the real promise/potential here is the ability to create an almost second brain — one with flawless memory — that can be indexed and automated. This sounds creepy in 2025 because it is creepy in 2025. But in 2035, might we feel differently about such tech? I would not bet against it, depending on who is making such a product. Meta is seemingly doing well with their Ray-Ban smartglasses, and they have grander visions still, but as they go more mainstream, there will be speed bumps. Apple has an obvious advantage here thanks to consumer trust (not to mention hardware prowess) — trust which they’re slowly eroding with their own mishaps — but there might be a new opportunity for startups here. This current crop though, all feel way too early. [WSJ 🔒]
🧐 Thinking Machines Lab CEO Has Unusual Control
How hot are AI deals right now? So hot that not only are startups getting 10-digit valuations pre-product, they’re also apparently getting guaranteed total board control, before the board of directors is even formed. This effectively makes it more like a board of advisors. Which is honestly the case quite often in many companies, but again, usually not the earliest stage companies just getting off the ground, being run by a first-time CEO, no less. Fine, technically second time, if you count the three days or so where Mira Murati took over control of OpenAI for Sam Altman after that board ousted him (which she seemingly inadvertently played a major role in). And that perhaps speaks to exactly why she’s demanding this governance structure here. It’s just sort of wild that investors are okay with this. Look, if everything goes perfectly, it probably won’t matter. And how often do things go perfectly with startups? That would be never. Just ask OpenAI. [Information 🔒]
I Wrote...
The Year-Round iPhone
A staggered release schedule makes sense for a few reasons...
Good Vibrations Between Apple & Anthropic
It makes sense for the two to partner on "vibe coding" – might we hear more at WWDC?
I Note...
I knew that Apple was releasing a documentary with Bono, seemingly sort of in the vein of the old VH1 ‘Behind the Music’. But I missed that it’s the first feature-length content being released in the Apple Immersive format. Given that I liked Edward Berger’s first attempt at cinematic content for the Vision Pro, and more recently loved the Metallica three-song experience, I assume this will be good. Then again, Apple and U2 have a mixed track record together — boop. [9to5Mac]
Apple also quietly rolled out these sort of strange celebrity profile pages on their website. Ping 2.0? [9to5Mac]
Google is hosting some sort of Android event live on YouTube a week before Google I/O, which seems to indicate both how jam-packed the main event will be and just how focused on AI it will undoubtedly be. End of an I/O era? [Verge]
Revenue is falling and losses are rising at Meta’s Reality Labs division. As they race towards $100B in cumulative losses, they really need to ensure the actual cash cows aren’t disrupted — so far, so good. Meanwhile, those AR glasses can’t get here soon enough… [CNBC]
Snapchat has had to scrap a “simple” redesign for the second time, as the user base keeps rejecting it. Meanwhile, Wall Street is clearly rejecting Snap right now. It’s too bad, they’re doing interesting things, and while as a $15B company, they’re still ahead of what Zuckerberg offered, but they’re also well below their IPO price, five years later. How does this end? [Verge]
NotebookLM, Google’s fun/useful AI tool is getting its own stand-alone app for Android and iOS soon. Native AI apps, so hot right now — but shouldn’t this just be a part of the Gemini app? Are we already heading towards a world of 15 different Google AI apps? [9to5Google]
Paramount shelved a Daily Show-branded civic participation initiative out of fear the politics could further delay the deal with Skydance. I’m sure Jon Apple-wanted-to-censor-me Stewart will love this one. [Semafor]
Apple has officially filed a notice of appeal to the Ninth Circuit court of the epic smackdown they received from Judge Yvonne Gonzalez Rogers — even if they win on appeal, how on Earth are they going to roll the changes back without looking even worse? [Verge]
Meanwhile, Tim Sweeney says they’re bringing back Fortnite to the App Store (even though he previously said they’d only do it if Apple rolled out the new App Store changes worldwide — which they did not do). As for how they’ll do this (given that Epic’s US developer account remains suspended, and this ruling doesn’t change that), they’re going to use their Swedish account, and apparently ran this plan by Apple (but I would take this all with a pretty big grain of salt given the history here). [Xitter]
How many times was the word “tariffs” mentioned during Apple’s earnings call? 27 times. Amazon’s? 17 times. [NYT]
But hey, at least Tim Cook got the opportunity to thank President Trump for a bunch of policies that President Biden put in place. Same old, same old. [9to5Mac]
Also lost in the rest of the endless Apple news shuffle: they’re being ordered to pay a $502M in a patent dispute with a UK company which they say is just a patent squatter (so they’re appealing). [Reuters]
Because Satya Nadella clearly needed more ways to complicate his relationship with Sam Altman and OpenAI, Microsoft seems poised to host and offer xAI’s Grok models in their cloud. [Verge]
I Quote...
“In this update, we focused too much on short-term feedback, and did not fully account for how users’ interactions with ChatGPT evolve over time. As a result, GPT‑4o skewed towards responses that were overly supportive but disingenuous.”
— OpenAI, explaining the “sycophancy” screw-up that led them to roll-back an update to the GPT-4o model. I think it’s a pretty clear explanation, but what a weird world we live in. An AI world which also sometimes mirrors our own.
I Spy...
Oh hell yes. [via Daring Fireball via Saul Sutherland]