The Comunicano for Thursday May 1st 2025
Thinking Thursday
It appears that the age of machines not only continues—it accelerates.
In my lifetime, I bore witness to the landing of man on the moon. Today, I bear witness—if only in spirit—to the landing of the self: the delegation of human agency to intelligent systems, artificial though they may be.
Let us begin with UPS, the package delivery company, which is now exploring the use of humanoid robots. Imagine that—machines with legs and arms, assisting with packages instead of equations. This is not science fiction. It is science operationalized, mechanizing the mundane to solve a very human problem: labor shortages.
Then there is Meta—no longer just a company but seemingly a digital nation unto itself. It has released an AI assistant powered by Llama 4, a tool designed to accompany you across the vast plains of social media, solving problems and translating languages. Useful, yes. But always remember: tools shape their users.
Meanwhile, a new study accuses LMArena of permitting manipulation of AI performance benchmarks. When I speek of the importance of ethics in science, this is what I meant. A skewed benchmark is no different from a biased experiment. It tells us not what is real, but what is convenient.
Privacy, too, stands at a crossroads. Meta’s encrypted AI chats in WhatsApp are promising—but encryption alone does not equal trust. And Microsoft’s “Recall” feature—allowing PCs to remember everything—offers power, yes, but power without careful thought is a recipe for regret.
Amazon, Google, Pinterest, even Yelp—all now wield AI as their primary instrument. Yelp’s phone assistant books your table. Pinterest filters AI from your art. Apple, in contrast, finds itself chastised by the law—a reminder that even the most powerful must answer to principles, not profits. The essence of all this? We are building machines that think, systems that learn, agents that act. But we must not forget what it means to be human.
Our task, like always, is not just to ask Can we? but to ask Should we? In this vast, ever-expanding field of artificial intelligence, I implore you: proceed with wonder, but also with wisdom.
After all, the measure of progress is not how much we automate, but how well we elevate humanity in the process. Which is after all, why you are reading, THE COMUNICANO!!!
Andy Abramson
Robotics Watch
UPS Explores Humanoid Robots for Logistics Enhancement (Robotics & Automation News)—UPS is considering deploying humanoid robots from Figure AI to support warehouse operations. Initial trials are expected as part of a broader strategy to address labor shortages and enhance efficiency. The collaboration would involve the startup's humanoid robots becoming part of the shipping giant's day-to-day logistics operations. While discussions are ongoing, this move could signify a significant shift in warehouse automation for UPS. Read more here
Meta Watch
Meta Launches Standalone AI App Powered by Llama 4 (TechRepublic)—Meta has introduced a new AI assistant app powered by its Llama 4 model. Designed for personalized assistance, the app integrates with platforms like Instagram and Facebook, offering features such as problem-solving and daily navigation. Additionally, Meta's Ray-Ban smart glasses now support offline translation, enhancing their utility for travelers and international communication. Read more here
Study Accuses LMArena of Helping Top AI Labs Game Its Benchmark (TechCrunch)—A recent study by researchers from Cohere, Stanford, MIT, and AI2 alleges that LMArena allowed major AI companies to manipulate benchmark results. Companies like Meta, OpenAI, Google, and Amazon reportedly tested multiple model variants privately and only published top-performing scores. Meta's submission of an optimized Llama 4 version, differing from its public release, has drawn particular scrutiny. Read more here
WhatsApp Tests Encrypted AI Conversations (Wired)—Meta is testing a new privacy-focused AI chat feature in WhatsApp that maintains full encryption and avoids data exposure. Conversations are processed securely on remote servers with no access from Meta or WhatsApp, aiming to merge the convenience of AI with WhatsApp’s commitment to end-to-end encryption. Read more here
Microsoft Watch
Microsoft's Controversial Recall AI Feature Now Available on All Copilot+ PCs (TweakTown)—Microsoft's Recall feature for Copilot+ PCs, which lets users search through snapshots or screenshots of their PC use history, is now available. Despite initial privacy concerns, Microsoft has implemented enhanced security measures, making the feature opt-in rather than enabled by default. Users can now utilize this AI-powered photographic memory to find and jump back into previous activities on their PCs. Read more here
AI Watch
