<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Paulina (Paula) Pluta on Medium]]></title>
        <description><![CDATA[Stories by Paulina (Paula) Pluta on Medium]]></description>
        <link>https://medium.com/@paula.pluta?source=rss-3d135aa078c9------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*XFIZKw8jXpWd_8o4kzdqFw.jpeg</url>
            <title>Stories by Paulina (Paula) Pluta on Medium</title>
            <link>https://medium.com/@paula.pluta?source=rss-3d135aa078c9------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 12:32:40 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@paula.pluta/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Invisible Protocol: Why Ambient AI Needs Human-Side Infrastructure]]></title>
            <link>https://medium.com/@paula.pluta/the-invisible-protocol-why-ambient-ai-needs-human-side-infrastructure-9fbb3c1de825?source=rss-3d135aa078c9------2</link>
            <guid isPermaLink="false">https://medium.com/p/9fbb3c1de825</guid>
            <category><![CDATA[technology-ethics]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[wearable-technology]]></category>
            <category><![CDATA[humancomputer-interaction]]></category>
            <dc:creator><![CDATA[Paulina (Paula) Pluta]]></dc:creator>
            <pubDate>Sun, 18 Jan 2026 23:40:09 GMT</pubDate>
            <atom:updated>2026-01-18T23:40:09.897Z</atom:updated>
            <content:encoded><![CDATA[<h4><strong>What if every wearable around us were a virtual assistant? Is there a universal sign for “P<em>lease don’t model my behavior today”</em>?</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qNFSZaqc3-uaLq1gEdxFTA.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/@jezar?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Jezael Melgoza</a> on <a href="https://unsplash.com/photos/people-riding-train-IdnphxeOpuQ?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>A recent visit to Tokyo turned out to be more than a cultural experience; it sent my thinking somewhere unexpected. On a crowded metro train, my friend and I noticed a man in heavy glasses watching us — likely curious about the two tall foreigners in front of him. That moment sparked a conversation: what if every wearable on the train were a virtual assistant? How would we feel? How would we know, and could we choose not to interact with them? Is there a universal sign for “Please don’t model my behavior today”?</p><p>Welcome to the Ambient AI era.</p><h3>Two Futures</h3><p>Fast-forward a few years, and I can see two distinct futures unfolding.</p><h3>Future 1: The Protocol of Presence</h3><p>Ambient AI becomes an invisible public utility within a decade.</p><p>In this version, ambient AI — the kind of AI embedded in our wearables and environments — becomes as seamless and reliable as the internet itself. Because data never leaves your device, it’s faster, more efficient, and inherently more private. This is powered by a local-first architecture, where computation happens on-device rather than in the cloud. Specialized neural-processing units (NPUs) run domain-specific models inside a hardware-isolated secure zone (Trusted Execution Environments, TEE).</p><p>New interaction layers emerge: instead of silently capturing data, reciprocal protocols request context through shared gestures — like a handshake. Human-side infrastructure (HSI) gives people tools to shape how machines perceive them. And digital “do-not-disturb” and “do-not-sense” preferences are enforced technically, and by design.</p><h3>Future 2: The Interaction Tax</h3><p>Ambient AI triggers backlash, delaying mainstream adoption for a generation.</p><p>In this future, cloud-native architectures dominate because centralized world models outperform anything running locally. Or industry optimizes for maximal data capture, and human presence turns into continuous telemetry — processed, monetized, and traded in ways people never fully understand. Or the ecosystem fractures into incompatible platforms, repeating the smart-home mess. Or the consumer platform never forms because human-centered design is treated as optional — and by the time anyone notices, the problems are already too complex to untangle.</p><p>In response, people start covering sensors, leaving wearables behind, avoiding ambient spaces altogether. The social cost of being modeled is too high.</p><p>Human-side infrastructure still emerges — but defensively: as resistance, as self-protection. People don’t want to shape how they’re sensed; they want to disappear from sensing entirely.</p><h3>What I’m Watching For</h3><p>Which future unfolds depends on developments across hardware, software, and the human layer:</p><p><strong>Hardware first:</strong> I’d lean toward Future 1 once NPUs and TEEs show up in mid-tier devices, not just flagships. When the hardware is everywhere, quiet software follows.</p><p><strong>Then efficiency:</strong> Can on-device models really compete with the cloud without killing batteries or overheating devices? If not, ambient AI never becomes a true utility.