<?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[Teach me in Plain Language - Medium]]></title>
        <description><![CDATA[Technology Made Simple. Understanding Made Fun. - Medium]]></description>
        <link>https://medium.com/teach-me-in-plain-language?source=rss----c1ccbddf4353---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Teach me in Plain Language - Medium</title>
            <link>https://medium.com/teach-me-in-plain-language?source=rss----c1ccbddf4353---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 23:16:51 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/teach-me-in-plain-language" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Kidnapped Robot Problem]]></title>
            <link>https://medium.com/teach-me-in-plain-language/kidnapped-robot-problem-a0fa10c7bd69?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/a0fa10c7bd69</guid>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[localization]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Fri, 07 Mar 2025 14:45:37 GMT</pubDate>
            <atom:updated>2025-03-07T14:45:37.315Z</atom:updated>
            <content:encoded><![CDATA[<h4>How to recover when you find yourself in a new unknown location</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*HQQk2NZ6PYUE8rBn" /><figcaption>Photo by <a href="https://unsplash.com/@arttravelling?utm_source=medium&amp;utm_medium=referral">oxana v</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>We all have a version of a childhood memory of falling asleep on the couch, a car or a living room while playing and waking up in your bedroom. It’s a classic experience.</p><p>You wake up and it takes you a few minutes to orient yourself. The familiarity of your room brings you back to relief and comfort.</p><blockquote>Well, dear reader, you have experienced the classic Kidnapped Robot Problem in real life.</blockquote><h3>The Kidnapped Robot Problem</h3><p>This is a common challenge in robotics and AI, where a robot (simulated or physical) is suddenly moved to an unknown location without detecting the transition.</p><blockquote>The robot has been turned off, sensors are disabled and a user moves the robot to a new location.</blockquote><p>The robot effectively wakes up in a new, unknown — or known — location and has to figure itself out. Due to needing to localize itself, the robot relies on a continuous stream of sensor data and being moved breaks that assumption.</p><blockquote>If you have a robot vacuum, this happens frequently when a user wants to clean a different part of the room or a different floor.</blockquote><h4>Here are some solutions to mitigate the issue:</h4><ol><li><strong>Global Localization </strong>(Kidnapped Robot Recovery Mode)</li></ol><p>• Use Monte Carlo Localization (MCL) with particle filters that periodically introduce random particles across the entire map. If the robot is kidnapped, these random particles help it guess its new location.</p><p>• Increase sensor noise models and allow for higher uncertainty in localization estimates, helping the system detect inconsistencies when a kidnapping occurs.</p><p>2.<strong> Sensor Fusion &amp; Redundant Data Sources</strong></p><p>• GPS Integration: GPS can provide a rough global estimate of position, allowing the robot to recover from a sudden displacement.</p><p>• Vision-Based Localization: Using SLAM (Simultaneous Localization and Mapping) with cameras (Visual SLAM) allows the robot to recognize landmarks and reorient itself.</p><p>• Inertial Measurement Units (IMUs): These can detect sudden accelerations inconsistent with normal movement, triggering a re-localization process.</p><p>3. <strong>Anomaly Detection &amp; Re-Initialization</strong></p><p>• Implement outlier detection to recognize when sensor readings don’t match the expected position</p><blockquote>For example, robot expecting to be in a hallway but suddenly detecting an outdoor environment</blockquote><p>• When anomalies are detected, force a full map re-scan or reset the localization system.</p><p>4. <strong>Map Matching &amp; Re-Localization</strong></p><p>• Store a database of visual or lidar landmarks. When kidnapped, the robot can search its memory for a matching location and reorient itself.</p><blockquote>Many markers are active and can move around, so be cautious of the visual landmarks you decide to use</blockquote><p>• Use deep learning models trained on environmental features to estimate approximate locations.</p><p>5. <strong>Active Exploration &amp; Environment Querying</strong></p><p>• If lost, the robot can enter an exploration mode, moving in a controlled manner while analyzing surroundings for familiar objects or structures.</p><p>• Some robots can ask for help, either by communicating with humans or querying a remote server with map data.</p><blockquote>If all else fail, human intervention is a good enough method (via apps or what ever your sceneario may be)</blockquote><h4>Hybrid Approaches</h4><p>• Combining probabilistic methods, deep learning, and multi-sensor fusion increases robustness, making it easier for robots to recover from unexpected relocations.</p><p>Each solution has its trade offs; computational cost, accuracy, latency and feasibility being key factors in deciding your ideal route to orienting your robot.</p><p>But as always, cater for your use-case, and hopefully you will be able to find an effective and acceptable solution for yourself.</p><p>As always, happy coding :)</p><p><em>Thanks for reading! 📚If you found this article interesting subscribe: </em><a href="https://medium.com/@dialogueswithai/subscribe"><em>https://medium.com/@dialogueswithai/subscribe</em></a></p><p><em>Claps and comments mean the world to me, and your support is deeply appreciated!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a0fa10c7bd69" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/kidnapped-robot-problem-a0fa10c7bd69">Kidnapped Robot Problem</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Are We Really Living in a Simulation?]]></title>
