<?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 Victoria Jhong on Medium]]></title>
        <description><![CDATA[Stories by Victoria Jhong on Medium]]></description>
        <link>https://medium.com/@victoriajhong_78089?source=rss-b27e973a5c83------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*NL8GpmT218hkhGAB.</url>
            <title>Stories by Victoria Jhong on Medium</title>
            <link>https://medium.com/@victoriajhong_78089?source=rss-b27e973a5c83------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 08:16:20 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@victoriajhong_78089/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[Dear Diary-2, What Artificial Intelligence Actually Is (and Isn’t)]]></title>
            <link>https://medium.com/@victoriajhong_78089/dear-diary-2-what-artificial-intelligence-actually-is-and-isnt-bedaf4c191b2?source=rss-b27e973a5c83------2</link>
            <guid isPermaLink="false">https://medium.com/p/bedaf4c191b2</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Victoria Jhong]]></dc:creator>
            <pubDate>Fri, 26 Dec 2025 14:26:08 GMT</pubDate>
            <atom:updated>2025-12-26T14:26:08.379Z</atom:updated>
            <content:encoded><![CDATA[<p>When people talk about Artificial Intelligence, they often jump straight to models.</p><p><em>Neural networks.<br>Deep learning.<br>Chatbots.<br>Automation.</em></p><p>But before learning <em>how</em> AI works, I realized I needed to answer a more basic question:</p><p><strong>What problem is Artificial Intelligence actually trying to solve?</strong></p><p>Without that, everything else feels fragmented.</p><h3>Artificial Intelligence is not one thing</h3><p>Artificial Intelligence is not a single technique or model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/1*Et-UnrhajG7Z24k11I1w8g.png" /></figure><p>It is a <strong>family of ideas</strong> about how machines can perform tasks that normally require human intelligence, such as:</p><ul><li>reasoning</li><li>learning from experience</li><li>making decisions under uncertainty</li><li>adapting to changing environments</li></ul><p>Different approaches emerged because <strong>people disagreed on what “intelligence” really means</strong>.</p><p>Understanding AI starts with understanding those disagreements.</p><h3>Four ways people have tried to build intelligence</h3><p>Over time, AI research has followed four major paths.<br> Most modern systems are combinations of these, but the ideas are distinct.</p><h3>1. Intelligence as rules and logic</h3><p>This is the oldest view of AI.</p><p>The idea is simple:</p><blockquote><em>If human experts can explain their reasoning as rules, machines can follow those rules.</em></blockquote><p>This approach represents knowledge explicitly:</p><ul><li><em>symbols</em></li><li><em>rules</em></li><li><em>logic</em></li><li><em>if–then statements</em></li></ul><p><strong>It works well when:</strong></p><ul><li>rules are stable</li><li>the domain is well-defined</li><li>explanations are critical</li></ul><p><strong>Its limitation is obvious:</strong></p><ul><li>the real world is messy</li><li>rules are incomplete</li><li>uncertainty is everywhere</li></ul><p>Still, this way of thinking never disappeared.<br>It lives on in expert systems, business rules, and governance logic.</p><h3>2. Intelligence as probability and statistics</h3><p>Another view treats intelligence as <strong>reasoning under uncertainty</strong>.</p><p>Instead of asking:</p><blockquote><em>What rule applies?</em></blockquote><p>It asks:</p><blockquote><em>What is the most likely explanation, given incomplete information?</em></blockquote><p>This approach relies on:</p><ul><li>probability</li><li>statistics</li><li>time-series analysis</li><li>inference</li></ul><p>It is powerful when:</p><ul><li>data is noisy</li><li>outcomes are uncertain</li><li>trends and patterns matter</li></ul><p>Many “classic” models used in forecasting and decision support come from this tradition.</p><p>Importantly, this view does not try to imitate human thinking.<br> It tries to make <strong>optimal decisions under uncertainty</strong>.</p><h3>3. Intelligence as learning from data</h3><p>This is where modern machine learning and deep learning sit.</p><p>The core idea is:</p><blockquote><em>Instead of telling the system how to behave, let it learn patterns from data.</em></blockquote><p>Neural networks, including deep learning models, belong here.</p><p>They are especially strong at:</p><ul><li>complex patterns</li><li>non-linear relationships</li><li>high-dimensional data</li></ul><p>But they come with trade-offs:</p><ul><li>heavy data requirements</li><li>limited transparency</li><li>sensitivity to how problems are framed</li></ul><p>Learning from data is powerful, but it does not automatically mean understanding.</p><h3>4. Intelligence as interaction and control</h3><p>Another tradition views intelligence as something that emerges from interaction with an environment.</p><p>This includes:</p><ul><li>control systems</li><li>feedback loops</li><li>adaptation over time</li></ul><p>Here, intelligence is not about prediction alone, but about:</p><ul><li>sensing</li><li>acting</li><li>adjusting based on outcomes</li></ul><p>This way of thinking is central in robotics, operations, and system optimization.</p><h3>Modern AI is a mixture, not a winner</h3><p>One mistake beginners often make is assuming that newer approaches replaced older ones.</p><p>In reality:</p><ul><li>rules are still everywhere</li><li>statistics still matter</li><li>learning models are layered on top</li><li>control logic constrains everything</li></ul><p>Modern AI systems work because <strong>different ideas compensate for each other’s weaknesses</strong>.