<?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 Ambica on Medium]]></title>
        <description><![CDATA[Stories by Ambica on Medium]]></description>
        <link>https://medium.com/@ambicaghai?source=rss-eccc9a102e7d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*qrRsq_m5ZsciUfWiuE_RxQ.jpeg</url>
            <title>Stories by Ambica on Medium</title>
            <link>https://medium.com/@ambicaghai?source=rss-eccc9a102e7d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 18:45:08 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ambicaghai/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 Quiet Grace of Persistence]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@ambicaghai/the-quiet-grace-of-persistence-57c047cb3a60?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*AV7DX2m5BQZ0gKS6" width="3600"></a></p><p class="medium-feed-snippet">Lessons from the PhD Journey</p><p class="medium-feed-link"><a href="https://medium.com/@ambicaghai/the-quiet-grace-of-persistence-57c047cb3a60?source=rss-eccc9a102e7d------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@ambicaghai/the-quiet-grace-of-persistence-57c047cb3a60?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/57c047cb3a60</guid>
            <category><![CDATA[persistence]]></category>
            <category><![CDATA[research]]></category>
            <category><![CDATA[uncertainty]]></category>
            <category><![CDATA[phd]]></category>
            <category><![CDATA[personal-development]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Tue, 27 May 2025 18:02:35 GMT</pubDate>
            <atom:updated>2025-05-27T18:02:35.133Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[What I’m Learning About Resilience]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/write-a-catalyst/what-im-learning-about-resilience-fe8cc77c064a?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*L6mZG-Z8yIRLmQjb" width="5184"></a></p><p class="medium-feed-snippet">It&#x2019;s not only about bouncing back.</p><p class="medium-feed-link"><a href="https://medium.com/write-a-catalyst/what-im-learning-about-resilience-fe8cc77c064a?source=rss-eccc9a102e7d------2">Continue reading on Write A Catalyst »</a></p></div>]]></description>
            <link>https://medium.com/write-a-catalyst/what-im-learning-about-resilience-fe8cc77c064a?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/fe8cc77c064a</guid>
            <category><![CDATA[life]]></category>
            <category><![CDATA[slowliving]]></category>
            <category><![CDATA[lessons-learned]]></category>
            <category><![CDATA[personal-growth]]></category>
            <category><![CDATA[resilience]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Tue, 13 May 2025 06:27:52 GMT</pubDate>
            <atom:updated>2025-05-13T06:27:52.209Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Unlocking Meaning at Scale: Why Sentence-BERT Matters]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/it-chronicles/unlocking-meaning-at-scale-why-sentence-bert-matters-f2bb039ebcac?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/1455/1*lVelBWQXsw7U57uC9pU0OA.png" width="1455"></a></p><p class="medium-feed-snippet">Making BERT Practical for Semantic Search and Clustering</p><p class="medium-feed-link"><a href="https://medium.com/it-chronicles/unlocking-meaning-at-scale-why-sentence-bert-matters-f2bb039ebcac?source=rss-eccc9a102e7d------2">Continue reading on IT Chronicles »</a></p></div>]]></description>
            <link>https://medium.com/it-chronicles/unlocking-meaning-at-scale-why-sentence-bert-matters-f2bb039ebcac?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/f2bb039ebcac</guid>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[mlops]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Thu, 08 May 2025 11:09:16 GMT</pubDate>
            <atom:updated>2025-05-08T11:09:16.896Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Returning to Wisdom: What The Root Cause Reminds Us About Health, Mind, and Higher Knowledge]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/illumination/returning-to-wisdom-what-the-root-cause-reminds-us-about-health-mind-and-higher-knowledge-b462a5db9ead?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*C05LmfkHkiK8bq5V" width="5868"></a></p><p class="medium-feed-snippet">Reflections Inspired by Mr. Alok Jain&#x2019;s Work</p><p class="medium-feed-link"><a href="https://medium.com/illumination/returning-to-wisdom-what-the-root-cause-reminds-us-about-health-mind-and-higher-knowledge-b462a5db9ead?source=rss-eccc9a102e7d------2">Continue reading on ILLUMINATION »</a></p></div>]]></description>
            <link>https://medium.com/illumination/returning-to-wisdom-what-the-root-cause-reminds-us-about-health-mind-and-higher-knowledge-b462a5db9ead?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/b462a5db9ead</guid>
            <category><![CDATA[synergy]]></category>
            <category><![CDATA[spirituality]]></category>
            <category><![CDATA[body]]></category>
            <category><![CDATA[mind]]></category>
            <category><![CDATA[astrology]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Thu, 01 May 2025 07:06:17 GMT</pubDate>