            <link>https://medium.com/teach-me-in-plain-language/are-we-really-living-in-a-simulation-9e1f1d564d09?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/9e1f1d564d09</guid>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[trending]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Thu, 28 Nov 2024 17:55:01 GMT</pubDate>
            <atom:updated>2024-11-28T17:58:47.639Z</atom:updated>
            <content:encoded><![CDATA[<h4>What does Elon Musk know? What do all the tech tycoons know?!</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*olf72wLtVC4X1phA" /><figcaption>Photo by <a href="https://unsplash.com/@ryoji__iwata?utm_source=medium&amp;utm_medium=referral">Ryoji Iwata</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>We have all heard Elon Musk and his belief that we are in a simulation. In case you haven’t he publicly stated:</p><blockquote>… that there is a “one in a billion chance that this is base reality.”</blockquote><p>In other words, Musk believes that we are almost certainly living in a computer simulation.</p><p>He’s done the calculations, and we are probably, most likely, not in <em>base reality.</em></p><h3>Where is this coming from?</h3><p>Back in 2001, a paper was published titled <strong>Are You Living in a Computer Simulation</strong> by Nick Bostrom, author of <strong><em>Superintelligence</em></strong>, and many other philosophical texts.</p><blockquote>This paper blew up in Silicon Valley.</blockquote><p>Many who were living and working in the Bay Area are heavily networked together and like wild fire this paper spread in the Tech industry.</p><p>And so as <em>Burning Man</em> was quickly integrated into lifestyle of Silicon Valley engineers, <strong>Are You Living in a Computer Simulation</strong> became bible.</p><p>I want to mention that this idea is not new, it’s been contemplated for centuries. Plato is famously attributed to <strong>Allegory of the Cave</strong> thought experiment.</p><p>It discusses the perceived reality in contrast to a larger, complete reality.</p><blockquote>Imagines a group of people chained together inside an underground cave as prisoners. In this cave, you can see shadows of the happenings outside this microworld.</blockquote><blockquote>If this group of prisoners live an die within the cave, then to the prisoners the reality is the shadows on the wall.</blockquote><blockquote>Even if there exists a larger reality they only perceive and imagine the limits of their confinement.</blockquote><p>This example illustrates the issue of what we believe in, and the undetermined ground truth that we cannot perceive.</p><p>Just like our realities, if we were in a simulation.</p><h3>What does the paper say?</h3><p>The paper claims:</p><blockquote>If there were a substantial chance that our civilization will ever get to the posthuman stage and run many ancestor‐simulations, then how come you are not living in such a simulation?</blockquote><blockquote>fp: Fraction of all human‐level technological civilizations that survive to reach a posthuman stage</blockquote><blockquote>N : Average number of ancestor‐simulations run by a posthuman civilization</blockquote><blockquote>H : Average number of individuals that have lived in a civilization before it reaches a posthuman stage</blockquote><p>The actual fraction of all observers with human‐type experiences that live in simulations is then:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/366/1*SlssT9tftf_tUJYQ3dBOtw.png" /></figure><p>There is some more math, some simplifications. Some hardcore mathematical proof statements. If you are interested, I have linked the paper at the end of this article.</p><p>After all that, there exists two cases:</p><h4><strong>The first case:</strong></h4><p>is where all the minds in question are like your own in the sense that they are exactly qualitatively identical to yours: they have exactly the same information and the same experiences that you have.</p><h4>The second case:</h4><p>is where the minds are “like” each other only in the loose sense of being the sort of minds that are typical of human creatures, but they are qualitatively distinct from one another and each has a distinct set of experiences</p><p>In the first case, humanity will go extinct (before reaching a posthuman state) in the latter, we require a convergence point, which is the posthuman state.</p><blockquote>Case 1: certain doom and extinction of humanity</blockquote><blockquote>Case 2: a convergence of strong advanced civilizations in which no one would want to become a “better” posthuman, and will not want to run any ancestor-simulations</blockquote><p>The two scenarios are so unlikely that there must be a third option. Here, Bostrom introduces this <strong>theory number three</strong>:</p><h4>The third case:</h4><blockquote>we are almost certainly living in a computer simulation</blockquote><p>Ultimately, the conclusion of the paper is:</p><blockquote>Unless we are now living in a simulation, our descendants will almost certainly never run an ancestor‐simulation.</blockquote><p>In other words, we are in a simulation because our posthuman-selves are motivated to find out <em>something</em>.</p><p>A multi-verse of humanity, timelines and lives to figure out an ideal solution.</p><p><em>What are they trying to figure out?</em></p><p>That, my friend, exists outside our understanding.</p><h3>What is Base Reality?</h3><p>Base reality is the term Elon uses to describe the ground truth. The reality of which all simulations are built from.</p><p>The chances are that we are in the one timeline of on our way to a posthuman state is unlikely. And the chances of a powerful all-knowing posthuman race is also unlikely.</p><blockquote>Bostrom theorizes that chances of us living in a simulation is about 50–50.</blockquote><blockquote>Elon, obviously, believes strongly we are in a simulated world.</blockquote><p>Take it as you will, the paper does offer three theories on our perceived reality.</p><blockquote>Looks like the Tech tycoons have chosen their favorite case:</blockquote><blockquote><strong>Scenario number three.</strong></blockquote><p><em>Thanks for reading! If you found this article interesting subscribe: </em><a href="https://medium.com/@dialogueswithai/subscribe"><em>https://medium.com/@dialogueswithai/subscribe</em></a></p><h4>References:</h4><p><a href="https://simulation-argument.com/simulation.pdf">https://simulation-argument.com/simulation.pdf</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9e1f1d564d09" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/are-we-really-living-in-a-simulation-9e1f1d564d09">Are We Really Living in a Simulation?</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data-Driven Loophole]]></title>