Yelp Introduces AI Assistant for Restaurant Calls (TechCrunch)—Yelp has launched an AI-powered phone assistant designed to handle restaurant reservations and customer inquiries. This feature aims to streamline operations for restaurant owners by automating tasks such as booking tables and answering common questions about hours and services. The assistant integrates with Yelp's existing platform, providing a seamless experience for both businesses and customers. Read more here
Pinterest Rolls Out AI Image Labels and Filters for Greater Transparency (The Verge)—Pinterest is stepping up its efforts to manage AI-generated content by launching new features that clearly label and filter such images. Images created or modified using generative AI will now carry a visible "AI modified" tag in the bottom-left corner of each pin. This labeling uses a mix of metadata analysis and proprietary classifiers, with an appeal option for wrongly flagged content. Additionally, Pinterest is introducing a "see fewer" toggle to help users limit AI content in visually saturated categories like beauty and art. These tools aim to enhance user trust and give more control over the browsing experience. Read more here
Alibaba Unveils Qwen 3 Open-Source AI Model (Reuters)—Alibaba has released Qwen 3, an open-source large language model series that strengthens multilingual processing, especially in Mandarin. The Qwen 3 family offers a viable alternative to Western proprietary models, with strong reasoning and understanding capabilities. This release signals a strategic push by Alibaba to compete globally in the generative AI space. Read more here
Apple Watch
The Core Rot: Judge Skewers Apple for 'Willful' App Store Violations (Reuters)—A federal judge has ruled that Apple violated her 2021 injunction requiring greater App Store competition, referring the company to prosecutors for possible criminal contempt. Judge Yvonne Gonzalez Rogers stated, "Apple's continued attempts to interfere with competition will not be tolerated," adding, "This is an injunction, not a negotiation." The ruling is a major win for Epic Games in its long-running battle against Apple's App Store policies. Read more here
Google Watch
Channel Surfing: Google Finally Lifts the Curtain on Performance Max (Google Blog)—Google is introducing channel-level reporting to Performance Max campaigns, giving advertisers unprecedented visibility into how different platforms contribute to their campaigns. Google's announcement explains that the new feature "helps you understand how your Performance Max campaign is delivering results across Google's full range of channels and inventory." The beta will launch alongside full search terms reporting and expanded asset performance metrics this month. Read more here
Amazon Watch
From Tokens to Triumph: Amazon's Nova Premier Joins the AI Big Leagues (Techcrunch)—Amazon has released Nova Premier, its most advanced AI model to date, featuring a massive 1 million token context window and multimodal capabilities. TechCrunch notes Nova Premier "can analyze around 750,000 words in one go" though it "is weaker on certain benchmarks than flagship models from rival AI companies such as Google." The model is priced competitively with other leading offerings at $2.50 per million input tokens. Read more here
Fintech Watch
Visa Develops AI-Ready Credit Cards for Autonomous Shopping (AP News)—Visa is rolling out a new generation of credit cards equipped to interact with AI shopping assistants. These AI-ready cards will allow automated purchases within user-defined rules, providing secure transactions while maintaining consumer control. Visa’s move highlights a future where digital agents could handle everyday purchases on your behalf. Read more here
AgenticAI Watch
AgenticFlow Launches No-Code AI Agent Platform (Martech Zone)—AgenticFlow has debuted a no-code platform for creating AI agents, allowing businesses to automate workflows without technical expertise. The tool enables users to build, deploy, and manage agents for tasks like customer support and operations. It’s part of a larger trend making AI accessible to non-engineers across industries. Read more here
Edutech Watch
Duolingo Transitions to AI-Generated Content (UPI)—Duolingo is phasing out human content contributors in favor of AI-generated lessons. The language-learning platform is using AI to create exercises, explanations, and quizzes, which allows for rapid scaling and content personalization. The shift has led to contractor layoffs but reflects broader adoption of generative AI in education. Read more here
Circus Watch
Washington Makes History: Wild Animal Circus Ban Becomes Law (aldf.org)—Washington Governor Ferguson signs legislation prohibiting elephants, bears, wild cats, and nonhuman primates from being used in traveling circuses and animal acts across Washington State. The bipartisan-supported Senate Bill 5065, championed by Senator Marko Liias, makes Washington the 12th state to enact such protections. Animal welfare advocates celebrated the law as ending cruel transportation and performance conditions for these animals, particularly benefiting elderly elephants like Viola and Isa who have been forced to perform despite suffering from mobility issues. Read more here