            <atom:updated>2025-05-01T07:06:17.079Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Art of Taking It Easy]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/write-a-catalyst/the-art-of-taking-it-easy-a511dfd11a91?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*o8TYfaE5a8ipTkum" width="3024"></a></p><p class="medium-feed-snippet">The Art of Slow Living in a Fast World</p><p class="medium-feed-link"><a href="https://medium.com/write-a-catalyst/the-art-of-taking-it-easy-a511dfd11a91?source=rss-eccc9a102e7d------2">Continue reading on Write A Catalyst »</a></p></div>]]></description>
            <link>https://medium.com/write-a-catalyst/the-art-of-taking-it-easy-a511dfd11a91?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/a511dfd11a91</guid>
            <category><![CDATA[fast]]></category>
            <category><![CDATA[slowliving]]></category>
            <category><![CDATA[mental-health]]></category>
            <category><![CDATA[life]]></category>
            <category><![CDATA[mindfulness]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Thu, 01 May 2025 04:10:22 GMT</pubDate>
            <atom:updated>2025-05-01T05:38:20.136Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Mental Health in a Machine World: Why Stillness Is No Longer Optional]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/health-science/mental-health-in-a-machine-world-why-stillness-is-no-longer-optional-76ae17c520d3?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*2QP3RCsAc8h6ZszJ" width="4000"></a></p><p class="medium-feed-snippet">Staying Human in the Age of Artificial Minds</p><p class="medium-feed-link"><a href="https://medium.com/health-science/mental-health-in-a-machine-world-why-stillness-is-no-longer-optional-76ae17c520d3?source=rss-eccc9a102e7d------2">Continue reading on Health and  Science »</a></p></div>]]></description>
            <link>https://medium.com/health-science/mental-health-in-a-machine-world-why-stillness-is-no-longer-optional-76ae17c520d3?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/76ae17c520d3</guid>
            <category><![CDATA[mental-health]]></category>
            <category><![CDATA[spirituality]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[mindfulness]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Mon, 28 Apr 2025 06:34:40 GMT</pubDate>
            <atom:updated>2025-04-28T06:34:40.410Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[You’re Being Hard on Yourself, Ambica]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/write-a-catalyst/youre-being-hard-on-yourself-ambica-ff63df8ec046?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/2448/0*Tv2gMdr1qzFxgL7F" width="2448"></a></p><p class="medium-feed-snippet">Healing isn&#x2019;t a performance</p><p class="medium-feed-link"><a href="https://medium.com/write-a-catalyst/youre-being-hard-on-yourself-ambica-ff63df8ec046?source=rss-eccc9a102e7d------2">Continue reading on Write A Catalyst »</a></p></div>]]></description>
            <link>https://medium.com/write-a-catalyst/youre-being-hard-on-yourself-ambica-ff63df8ec046?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/ff63df8ec046</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[therapy]]></category>
            <category><![CDATA[spirituality]]></category>
            <category><![CDATA[healing]]></category>
            <category><![CDATA[mental-health]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Thu, 24 Apr 2025 01:20:06 GMT</pubDate>
            <atom:updated>2025-04-24T01:20:06.378Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Video-RAG: A Practical Breakthrough in Long Video Understanding for Multimodal AI]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/it-chronicles/video-rag-a-practical-breakthrough-in-long-video-understanding-for-multimodal-ai-5803a7409105?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*wNTR9aRB4j4L9h_m" width="2848"></a></p><p class="medium-feed-snippet">A plug-and-play method, without retraining or breaking your inference budget.</p><p class="medium-feed-link"><a href="https://medium.com/it-chronicles/video-rag-a-practical-breakthrough-in-long-video-understanding-for-multimodal-ai-5803a7409105?source=rss-eccc9a102e7d------2">Continue reading on IT Chronicles »</a></p></div>]]></description>
            <link>https://medium.com/it-chronicles/video-rag-a-practical-breakthrough-in-long-video-understanding-for-multimodal-ai-5803a7409105?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/5803a7409105</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[lvm]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Wed, 23 Apr 2025 20:06:56 GMT</pubDate>
            <atom:updated>2025-04-23T21:21:32.093Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Mindfulness Is Not About Escaping the Mind — It’s About Befriending It]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/illumination/mindfulness-is-not-about-escaping-the-mind-its-about-befriending-it-cbe384a79d74?source=rss-eccc9a102e7d------2"><img src="https://cdn-images-1.medium.com/max/2600/0*kW9p43PGbIJveAdA" width="5956"></a></p><p class="medium-feed-snippet">What Kudesia&#x2019;s &#x201C;Metacognitive Practice&#x201D; Can Teach Us About Spiritual Presence in an AI World</p><p class="medium-feed-link"><a href="https://medium.com/illumination/mindfulness-is-not-about-escaping-the-mind-its-about-befriending-it-cbe384a79d74?source=rss-eccc9a102e7d------2">Continue reading on ILLUMINATION »</a></p></div>]]></description>