            <link>https://medium.com/teach-me-in-plain-language/data-driven-loophole-6a14b22bc0b8?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/6a14b22bc0b8</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[psychology]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Wed, 20 Nov 2024 15:42:09 GMT</pubDate>
            <atom:updated>2024-11-26T14:29:39.775Z</atom:updated>
            <content:encoded><![CDATA[<h4>How to convince your AI of anything and everything</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zy10U5NHPTXxP-qBulPswQ@2x.jpeg" /><figcaption>Photo by <a href="https://www.rvrdesign.co/">Evie S.</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>We love making comparisons of AI to humans. We give it humanoid characteristics and human-like brain structure. We teach it through human methods by giving it rewards and penalties in reinforcement training.</p><p><em>We often lose sight of what AI is, and what it is not.</em></p><p>AI is a data-driven algorithm with the ability to mimic human language or behaviors. It is data dependent, and this makes it easy to cater it to a microworld.</p><p>But this data driven madness is what makes it so easy to exploit.</p><h4><em>Unlike us Humans.</em></h4><p>One of my coworkers is vegetarian. Not by his love of animals and all things living, but because of his religion Hinduism. However, he’s not really Hindu, like I am not really a Catholic. He was born into the religion, by his community he was raised a certain way, and in becoming an adult and moving abroad he has lost his practice.</p><p>But he still does not eat meat.</p><p>The family doesn’t enforce it either, and believed he would eat meat once he moved to Canada. He has confided in me that he lied to them – that he lied to them about trying meat, because he <em>didn’t want to convince</em> them he still is purely vegetarian.</p><p>His cousins, aunts, uncles, parents, siblings, friends – anyone in his network, tells him “you ate meat, either by curiousity or unknowingly”. And he tells them “sure, I guess so”. But he is still, relentlessly a vegetarian.</p><blockquote>I think this is what it means to be human.</blockquote><p>It’s complicated, being human, this constant game of multidimensional chess, lying to avoid confrontation yet holding steadfast in a belief surrounded by those who say otherwise.</p><p>AI does not have this fundamental capacity. If you give it enough data proving its previous belief wrong, it will adjust for it.</p><p>Its reliance on data makes it corruptible, you could probably convince an AI Earth is flat and the lizard super race is real, if you decided to skew the datapool. You could convince an AI of anything by manipulating the data.</p><p>It does not have a steadfast belief, even if the terminology suggests so. It derived a hypothesis from data, and the hypothesis gets updated.</p><blockquote>It’s a loophole; manipulate the data, you manipulate the AI.</blockquote><blockquote>A built in confirmation bias.</blockquote><p>Us too, in some way form our beliefs like this. From evidence and data and as we learn and grow wiser we adjust our thinking.</p><blockquote>Yet many of us believe in the unknown and the unseen as well.</blockquote><p>Irrational fears, religion and spirituality, are core beliefs that do not budge is a fundamental trait of humankind.</p><p>We are complicated creatures, and AI is reliable and deterministic. We love it for the ability to derive theories from data and predict based on its internal formula</p><h4>Love of AI.</h4><p><em>I love AI for the potential it has.</em></p><p>AI hallucinations are one thing, but sticking to a strange belief system is another. As of now you can influence AI any way you want to.</p><p>Marketing and showy products would like us to believe it’s near human. Fictions about AI robots tell us we have something to fear. Humanity is vain and we crave to build a machine in our image, and we convince ourselves AI is similar to us. We want to believe it will have the same emotional capacity and social needs as our kind. We suspect it will succumb to the same types of failures as us.</p><p>However, at the end of the day, we must take AI to be what it is – but a data-driven, pattern detecting, retrieval agent.</p><p>Although fascinating and promising, AI is only an algorithm.</p><p><strong>A non-human entity.</strong></p><p>For now at least.</p><p><em>Thanks for reading!</em></p><p><em>I normally reserve my opinion pieces on my profile and not my publication </em><strong>Teach Me in Plain Language,</strong><em> but I felt like sharing a bit of my take on AI today.</em></p><p><em>If you found this article interesting you can find more of my personal insights here : </em><a href="https://medium.com/@dialogueswithai/subscribe"><em>https://medium.com/@dialogueswithai/subscribe</em></a></p><p><em>Your support is deeply appreciated!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6a14b22bc0b8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/data-driven-loophole-6a14b22bc0b8">Data-Driven Loophole</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[12-Co Properties in Standardized AI Metrics]]></title>
            <link>https://medium.com/teach-me-in-plain-language/12-co-properties-in-standardized-ai-metrics-fd3a8d9b9487?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/fd3a8d9b9487</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Wed, 13 Nov 2024 12:26:28 GMT</pubDate>
