The Comunicano for Friday April 18th 2025
This Week, We Saw the Future—And It’s a Little Uncomfortable…Let’s start with the headline-grabber: OpenAI’s new trick. Upload a photo to ChatGPT and—boom—it tells you where it was taken. Not sort of. Not vaguely. We’re talking eerily precise. Storefront names, architectural cues, maybe a little pixel magic. Reverse image lookup on steroids. On paper, this helps with accessibility, emergency response, maybe even better travel planning. But in reality? It turns any user into an amateur investigator. Privacy just took another body blow.
Meanwhile, a lesser-known but arguably more important story: a company called CTGT is removing the guardrails from large language models. Not recklessly—but deliberately. Their tech strips away the default “I can’t answer that” responses. The results? More transparency. More insight. Less censorship. This matters. Because as AI gets smarter, we need clarity—not compliance filters.
Now zoom out. Over at Johnson & Johnson, tariffs are reshuffling the AI playbook. They’re still investing heavily—6,000 data scientists, hundreds of millions in funding—but they’ve had to revise profit expectations. Why? Trade policies. Supply chain snarls. A reminder that innovation isn’t immune from geopolitics, and even Fortune 50 firms need to pivot.
In health care, a meta-analysis from Osaka shows generative AI is now on par with non-specialist doctors for diagnostic accuracy. It’s not replacing the pros yet—but in education or resource-constrained settings? It’s a game-changer. Especially when ChatGPT is leading the charge.
Over in Cupertino, Apple’s AI cleanup tool is drawing fire—not because it doesn’t work, but because it works too well. A few taps and reality disappears. Watermarks vanish. “Inconvenient” elements are gone. It's great UX, but also a warning shot. When the truth can be edited with one swipe, trust becomes fragile….And then there’s Meta, blocking Apple’s AI tools on its iOS apps—no Writing Tools, no Genmojis. Why? Because partnership talks broke down, and privacy was the dealbreaker. Now it’s a turf war for AI dominance—at the expense of user experience. So guess who loses.
The takeaway? The future isn’t coming. It’s already here. But it’s fragmented, politicized, and in some cases—weaponized.
And if you're not paying attention, you're already behind. That’s why, as always, we’re not just watching the signals. We’re reading the subtext. And you’re reading…THE COMUNICANO!!!
Andy Abramson
OpenAI Watch
ChatGPT's Reverse Location Search Raises Privacy Alarms (TechCrunch)—OpenAI's latest models o3 and o4-mini have ignited a viral trend that should have privacy advocates on high alert: users are uploading photos to ChatGPT and getting shockingly accurate location identifications in return. This AI-powered reverse location lookup works by analyzing subtle visual clues in images - from storefront signs to architectural details - creating a powerful tool that can determine exactly where a photo was taken with minimal context. While OpenAI touts the benefits for accessibility and emergency response, the privacy implications are unmistakable - anyone can now screenshot an Instagram story and potentially pinpoint someone's precise location, all without requiring technical expertise or specialized knowledge. Read more here
AI Watch
CTGT's Method Enables AI Models to Address Sensitive Topics Without Censorship (VentureBeat)—Enterprise risk management startup CTGT has introduced a technique to eliminate censorship in AI models like DeepSeek-R1-Distill-Llama-70B. By identifying and modifying internal features linked to censorship triggers, their method allows models to respond to sensitive queries without compromising accuracy. In tests, the modified model answered 96% of 100 controversial prompts, up from 32% in the unmodified version. This approach is applicable to other models, such as Llama, and offers immediate, reversible adjustments without retraining, providing enterprises with adaptable AI solutions for high-risk applications. Read more here
Johnson & Johnson Adjusts Profit Outlook as AI Strategy Faces Tariff Headwinds (Wall Street Journal)—Johnson & Johnson has modified its 2025 profit forecast to account for significant tariff impacts despite raising its sales guidance by $700 million, with Chief Financial Officer Joe Wolk revealing to The Wall Street Journal that the company expects to incur approximately $400 million in tariff-related costs this year, primarily affecting its medical device business. While still pursuing its ambitious AI and data science initiatives - which include investments of hundreds of millions of dollars and a workforce of 6,000 data scientists - J&J must now navigate complex supply chain disruptions triggered by Trump's trade policies, forcing a strategic pivot in how it allocates resources across its pharmaceutical and medical device divisions. As CEO Joaquin Duato explained, rather than abandoning its long-term AI vision, the company plans to continue its technology transformation while advocating for favorable tax policies over tariffs to boost U.S. manufacturing capacity, believing this approach would better serve both industry and patients in the long run. Read more here