            <link>https://medium.com/illumination/mindfulness-is-not-about-escaping-the-mind-its-about-befriending-it-cbe384a79d74?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/cbe384a79d74</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[reflections]]></category>
            <category><![CDATA[mindfulness]]></category>
            <category><![CDATA[growth]]></category>
            <category><![CDATA[spirituality]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Tue, 15 Apr 2025 19:20:57 GMT</pubDate>
            <atom:updated>2025-04-15T19:20:57.047Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[From human-in-the-loop to human-on-the-loop]]></title>
            <link>https://medium.com/@ambicaghai/from-human-in-the-loop-to-human-on-the-loop-688534c53cc4?source=rss-eccc9a102e7d------2</link>
            <guid isPermaLink="false">https://medium.com/p/688534c53cc4</guid>
            <category><![CDATA[machine-learnin]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[human-in-the-loop]]></category>
            <category><![CDATA[mcps]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Ambica]]></dc:creator>
            <pubDate>Tue, 15 Apr 2025 17:02:47 GMT</pubDate>
            <atom:updated>2025-04-15T17:02:47.950Z</atom:updated>
            <content:encoded><![CDATA[<h4>Rethinking Oversight and Autonomy in an Age of AI Agents</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KxT4fhc9VkpBsOyk" /><figcaption>Photo by <a href="https://unsplash.com/@kommumikation?utm_source=medium&amp;utm_medium=referral">Mika Baumeister</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>As AI agents assume ever increasing roles across fields — from healthcare to finance, logistics to legal — our relationship with automation is evolving. We’re not just scaling up the capabilities of large language models; we’re revolutionizing the way humans remain in the loop.</p><p><em>The transition from human-in-the-loop (HITL) to human-on-the-loop (HOTL) is more than semantic:</em> It represents a fundamental change in the way decisions are made, who makes them, and when — or if — human judgment should step in.</p><p>Central to this evolution is the emergence of machine-centric processes (MCPs) and ever-autonomous AI agents. These systems are promising speed, consistency and scale. But they also elevate hard questions about trust, oversight and the boundaries of delegation.</p><p>This article investigates the transition from HITL to HOTL, the types of AI agents in our current workflows, the increasing power of MCPs, and — most importantly — the trade-offs that accompany each configuration.</p><h3>Human-in-the-Loop: Control and Assurance Comes at a Price</h3><p>A Human-in-the-Loop system is where humans continue to be directly involved in the decision-making loop. AI can produce predictions, rankings or classifications — but nothing is taking any action without human intervention. Such setups are common in high-stakes environments: medical diagnoses, military targeting, financial fraud detection, and legal judgments.</p><h4>Why it works:</h4><p>HITL systems add a human lens, something that the algorithm-driven systems on their own lack. They empower domain experts to introduce context and values and discretion into the fold — things that data alone can’t provide.</p><h4>But there’s a cost:</h4><ol><li>These systems do not easily scale for real-time or high-frequency decisions.</li><li>They demand constant human supervision, which can become both resource-intensive and mentally exhausting.</li><li>They gunk up workflows that otherwise stand ready for automation.</li></ol><p>In summary, HITL is safe, but not necessarily practical.</p><h3>Human-on-the-Loop: Control Without Micromanagement</h3><p>Human-on-the-Loop system is a more scalable solution. Here, AI systems decide after human following their behavior, audit outputs periodically and also have a framework to intervene if things go wrong. It’s not endorsing every decision — it’s overseeing the larger system.</p><p>It works well in environments for which speed is critical, but complete autonomy places them at risk. Like walking a cybersecurity dashboard where the tech highlights unusual behavior but a human makes the call when to escalate. Or a trading bot that runs freely until it trips over volatility levels defined by a risk officer.</p><h4>What it enables:</h4><ol><li>Quicker processes with less downtime.</li><li>This is known as a more general dynamic system that can adapt to multipath environments and yet still be observed.</li><li>The ideal compromise between human expertise and machine autonomy.</li></ol><h4>But it does face some challenges:</h4><ol><li>There’s a danger of complacency around automation — when humans overtrust the system, and don’t intervene when they should.</li><li>Within HOTL, oversight is generally reactive. It can take time before humans see a problem; the damage may have already begun.