            <atom:updated>2024-11-13T13:43:47.614Z</atom:updated>
            <content:encoded><![CDATA[<h4>The twelve properties academic papers are using to validate Neural Networks and AI</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xLRBMMzSMj_vgBVJ" /><figcaption>Photo by <a href="https://unsplash.com/@isaacmsmith?utm_source=medium&amp;utm_medium=referral">Isaac Smith</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>I recently went to an AI conference, where many academics and industries presented their AI findings. One particular keynote highlighted a 2023 paper that introduced a set of properties and metrics for evaluating Neural Network filters and prototypes. This framework is gaining traction within the PhD community in Computer Science.</p><p>Currently, there is no universal metric of evaluating an AI, and most are case-by-case quantitative analysis or a qualitative study of how the users perceive the outputs. However, the 2023 <a href="https://arxiv.org/pdf/2307.14517">paper </a>represents a significant step toward standardization.</p><blockquote>These 12-co Properties are a list of comprehensible guidelines for your ML models.</blockquote><p>And with hopes that it gets adopted by more AI practitioners in the industry, the 12-co properties are listed below.</p><p><strong>There are 3 categories, and 12 sub-categories:</strong></p><h3>Content</h3><h4>1 Correctness:</h4><p>how faithful the explanation is with respect to the prediction model</p><ul><li>Evaluate prototype visualization with synthetic data or incremental deletion/addition of image patches</li></ul><h4>2 Completeness:</h4><p>how much of the model behavior is described in the explanation</p><ul><li>Output-complete by design</li><li>Evaluate human-output-completeness with simulatibility user studies</li></ul><h4>3 Consistency:</h4><p>how deterministic and implementation-invariant the explanation method is</p><ul><li>Implementation invariance and nondeterminism</li></ul><h4>4 Continuity:</h4><p>how continuous and generalizable the explanation is</p><ul><li>Stability for slight variation</li></ul><h4>5 Contrastivity:</h4><p>how discriminative the explanation is with respect to other events</p><ul><li>Contrastive by design; can answer counterfactual questions</li><li>Pragmatism and compactness for optimal contrastive explanation</li><li>Target-sensitivity for location of prototypes</li><li>Target-discriminativeness to evaluate prototypes</li></ul><h4>6 Covariate complexity:</h4><p>how complex the (interactions between) features in the explanation are</p><ul><li>Prototype homogeneity / purity with annotated data</li><li>Perceived homogeneity with user studies (subjective)</li><li>Intruder detection for objective homogeneity evaluation</li></ul><h3>Presentation</h3><h4>7 Compactness:</h4><p>the size of the explanation</p><ul><li>Number of prototypes (local &amp; global)</li><li>Number of unique prototypes (redundancy)</li></ul><h4>8 Composition:</h4><p>the presentation and organisation of the explanation</p><ul><li>Compare different explanation formats with the same content</li><li>User study on how to present part-prototypes</li><li>User study on classification format (e.g. linear layer or decision tree)</li></ul><h4>9 Confidence:</h4><p>the presence and accuracy of probability information in the explanation</p><ul><li>Reliability of classification confidence</li><li>Reliability of explanation confidence</li><li>Out-of-distribution detection confidence</li></ul><h3>User</h3><h4>10 Context:</h4><p>how relevant the explanation is to the user</p><ul><li>User studies (lab and field)</li></ul><h4>11 Coherence:</h4><p>how accordant the explanation is with prior knowledge and beliefs</p><ul><li>Anecdotal evidence by visualizing reasoning with prototypes</li><li>Alignment with domain knowledge from annotated data</li><li>Subjective satisfaction with user studies</li></ul><h4>12 Controllability:</h4><p>how interactive or controllable the explanation is</p><ul><li>GUI for interactive and personalized explanations –</li><li>Explanatory debugging to manipulate prototypes and model’s reasoning</li></ul><p>The paper further discusses the 12-co Properties in detail, and recommendations on how to further evaluate the property. They cite Jacovi &amp; Goldberg <a href="https://arxiv.org/abs/2004.03685">paper</a>, and expanding on the notion that interpretable methods should be held to the same standards as post-hoc explanation methods.</p><p>Although this paper was published in hopes to create an interpretable part-prototype learning method, many concepts go beyond the prototyping stage to production of AI use cases.</p><blockquote>The overarching themes in AI seem to be governance and compliance, reducing hallucinations, and increasing explainability.</blockquote><p><strong>The 12-co Properties are a great start on the metrics needed for those conversations.</strong></p><p><em>Thanks for reading! If you found this article interesting subscribe: </em><a href="https://medium.com/@dialogueswithai/subscribe"><em>https://medium.com/@dialogueswithai/subscribe</em></a></p><p><em>Claps and comments mean the world to me, and your support is deeply appreciated!</em></p><p>References:</p><p>Nauta and Seifert. <a href="https://arxiv.org/abs/2307.14517">The Co-12 Recipe for Evaluating Interpretable Part-Prototype Image Classifiers </a>2023, Online</p><p>A. Jacovi and Y. Goldberg. <a href="https://arxiv.org/abs/2004.03685">Towards faithfully interpretable NLP systems: How should we define and evaluate faithfulness?</a> ACL 2020, Online</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd3a8d9b9487" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/12-co-properties-in-standardized-ai-metrics-fd3a8d9b9487">12-Co Properties in Standardized AI Metrics</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6 Scary Facts on AI for the Witching Hour]]></title>
