There can be no doubt that yesterday’s big Made By Google hardware event was designed to take the wind out of Apple’s sails.
Google typically unveils its latest Pixel smartphones in the fall, after Apple’s September iPhone reveal; we are very much still enjoying the summer. And Google missed no opportunity to praise the AI capabilities, powered by the company’s own Gemini models, in its new devices—and in other capable Android phones.
“Gemini is deeply integrated with Google apps and Android and can handle complex queries without hand-off to third-party AI providers you may not know or trust,” Google said pointedly in a blog post. It was a likely reference to the fact that Apple, still lacking a powerful AI model of its own, will have to rely on OpenAI and potentially other providers for the heavyweight AI functionality in its upcoming iPhones.
There’s an element of FUD in Google’s messaging here, given that Apple claims it in-house AI models are highly capable, and it offers privacy protections when handing queries over to third-party AI providers. But nonetheless, Google does hold a clear advantage over Apple by having developed its own models.
Google also stole a march on OpenAI by yesterday rolling out Gemini Live, its conversational chatbot, which offers very similar functionality to the voice-chat features for ChatGPT that OpenAI has demonstrated but not yet released. According to TechCrunch, Gemini Live is impressive, though it does make stuff up sometimes. The Verge says it’s prone to mansplaining.
Gemini Live will feature in the new Pixel Buds Pro 2 earphones, which now feature a variant of Google’s custom Tensor AI chip. This is apparently useful for better noise cancellation.
Of course, the stars of the hardware show were the handsets—the Pixel 9, Pixel 9 Pro, Pixel 9 Pro XL, and Pixel 9 Pro Fold—which run Google’s beefy Tensor G4 processor and come with not only Gemini Live, but also the multimodal version of the on-device Gemini Nano model. This should be able to handle image, audio, and speech inputs as well as text, and it will be very interesting to see how it stacks up against the on-device “Apple Intelligence” functionality in this year’s iPhones.
With Gemini Live coming to other Android handsets, including those from top manufacturer Samsung, Google’s AI efforts clearly have a lot of momentum and Apple’s response will need to be strong. Given that many Apple Intelligence features won’t be available on the upcoming iOS 18 when it first comes out, it looks like Google will have the upper hand for a while.
However, that doesn’t mean all is well in Google-land. A court found last week that Google’s had used its monopolization of the online search market to break antitrust law and, while the judge is yet to detail the remedies for this lawbreaking, Bloomberg now reports that the Justice Department may push for a Google breakup. DOJ insiders told the publication that they could seek the divestment of the Android operating system and Chrome browser, and/or Google’s AdWords platform.
I still can’t quite believe a Google breakup is really on the cards—the government did once try to do this to Microsoft and failed—but if Google were to lose control of Android, that could mean losing the primary vehicle for its AI ambitions. So stay tuned.
More news below.
David Meyer
Want to send thoughts or suggestions to Data Sheet? Drop a line here.
Update: This essay was updated on Aug. 15th to clarify the protections in Apple’s AI approach.
NEWSWORTHY
SpaceX pollution. Texas environmental regulators say Elon Musk’s SpaceX has been illegally discharging polluted wastewater into bodies of water in the state. Ecologists are highly concerned about the levels of mercury being discharged. But, as CNBC reports, the Federal Aviation Administration seems fine with SpaceX’s deluge system continuing. In other Musk news, his X social network has released version 2 of its Grok chatbot, which can now generate images—without industry-standard guardrails against creating images of political figures.
Chinese robo-taxis in Cali. Californian regulators have granted China’s WeRide permission to test its driverless vehicles with passengers, TechCrunch reports. WeRide can now test robo-taxis without a human safety driver present, but it still can’t offer paid services to public. In related news, The Verge reports that Waymo has updated its autonomous vehicles so they don’t repeatedly honk at each other in parking lots in the wee hours of the morning.
Big geofencing ruling. A U.S. appeals court with jurisdiction over Louisiana, Mississippi, and Texas has declared that geofence warrants are unconstitutional, TechCrunch reports. Geofence warrants let cops designate an area around a crime scene and demand that companies like Google hand over location data for every device falling into that area at the time of the incident. Friday’s Fifth Circuit opinion backs up those advocates who say geofence warrants allow over-broad searches of people’s location data—but a recent Fourth Circuit ruling went the other way, declaring that geofence warrants are legal in North Carolina, Virginia, and West Virginia.
ON OUR FEED
“[EU Digital Services Act] implementation is too important to be misused by an attention-seeking politician in search of his next big job.”
—An unnamed EU official tells Politico that the European Commission’s digital chief, Thierry Breton, may have gone too far in the letter he sent Elon Musk ahead of Musk’s X Spaces conversation with Donald Trump earlier this week. The letter warned Musk against violating the EU’s DSA by spreading misinformation, but X, the Trump campaign, and even some in the EU think it reeked of attempted interference in the U.S. election.
IN CASE YOU MISSED IT
Newsom pushes for school phone ban as social media concerns surge, by the Associated Press
X ordered to pay €550,000 to Irish employee fired for not replying to Elon Musk’s yes-or-resign ‘extremely hardcore’ ultimatum, by Bloomberg
Trump praised Elon Musk for firing striking workers, sparking a clash with the UAW, by the Associated Press
Microsoft’s AI Copilot can be weaponized as an ‘automated phishing machine,’ but the problem is bigger than one company, by Sage Lazzaro
Sam Altman warned OpenAI will ‘steamroll’ AI startups. I run one. Here’s why I’m not worried, by Eric Olson (Commentary)
Polish billionaire ‘Locker King’ weighs legal action against Meta over deepfake ads, by Bloomberg
Norway’s $1.7 trillion wealth fund made billions from tech in 2024—now it’s slashing its stake in Meta, Novo Nordisk and ASML, by Bloomberg
BEFORE YOU GO
Disney+ terms vs wrongful death lawsuit. A New York doctor died last year after waitstaff at a Disney Springs restaurant assured her that her order contained no dairy or nuts, and she died after eating food that did trigger her severe allergies. Now, Disney has reportedly asked for the dismissal of a lawsuit by her grieving widower, on the basis that he agreed to “arbitrate all disputes with the company” when he signed up for a Disney+ trial five years ago, and again when buying EPCOT tickets last year.