</p><p>There is no single “best” model — only models that fit certain contexts.</p><h3>Why this matters before learning models</h3><p>If you start learning AI from tools alone, everything feels scattered:</p><ul><li>too many algorithms</li><li>conflicting advice</li><li>unclear trade-offs</li></ul><p>Once you understand the <strong>theoretical landscape</strong>, new concepts have somewhere to land.</p><p>Instead of asking:</p><blockquote><em>Should I learn this model?</em></blockquote><p>You start asking:</p><blockquote><em>What assumption does this model make about intelligence?</em></blockquote><p>That question changes everything.</p><p>Dear Diary,</p><p>After today, you should clearly know:</p><ul><li>why AI exists at all</li><li>what problem it is fundamentally solving</li><li>why multiple AI approaches coexist</li><li>why models are tools, not intelligence</li></ul><p>Let’s call it a day.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bedaf4c191b2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Dear Diary-1, I’m Learning Data Science, but Why Does My Knowledge Feel So Scattered?]]></title>
            <link>https://medium.com/@victoriajhong_78089/dear-diary-1-im-learning-data-science-but-why-does-my-knowledge-feel-so-scattered-0ea2689875d2?source=rss-b27e973a5c83------2</link>
            <guid isPermaLink="false">https://medium.com/p/0ea2689875d2</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[deep-learning]]></category>
            <dc:creator><![CDATA[Victoria Jhong]]></dc:creator>
            <pubDate>Fri, 26 Dec 2025 13:54:28 GMT</pubDate>
            <atom:updated>2025-12-26T13:54:28.999Z</atom:updated>
            <content:encoded><![CDATA[<p>When I first started learning Data Science, I did what most people do.</p><p>I learned models.</p><p>GRU/LSTM/RNN</p><p>I could run the code. I could produce results.<br>And yet, at some point, I realized something uncomfortable:</p><p><strong>I knew how to run models, but I couldn’t clearly explain what problem I was actually solving.</strong></p><p>That wasn’t a technical issue.<br>It was a structural one.</p><h3>Feeling “scattered” doesn’t mean you’re bad at Data Science</h3><p>For a long time, I thought my problem was that I hadn’t learned <em>enough</em>.</p><p>More models.<br>More calculation.<br>More math.</p><p>But eventually I realized something counterintuitive:</p><blockquote><em>People who feel their knowledge is scattered are often the ones who have already crossed into real-world problems too early.</em></blockquote><p>Data Science doesn’t come from a single discipline.<br> It sits at the intersection of:</p><p><em>statistics</em></p><p><em>programming</em></p><p><em>systems and decision-making</em></p><p>If you stay inside just one of these, things feel clean and confident.<br>The moment you try to connect them, everything starts to feel messy.</p><p>That mess is not failure.<br>It’s the beginning of systems thinking.</p><h3>The moment I stopped blaming the models</h3><p>While working on short-term cooling load prediction for data centers, I tested different approaches:</p><ul><li>traditional time-series models (ARIMA)</li><li>deep learning sequence models (GRU)</li><li>state-driven regression models (tree-based methods)</li></ul><p>The result surprised me.</p><p>The performance differences were <strong>not dramatic</strong>.</p><p>That forced me to ask a deeper question:</p><blockquote><em>If different models can explain the data reasonably well, what actually matters?</em></blockquote><p>The answer wasn’t “use a more complex model”.</p><p>It was <strong>how the problem was framed</strong>, and <strong>how reality was translated into features</strong>.</p><h3>Data Science finally made sense when I stopped centering it on models</h3><p>To stop feeling lost, I started organizing everything I learned into a simple structure.</p><p>Not tools.<br>Not algorithms.<br>A structure.</p><h3>The five layers I now use to think about Data Science:</h3><p><strong>Problem layer</strong><br>Why are we predicting this? Who makes decisions with the result?</p><p><strong>Data layer</strong><br>What data do we actually have? At what frequency? With what limitations?</p><p><strong>Feature layer</strong><br> How do we represent reality? States, thresholds, aggregations, delays.</p><p><strong>Model layer</strong><br>Which methods are appropriate — and why?</p><p><strong>Evaluation layer</strong><br>Is this usable? Stable? Interpretable? Deployable?</p><p>When I placed ARIMA, GRU, and LightGBM into this framework, something clicked:</p><p><strong>They weren’t competing with each other.<br>They were answering different questions.</strong></p><h3>Accepting that “scattered” is a transitional state</h3><p>Real Data Science is not about knowing the most models.</p><p>It’s about being able to answer this:</p><blockquote><em>In this context, what complexity is actually justified?</em></blockquote><p>If you’re learning Data Science and feel that:</p><ul><li>your knowledge is fragmented</li><li>models blur together</li><li>you’re unsure whether you “really understand”</li></ul><p>That might not mean you’re behind.</p><p>It might mean you’ve reached the point where technique alone is no longer enough.</p><h3>What I plan to write here</h3><p>I want to write about:</p><ul><li>how technical pieces connect into systems</li><li>why simpler models often survive real-world constraints</li><li>how Data Science supports decisions, not just predictions</li></ul><p>This space is for documenting the process of my growth.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0ea2689875d2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>