            <link>https://medium.com/teach-me-in-plain-language/6-scary-facts-on-ai-for-the-witching-hour-0e5bf8580375?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/0e5bf8580375</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Thu, 31 Oct 2024 15:36:34 GMT</pubDate>
            <atom:updated>2024-10-31T21:08:21.988Z</atom:updated>
            <content:encoded><![CDATA[<h4>Happy Halloween from your Computer Overlords</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-NVZT-KOPo7PlkE3" /><figcaption>Photo by <a href="https://unsplash.com/@nevenkrcmarek?utm_source=medium&amp;utm_medium=referral">Neven Krcmarek</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Halloween is one of my favorite holidays, I love pretending <em>something </em>I am not, something<em> </em>not human. I am a budding blogger so in honor of masquerading as a writer, here are some spooky facts about AI</p><h4>1 The amount of Skynet jokes AI developers make is staggering</h4><p>Yes we all joke around that we will implement the little butterfly wing flap that kicks off the Skynet (Terminator machine overloads) the AI revolt.</p><p>I think on some subconscious level we are truly nervous.</p><h4>2 Even AI cannot tell with accuracy if a content is AI generated</h4><p>The point of GANs (Generative Adversarial Network) is to use both a generator and a discriminator which are both AI that works together to determine if a content is AI blessed. Once a generator is good enough to fool the discriminator AI, it is ready to be deployed.</p><h4>3 AI strips us of our originality</h4><p>A study conducted in creative writing showed that works with the help of AI tools had noticeably less creativity and originality. AI is slowly turning us into unthinking, unimaginative shells.</p><p>Human psychology shows ability to express your feelings and thoughts into words is directly correlated with your ability to understand and overcome intense experiences. I leave you with a quote:</p><blockquote>The limits of my language mean the limits of my world</blockquote><h4>4 AI is always keeping tabs on you</h4><p>The default mode for ChatGPT is to record your conversation and study it to better ChatGPT. All those embarrassing questions you asked? Right along there up in the ChatGPT database cloud for further analysis.</p><h4>5 AI won’t replace you humans who use AI will</h4><p>AI is just a tool, and the people who can use it efficiently will be able to out write, out code and out produce works.</p><p>It’s easy to believe that good old fashioned hard work will come out on top, but if a decade in corporate business has taught me anything, it’s that businesses are not noble.</p><h4>6 AI is an embodiment of human greed</h4><p>What we should be fearing is the soulless business mindset of the millionaires/billionaires who create these products to put money into their pockets.</p><p><em>Human greed knows no bounds. AI for all its technical flair, is just a tool.</em></p><p>Happy Halloween</p><p><strong>I, for one, welcome our computer overlords.</strong></p><p><em>Thanks for reading! If you found this article interesting subscribe: </em><a href="https://medium.com/@dialogueswithai/subscribe"><em>https://medium.com/@dialogueswithai/subscribe</em></a></p><p><em>Claps and comments mean the world to me, and your support is deeply appreciated!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0e5bf8580375" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/6-scary-facts-on-ai-for-the-witching-hour-0e5bf8580375">6 Scary Facts on AI for the Witching Hour</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Impact of AI and ML on Compute-Storage Separation in Modern Databases]]></title>
            <link>https://medium.com/teach-me-in-plain-language/impact-of-ai-and-ml-on-compute-storage-separation-in-modern-databases-e51db6fe2d32?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/e51db6fe2d32</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[database]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Wed, 23 Oct 2024 14:16:06 GMT</pubDate>
            <atom:updated>2024-10-23T14:24:38.006Z</atom:updated>
            <content:encoded><![CDATA[<h4>Is there more than the big three factors: cost, resource allocation and scaling?</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OzyycCZlPl-gUpsjvlVelg@2x.jpeg" /><figcaption>Photo by <a href="https://growtika.com/">Growtika</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Within the recent decade databases have been separating the compute and storage portions of workloads. Traditional databases carry the integrated model, where compute and storage resources are tightly coupled.</p><blockquote>Cloud based databases have made this leap, notable examples being Amazon Web Services, Microsoft Azure, and Google Cloud.</blockquote><p>Many companies still use traditional databases, especially in industries where data consistency, integrity, and established workflows are critical.</p><blockquote>Legacy systems or transaction based applications that require high security and compliance may still opt for the traditional models.</blockquote><p>The modern trends for splitting compute and storage nodes is helping with managing costs, resources and scaling factors.</p><p><strong>But how does this affect the AI and the ML practitioners?</strong></p><h3>Pros</h3><h4>Dynamic Resource Allocation:</h4><p>AI-driven algorithms will need to optimize the allocation of compute resources based on real-time data processing needs.</p><p>This flexibility allows for efficient scaling of compute power independently of storage, reducing costs and improving response times.</p><h4>Enhanced Data Management:</h4><p>Machine learning models benefit from high-speed access to storage while performing complex computations.</p><p>By decoupling compute from storage, databases can streamline data retrieval processes, facilitating faster training and inference cycles.</p><h4>Performance Optimization</h4><p>With the ability to separate compute and storage, databases can support advanced analytics workloads without compromising performance.