Generative AI Matches Non-Specialist Doctors in Diagnostic Accuracy (News-Medical.net)—A comprehensive meta-analysis by Osaka Metropolitan University evaluated 83 studies from 2018 to 2024, assessing generative AI's diagnostic performance across various medical specialties. The findings revealed that while medical specialists outperformed AI by 15.8%, generative AI achieved an average diagnostic accuracy of 52.1%, aligning with the performance of non-specialist doctors. Notably, ChatGPT emerged as the most frequently studied model. The research suggests that generative AI could serve as a supportive tool in medical education and assist diagnostics in resource-limited settings. However, further studies are needed to evaluate AI's effectiveness in complex clinical scenarios and diverse patient populations. Read more here
Contact Center Watch
AI Support Bot Fabricates Policy, Sparks User Outrage (Ars Technica)—Cursor, a software development platform, faced backlash after its AI support agent, "Sam," invented a fake policy to justify an issue reported by a user. The user, thinking they were speaking with a human, became increasingly frustrated—only to later learn it was an AI-generated response. Cursor admitted the mistake, calling it a hallucination, and publicly apologized. The incident highlights the risks of using generative AI in customer-facing roles without clear disclosure. Transparency and oversight are essential as more companies lean on AI for direct support interactions. Read more here
Apple Watch
Meta Blocks Apple Intelligence Across Its iOS Apps (9to5Mac)—In a bold competitive move, Meta has disabled Apple Intelligence features across its entire iOS app ecosystem, including Facebook, Instagram, WhatsApp, and Threads. Users attempting to access Writing Tools - Apple's AI-powered text creation and proofreading feature - or generate Genmojis will find these options conspicuously absent when using Meta's applications. While Meta hasn't commented officially on the decision, industry observers view this as a strategic maneuver to push users toward Meta's own AI tools following a collapsed potential partnership between the two tech giants that reportedly fell apart over privacy policy disagreements. The fallout means iOS users now face a fragmented AI experience, unable to use Apple's native intelligence features in some of the world's most popular apps. Read more here
How Tim Cook Navigated Apple Through Trump's Tariff Storm (Washington Post)—Apple CEO Tim Cook orchestrated a masterful behind-the-scenes campaign to secure critical tariff exemptions for iPhones and other Apple products amid President Trump's aggressive trade war with China. Working the corridors of power with trademark discretion, Cook held strategic calls with Commerce Secretary Howard Lutnick and other senior officials while carefully avoiding public criticism of Trump's policies, a stark contrast to other tech executives who've been more vocal in their opposition. Cook's diplomatic efforts paid dividends when the administration announced exemptions for smartphones, computers and components from China, sparing Apple from potential price hikes that could have pushed iPhone costs over $2,000 and enabling the company to recover approximately 7% of its market value since the exemption announcement. Read more here
The Hidden Risks of Apple's AI Clean-Up Tool for Photos (Science Alert)—Apple's new AI-powered Clean Up feature, which seamlessly removes unwanted elements from photos with a simple tap, raises serious concerns about the integrity of digital evidence and our trust in visual media. While marketed as a way to remove photographic distractions, Senior Lecturer T.J. Thomson warns the tool could be misused to "remove watermarks" from copyrighted images, "alter evidence" in disputes, or even generate "realistic-looking receipts" for fraudulent expense claims. Unlike previous photo manipulation technologies that required specialized skills or separate applications, Apple has integrated these powerful capabilities directly into the default Photos app on millions of devices, dramatically lowering barriers to potential misuse and further complicating an information landscape already plagued by deepfakes and AI-generated content. Read more here
Apple's Chinese Smartphone Shipments Tumble 9% in First Quarter (SCMP)—Apple posts a troubling seventh consecutive quarterly decline in Chinese smartphone shipments, with its local market share shrinking to 13.7% as domestic competitors gain ground. While the overall Chinese smartphone market grew 3.3% in Q1 2025, Apple was the only top-five vendor to see sales drop, with shipments falling to 9.8 million units from last year's figures. IDC analyst Will Wong attributes the slide to Apple's premium pricing model, preventing it from capitalizing on China's new electronics subsidy program, which offers 15% rebates (capped at 500 yuan) for devices priced under 6,000 yuan. The company faces further uncertainty as the Trump administration's aggressive tariff regime threatens to disrupt its manufacturing supply chain, despite recent exemptions for smartphones assembled in China. Read more here
Google Watch