</p><p><strong>Protocols matter too:</strong> I’m looking for something like a universal handshake for AI intent — APIs that make “do not disturb” and “do not record” actually work across devices and platforms. If that becomes easy to implement, we’re looking at a real ecosystem.</p><p><strong>Regulation and scandal:</strong> Do governments and major markets align or fragment? Could a major cloud-native scandal accelerate the shift to local-first? These forces could tip the balance either way.</p><p><strong>Finally, humans:</strong> Are people gaining real ways to shape how machines perceive them? If the tools are used to preserve expressiveness and ambiguity, that’s Future 1. If they’re used mainly to hide, that’s Future 2.</p><h3>The Gap in the Middle</h3><p>Back from the future to the present. Imagine walking into an Amazon Go store that pushes edge AI to its limits. Local vision sensors act as “smart eyes,” reading your movements without sending video to the cloud. Lattice FPGAs — the “ultra-fast secretaries” — organize data from hundreds of sensors so the system never loses track of you. An NVIDIA Jetson — the “brain” — runs real-time inference to determine when you’ve actually picked up an item. Together, the system turns your physical presence into continuous digital input so the store can charge you automatically.</p><p>To push back, you might wear an adversarial-ML “privacy hoodie,” whose strange pattern works like a digital cloak — confusing the system so you register as background noise rather than a person. The store can’t track you — but it also can’t sell you anything.</p><p>Right now, you’re either fully visible to ambient systems or fully blocked. There’s no middle ground — no edit, no pause, no way to say “sense this, but not that.” That gap is what obsesses me: <strong>human-side infrastructure (HSI)</strong> — tools that let people shape how they’re perceived by machines, not just whether they’re perceived at all.</p><h3>What’s Emerging</h3><p>Retail-ready examples are already appearing, though most remain static: there’s privacy-focused eyewear like Reflectacles, and clothing from independent designers such as Cap_able.</p><p>Researchers are also pushing the space forward. Pedro Lopes’ group at University of Chicago is developing wearable ultrasonic jammers that let you selectively disable microphones nearby — not all sensing, just audio, just when you choose. Katia Vega is experimenting with “beauty technology” at UC Davis, embedding computation into cosmetics and wearables, treating the body as an interactive surface rather than a passive target.</p><p>And of course, there’s the canary in the mine: art and fashion. Zach Blas’s Facial Weaponization Suite, Adam Harvey’s CV Dazzle, and Isao Hashimoto’s garments that distort silhouette and gait recognition all reframe “modulating digital presence” as both a technical and cultural practice.</p><p>What excites me is that the hardware is finally catching up. Ultra-low-power AI microcontrollers (Ambiq’s Apollo series, Nordic’s nRF54) can run inference at the edge of the body. Flexible e-ink and micro-LED surfaces make dynamic, adaptive “skins” possible — patterns that respond to what’s sensing you rather than static prints.</p><p>The bottleneck now isn’t hardware or material; it’s integration. Someone needs to assemble these pieces into products people actually want to wear.</p><h3>The Question I Want to Answer</h3><p>Which of these approaches is closest to breakout?</p><p>My hypothesis is that static adversarial fashion (Reflectacles, Cap_able) is a dead end — too narrow, too binary, too unfashionable for mainstream adoption. The opportunity is in active, adaptive systems that feel less like armor and more like a privacy setting you forget is running.</p><p>The infrastructure that lets us curate our digital presence online is mature. The equivalent for physical space barely exists. I want to map what’s emerging, identify what’s viable, and figure out whether we’re building these tools out of desire or fear. That will shape everything — the product design, the go-to-market, the adoption curve.</p><h3>Why This Matters</h3><p>In ambient AI, people and invisible software form feedback loops in which signals are sensed, interactions modeled, and inferences drawn — often continuously. As AI moves closer to everyday life, a product’s success will depend not only on performance, but on whether people feel comfortable, empowered, and even delighted living alongside it.</p><p>That raises questions beyond technical capability: how uncertainty is handled, how consent is expressed or bypassed, how ambiguity is treated, and how design choices shift power. These aren’t just philosophical concerns — they influence trust, adoption, and willingness to pay.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9fbb3c1de825" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>