</li><li>Moreover, as the complexity of AI systems increases, human operators may not completely grasp what’s happening behind the curtains — creating discrepancies in intervention quality.</li></ol><h3>The Rise of AI Agents and Machine-Centric Processes (MCPs)</h3><p>AI agents go a step further: more autonomy. They are not just models which take input and produce output — they are systems that act, learn, and adapt over time. These agents typically work across many tasks, without the constant need for supervision — be it an end-to-end customer service agent that solves problems or a financial copilot that generates reports.</p><p>This has resulted in machine-centric processes (MCPs) — the workflow built out around what a machine can do, not what a human must supervise. AI agents are the primary actors in MCPs. Humans are called in only for exceptions, to make new designs or to audit processes.</p><h4>Why organizations are moving toward MCPs:</h4><ol><li>They are a huge step in reducing manual workload.</li><li>They increase the speed and economy of operations.</li><li>They allow for new classes of services that just weren’t possible before.</li></ol><p>Yet MCPs are not a plug-and-play solution. They call for another way to think about responsibility, transparency and resilience.</p><h4>But There Are Trade-offs</h4><p>MCPs and AI agents have an attractive cost-benefit ratio, yet they also entail substantial trade-offs, particularly about the downstream visibility, trust, and risk-management.</p><p>Accountability is straightforward in Human-in-the-Loop systems. A human made the final call. In Human-on-the-Loop models, oversight is shared (and so some clarity is diluted), but there is still a traceable chain that leads back to some human being. In MCPs, though, it becomes more difficult to pin down who or what is responsible for a decision in the end — especially in edge cases.</p><p>Interpretability is lost as well. As AI agents get increasingly complex and interconnected, it’s more difficult for other humans to know how a specific result was achieved. And if people cannot understand or explain what the system is doing, the system becomes more difficult to trust, regulate or improve.</p><p>There is also the difficulty of trust calibration. If a system gets most things right, but is sometimes wrong in subtle ways, humans may over-trust it or under-trust it, and are equally capable of causing failures as a result. And in fully machine-centric workflows, by the time human attention is summoned, the machine may already have strayed too far off course.</p><p>These are not deal-breakers, but they are design constraints. If we want the end benefits of autonomous agents, we need to actively account for their limitations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kTAuBEET32g2xUOOZQG0lg.png" /><figcaption>Comparative strengths and weaknesses of HITL, HOTL, and MCP across five dimensions</figcaption></figure><h3>Redesigning Human Roles: From Decision-Makers to System Designers</h3><p>Human roles aren’t going away as we progress from HITL to HOTL — and further into MCP territory. They’re evolving.</p><p>Instead of making every decision, humans are now:</p><ol><li>Establishing guard rails and escalation path.</li><li>Training of operational envelopes for the agents.</li><li>Auditing results and improving model behavior.</li><li>Reconciliation of conflicting objectives or values.</li></ol><p>It is a move from doing the work to structuring the system that does the work. It demands new skills: systems thinking, ethical foresight, and a deep understanding of AI behaviors — not just domain knowledge.</p><h3>Conclusion: The Question Should Not Be One or the Other</h3><p>The fact is, there’s no one loop that suits all use cases. HITL remains important in high-stakes or tightly controlled domains. HOTL is a good choice when we need a middle ground between both speed and control. And MCPs they’re a great fit for repeatable, low-risk processes that have started to benefit from scale and consistency.</p><p>The key is intentionality. Don’t just replace humans — rethink where humans <em>add value</em>. Design systems where machine autonomy is supported by human awareness.</p><p>Build feedback loops, not just control panels.</p><p>And remember: as AI systems evolve, our oversight strategies must evolve with them — not after.</p><h3>Let’s Keep the Conversation Human</h3><p>As we continue building smarter AI systems and more autonomous agents, it’s essential that we pause and ask: <em>Where should humans still be part of the loop — and how?</em> This isn’t just a technical design question; it’s a values decision. The way we architect oversight today will shape how AI interacts with people, policies, and power tomorrow.</p><p>I’d love to hear from practitioners, product thinkers, and policy designers:<br> How are you approaching the balance between automation and human judgment in your own work? What oversight frameworks are working — and where are they falling short?</p><p>Drop your thoughts in the comments. Let’s build more responsible, transparent AI — together.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=688534c53cc4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>