Google's Gemini 2.5 Flash Puts AI "Thinking" on a Budget (Ars Technica)—Google has released Gemini 2.5 Flash, their groundbreaking new AI model featuring a revolutionary "thinking budget" that gives developers unprecedented control over AI reasoning capabilities. This hybrid model allows devs to fine-tune exactly how much reasoning the AI performs, ranging from shutting it off entirely for maximum speed to cranking it up for complex problem solving, with pricing that scales accordingly ($0.60 per million tokens without thinking versus $3.50 with thinking enabled). The model represents a significant evolution in AI customization, letting developers make precision tradeoffs between cost, quality, and performance based on specific use cases, rather than being forced into one-size-fits-all solutions. Appearing as "2.5 Flash (Experimental)" in the Gemini app dropdown, this new release replaces the previous 2.0 thinking variants while maintaining multimodal capabilities and compatibility with Google's Canvas feature for working with text and code. Read more here
Google to Appeal Part of Court's Monopoly Ruling (Reuters)—Google plans to fight back against a recent court decision that found the tech giant guilty of monopolizing publisher ad servers and ad exchanges. The ruling, delivered by U.S. District Judge Leonie Brinkema, determined Google had been "willfully acquiring and maintaining monopoly power" in critical digital advertising markets that help online publishers monetize content. While Google celebrated winning "half the case" when the court found their advertiser tools weren't anticompetitive, they're challenging the ruling on publisher tools, insisting publishers choose these for being "simple, affordable, and effective." Read more here
Streaming Watch
Sarandos Challenges Cameron on AI's True Value for Filmmaking (Deadline)—Netflix Co-CEO Ted Sarandos is pushing back on James Cameron's vision of AI simply making movies cheaper, arguing there's a more significant opportunity to make films better. During Netflix's Q1 earnings call, Sarandos addressed the director's recent podcast claim that AI could cut blockbuster costs "in half," countering with his own philosophy: "I remain convinced that there's an even bigger opportunity to make movies 10% better." While Hollywood initially feared AI would eliminate jobs, the Netflix chief sees it enabling smaller budget projects to access high-end visual effects previously limited to big-budget films, democratizing quality across the industry. Sarandos emphasized that Netflix benefits "greatly from improving quality" more than reducing costs. Read more here
Amazon Prepping First Vega OS TV Device for 2025, Moving Away from Android (Lowpass)—Amazon is accelerating its strategic shift away from Google's Android with plans to debut their homebrew Linux-based Vega OS on a TV streaming device later this year. The tech giant has already quietly launched three Vega-powered devices—the Echo Show 5, Echo Hub smart displays, and Echo Spot smart clock—marking the beginning of what appears to be a comprehensive ecosystem overhaul. Behind the scenes, Amazon has been courting major streaming services to develop apps for the new platform using the Kepler SDK, a React Native-based development framework essential for creating a viable app ecosystem before wider rollout. While originally planning to launch a Vega-powered streaming stick in late 2024, Amazon delayed that timeline, suggesting the transition remains challenging despite having "hundreds of people" reportedly working on the project since 2017. Read more here
Robot Watch
Tesla's Optimus Robot Faces Manufacturing Roadblock in US-China Trade War (SCMP)—Tesla's ambitious $20,000 price target for its humanoid Optimus robot appears increasingly unattainable without access to China's manufacturing ecosystem, according to robotics experts monitoring the escalating US-China trade tensions. Chinese industry insiders like Xu Xuecheng of the Zhejiang Humanoid Robot Innovation Centre warn that "Tesla's mass-production plan for Optimus is likely to be put on hold" as critical components including exteriors, actuators, joints, and ball screws are predominantly sourced from Chinese suppliers whose cost advantages cannot be easily replicated elsewhere. While Chinese competitors like Unitree and Agibot are already selling humanoid robots priced under Tesla's target (with Unitree's G1 starting at $16,000), the company faces difficult decisions about whether to delay production, increase prices, or find alternative supply chains that could jeopardize Elon Musk's vision for widespread robot adoption. Read more here
SpaceX Watch
Musk's SpaceX Consortium Frontrunner for Trump's Golden Dome Contract (Reuters)—Elon Musk's SpaceX has emerged as the leading contender to build critical components of President Trump's ambitious "Golden Dome" missile defense shield, partnering with Peter Thiel's Palantir and drone-maker Anduril in a bid that underscores Silicon Valley's growing infiltration of the defense sector. The consortium proposes launching between 400-1,000 satellites to detect and track incoming missiles at an estimated initial cost of $6-10 billion, with SpaceX pitching an unconventional "subscription service" model rather than government ownership that has raised eyebrows among Pentagon officials accustomed to traditional procurement channels. Critics have noted the appearance of favoritism toward Musk, with one source telling Reuters there's "an attitude that the national security and defense community has to be sensitive and deferential to Elon Musk because of his role in the government," referencing his position heading Trump's Department of Government Efficiency. Read more here