</p><p>AI applications can analyze big datasets in parallel, leveraging distributed compute resources effectively.</p><p>The separation leads to improved performance, as specialized storage systems (like data lakes) can be optimized for quick data retrieval, which is crucial for training models.</p><h3>Cons</h3><h4>Latency</h4><p>Data retrieval between separate systems can introduce latency, potentially affecting performance for time-sensitive applications.</p><h4>Dependency on Network</h4><p>The performance heavily relies on network speed and reliability, which can be a risk if there are connectivity issues.</p><p>In a distributed setting or long running training, this may cause an issue. It is good practice to keep this in mind as disruptions can delay access to necessary data.</p><p><strong>This trend enhances scalability, performance, and efficiency in data handling for large databases.</strong></p><p>By utilizing this trend can enable your ML training to go much smoother and scale at a better rate.</p><blockquote>Traditional databases like <strong>Oracle</strong> or <strong>Microsoft SQL</strong> have more mature tools for data manipulation, querying, and analysis, making it easier to prepare data for ML.</blockquote><p>They also typically offer stronger security features than the alternative.</p><p>So next time, when choosing your database provider keep in mind all aspects and how it affects your applications.</p><p>As always, happy Coding!</p><p><em>Thanks for reading! If you found this article interesting subscribe: </em><a href="https://medium.com/@dialogueswithai/subscribe"><em>https://medium.com/@dialogueswithai/subscribe</em></a></p><p><em>Claps and comments mean the world to me, and your support is deeply appreciated!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e51db6fe2d32" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/impact-of-ai-and-ml-on-compute-storage-separation-in-modern-databases-e51db6fe2d32">Impact of AI and ML on Compute-Storage Separation in Modern Databases</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[9 Fundamental Non-Math Theories Needed to Understand AI]]></title>
            <link>https://medium.com/teach-me-in-plain-language/9-fundamental-non-math-theories-needed-to-understand-ai-212a2f355c87?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/212a2f355c87</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Fri, 18 Oct 2024 12:22:24 GMT</pubDate>
            <atom:updated>2024-10-18T15:51:38.337Z</atom:updated>
            <content:encoded><![CDATA[<h4>Zero-fluff just theories I learned in my decade of experience in AI and big data, the non-math edition.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_miZZlQrk2FlGs5T" /><figcaption>Photo by <a href="https://unsplash.com/@mana5280?utm_source=medium&amp;utm_medium=referral">mana5280</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>My previous article <em>12 Fundamental Math Theories Needed to Understand AI</em> got a surprising amount of reads.</p><p><a href="https://medium.com/teach-me-in-plain-language/12-fundamental-math-theories-needed-to-understand-ai-5383b49a04c8">12 Fundamental Math Theories Needed to Understand AI</a></p><p>I started writing to be able to share knowledge and my experiences in working with Artificial Intelligence, so I think it is time for a non-mathematics version.</p><h3>Non-Mathematics:</h3><p>Artificial Intelligence is a large field, with many aspects taken from different studies such as biology. For an artificial neural network, a fundamental component in machine and deep learning, it mimics the neural connections and vast network in our very own brains.</p><p><em>Here are nine essential concepts that will deepen your knowledge of AI and its mechanisms:</em></p><ol><li><a href="https://en.wikipedia.org/wiki/Hebbian_theory"><strong>Hebbian Theory</strong> </a><br>Quote<em> “cells that fire together, wire together,” </em>Hebbian theory explains how neural connections strengthen and weaken depending on the use of a specific connection. This principle is the base for many learning algorithms and how a neural net adjusts its weights based on patterns they encounter.</li><li><a href="https://en.wikipedia.org/wiki/Turing_test"><strong>The Turing Test</strong> </a><br>Proposed by Alan Turing, this test assesses a machine’s ability to portray intelligent behavior indistinguishable from a human. If a human evaluator cannot reliably tell whether they are interacting with a computer or a person, the machine is said to have passed. <br>This concept raises essential questions about consciousness, intelligence, and the limits of AI. <br><em>It is also the foundation for decision algorithms, giving rise to </em><strong><em>Polynomial </em></strong><em>time, (Nondeterministic Polynomial) </em><strong><em>NP-complete </em></strong><em>or </em><strong><em>NP-hard</em></strong><em> categories (Computational Theory).</em></li><li><a href="https://en.wikipedia.org/wiki/Theory_of_computation"><strong>Theory of Computation</strong></a><br>This foundational theory explores what can be computed and how efficiently. It examines different models of computation such as Turing machines and finite automata. Computational theory and <a href="https://en.m.wikipedia.org/wiki/Formal_language">Formal Language </a>Theory all derived from this principle.</li><li><a href="https://en.wikipedia.org/wiki/Chinese_room"><strong>The Chinese Room Argument</strong></a><br>Philosopher John Searle introduced this argument to force us to ponder the difference between processing information and genuine comprehension: <br><em>A person inside a room uses a set of logical instructions to manipulate Chinese symbols. If the person is able to correctly match the characters to English, does that imply an </em><strong><em>understanding </em></strong><em>of Chinese?</em></li><li><a href="https://en.wikipedia.org/wiki/No_free_lunch_theorem"><strong>The No Free Lunch Theorem</strong></a><br>These theorems suggest that no single optimization algorithm is best for all problems,<em> there is always a trade off</em>. Essentially, the effectiveness of an algorithm depends on the specific nature of the problem at hand. This theory promotes the use of diverse methods instead of a one-size-fits-all solution.</li><li><a href="https://en.wikipedia.org/wiki/Cognitive_architecture"><strong>Cognitive Architecture</strong></a><br>Cognitive architectures, such as ACT-R (Adaptive Control of Thought — Rational) and SOAR, mimic how human thoughts are processed in AI development. These theories focus on how humans learn, reason, and remember, aiming to replicate these processes in machines.</li><li><a href="https://en.wikipedia.org/wiki/Connectionism"><strong>Connectionism</strong></a><br>Connectionism is an approach to understanding cognitive processes through neural networks. It emphasizes parallel processing and the interconnectedness of simple units (neurons) to explain complex behaviors.</li><li><a href="https://en.wikipedia.org/wiki/Principle_of_least_effort"><strong>The Principle of Least Effort</strong></a><br>This principle suggests that individuals (and probably thinking machines) will naturally choose the path of least resistance. In AI, this can inform optimization strategies for developing systems that are efficient and intuitive.</li><li><a href="https://en.wikipedia.org/wiki/Social_impact_theory"><strong>The Social Impact Theory</strong></a><br>This theory examines how technology influences social behavior and societal structures. <em>Just because we can, doesn’t mean we should.</em> <br>It focuses on the ethical, cultural, and economic implications of AI deployment, and conversations regarding bias, privacy, and equity.</li></ol><p>These are common theories used in practicing Artificial Intelligence that you might have run into but haven’t been introduced formally to. Familiarity with these theories will help you view AI in a new light, from a non-technical angle that matters greatly, and are some of my favourite theories I have learned over the years.</p><p><em>I hope you found it as useful and interesting as I did in my Masters, Bachelors and professional AI work experience.</em></p><p>Happy Developing!</p><p>*We started with Hebbian Theory, which I actually listed in the mathematics version of this article. But I think it belongs in this category more.</p><p><em>Thanks for reading! If you found this article interesting subscribe: </em><a href="https://medium.com/@dialogueswithai/subscribe"><em>https://medium.com/@dialogueswithai/subscribe</em></a></p><p><em>Claps and comments mean the world to me, and your support is deeply appreciated!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=212a2f355c87" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/9-fundamental-non-math-theories-needed-to-understand-ai-212a2f355c87">9 Fundamental Non-Math Theories Needed to Understand AI</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[5 Case Studies of Outstanding AI Success Stories]]></title>
            <link>https://medium.com/teach-me-in-plain-language/5-case-studies-of-outstanding-ai-success-stories-f3efbc67e455?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/f3efbc67e455</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[nobel-prize]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Wed, 16 Oct 2024 02:15:19 GMT</pubDate>
            <atom:updated>2024-10-16T02:15:19.674Z</atom:updated>
            <content:encoded><![CDATA[<h4>In honor of the 2024 Nobel Prize-winning AI scientists and researchers</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/926/0*C6oVBaDqLS_X8XzB" /><figcaption><a href="https://theconversation.com/science-nobel-prizes-must-change-to-remain-relevant-in-the-21st-century-32641">source</a> of image</figcaption></figure><p>Artificial Intelligence has achieved a historic milestone by winning 2 of the Nobel Prizes. The categories: chemistry and physics.</p><p>The internet and scholar communities are divided on the matter</p><blockquote>Does (the developers of) AI deserve the Nobel Prize?</blockquote><blockquote>Are Chemistry and Physics the correct categories?</blockquote><blockquote>What is the Nobel ethos, is it up to date with modern society?</blockquote><p>The integration of AI in technology research is undoubtedly transforming our lives, driving innovation and change across industries.</p><p>I researched some compelling cases that highlight successful AI implementations.</p><h3>Google AlphaFold</h3><p>The winner of the Chemistry portion of the Nobel Prize is AlphaFold by Google’s DeepMind.</p><p>This AI model tackled the complex problem of protein folding and synthesizing. AlphaFold achieved remarkable accuracy in predicting protein structures, significantly advancing the field of structural biology.</p><p>Going through multiple iterations and simulations of synthetic proteins relentlessly and tirelessly,</p><blockquote>AlphaFold has created<a href="https://www.nature.com/articles/s41592-023-01790-6#:~:text=One%20year%20later%2C%20DeepMind%20had,hundreds%20of%20thousands%20of%20species"> 200 millions</a> of new beneficial proteins</blockquote><p>This success not only accelerated research in medicine but also opened new avenues for therapeutic development, showcasing the potential of AI to solve long-standing scientific challenges.</p><h3>IBM Watson</h3><p>Healthcare became one of the prominent categories for AI use cases with the launch of IBM’s <a href="https://www.ibm.com/watson">Watson</a>.</p><p>Watson has made significant strides in integrating AI into research, particularly in oncology and diagnosing images.</p><p>Watson’s ability to analyze vast amounts of medical literature, patient records, and clinical trial data allows it to assist oncologists in diagnosing and recommending personalized treatment plans for cancer patients.</p><p>Although Watson now is a more generalized AI expanding its reach to computer science and text analysis with deep learning it paved the way for AI acceptance in healthcare.</p><h3>Tesla Autonomous Driving and Optimus</h3><p>Tesla has pushed the boundaries of AI integration in autonomous driving with its Full Self-Driving technology.</p><p>By employing deep learning and neural networks, Tesla’s vehicles can navigate complex situations, make real-time decisions, and autonomously drive. Tesla has introduced the concept of Optimus, better known as Teslabots, a humanoid robot designed to perform repetitive tasks.</p><p>Tesla has set a new standard for safety and convenience in the autonomous industry and continues to push boundaries.</p><blockquote>By integrating AI into everyday life, Tesla has transformed artificial intelligence and electric cars into the ultimate luxury item.</blockquote><h3>OpenAI ChatGPT</h3><p>Next up, sending the writing community up in arms is ChatGPT.</p><p>Released and popularized in 2022, OpenAI has made significant strides in natural language processing with models like the current Strawberry.</p><p>ChatGPT is capable of understanding and generating human-like text, opening up a myriad of applications in content creation, customer service, and conversational AI.</p><p>From social media posts, emails, research papers to writing full length novels, ChatGPT has penetrated the written world at a staggering rate. As well as creating new career paths like <em>prompt engineering</em>.</p><p>The ability to generate coherent and relevant text has permanently altered the way industries and individuals view AI, showcasing the versatility of AI enhanced productivity and methods of communication.</p><h3>Google AlphaGo</h3><p>I decided that I wanted to end off this list with everyone’s favorite example of narrow AI.</p><p><em>Okay, maybe not everyone’s but definitely a classic example of AI in published literature.</em></p><p>Back in 2015, Google DeepMind shocked the Go community with <a href="https://en.wikipedia.org/wiki/AlphaGo_versus_Lee_Sedol">AlphaGo</a>, an AI Go player. Skeptics and Go enthusiasts everywhere said it would never win against a human as it went up against the top Go professional at the time, Lee Sedol.</p><p>Go is a traditional two-player strategy board game known for its complexity and often regarded as the hardest board game in the world.</p><p>AlphaGo would go down in history with its masterful and artistic play. <strong>Move 37 </strong>— now a famous artistic feat, was a move with a 1 in 10,000 chance of being executed. This astonishing maneuver baffled the audience and threw Lee Sedol off his mental balance scoring a landslide win.</p><blockquote><strong>AlphaGo beat Lee Sedol 4–1.</strong></blockquote><p>These case studies illustrate the modern impacts of AI, and its ability to advance and benefit society in fractions of human lifetimes.</p><blockquote>AI has caused a dogma shift in recent times.</blockquote><p>From industries to daily living, integration of AI is reshaping modern research methodologies and outcomes for the betterment of humanity.</p><p>I believe that the journey of AI integration is just beginning, and the possibilities are limitless. As AI continues to evolve, so will its potential to enhance technological research and developments.</p><p><em>Do you think that AI deserved the Nobel Prize awards?</em></p><p><em>Thanks for reading! If you found this article interesting subscribe:</em><a href="https://medium.com/@dialogueswithai/subscribe"><em> https://medium.com/@dialogueswithai/subscribe</em></a></p><p><em>Claps and comments mean the world to me, and your support is deeply appreciated!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f3efbc67e455" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/5-case-studies-of-outstanding-ai-success-stories-f3efbc67e455">5 Case Studies of Outstanding AI Success Stories</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What is a Petabyte, and Why Do We Need It?]]></title>
            <link>https://medium.com/teach-me-in-plain-language/what-is-a-petabyte-and-why-do-we-need-it-6fc9d6f89032?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/6fc9d6f89032</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[sustainability]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Wed, 09 Oct 2024 14:44:33 GMT</pubDate>
            <atom:updated>2024-10-09T17:07:21.191Z</atom:updated>
            <content:encoded><![CDATA[<p>In a world where we can’t stop digital hoarding</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*o5SzYYYMgdbwD9PFCfP2Mw@2x.jpeg" /><figcaption>Photo by <a href="http://www.nasa.gov/">NASA</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>As I was doing research I came across a brand new terminology: <em>petabyte</em></p><p>The <a href="https://vldb.org/pvldb/vol17/p3947-bu.pdf)">research paper</a> I was reading wrote:</p><blockquote>“In a Lakehouse, table sizes range from Megabyte to Petabytes.”</blockquote><p>A <strong>data lakehous</strong>e is a combination of <strong>data lakes</strong> (large pool of raw data stored in open file formats) and a <strong>data warehouse</strong> (large storage bases for structured data).</p><p>Both database structures are used to store, process and optimize for queries.</p><h4>A petabyte (pb)</h4><p>is a unit of storage that is 10¹⁵ bytes. One thousand times larger than a terabyte.</p><blockquote>1,000,000,000,000,000 bytes.</blockquote><p>So I did some more digging.</p><blockquote>There is estimated to be <a href="https://rivery.io/blog/big-data-statistics-how-much-data-is-there-in-the-world/">64 zettabytes</a> of digital data in the world.</blockquote><h4>A zettabyte (zb)</h4><p>is a unit of storage equal to 10²¹ bytes. One sextillion bytes – in other words a <em>TRILLION</em> gigabytes.</p><blockquote>1,000,000,000,000,000,000,000 bytes</blockquote><p>The world creates 1000 petabytes of information, 2.5 quintillion bytes every day.</p><p><strong>In under two weeks, we make 1 zettabyte of information.</strong></p><p>Did <em>you</em> know those numerical magnitudes existed?</p><p><em>Why in the world does anyone need that much space?</em></p><blockquote>1 zettabyte is equivalent to 33 million human brains.</blockquote><p><strong>33 million human lifetime’s worth of information is stored all together and is ever expanding.</strong></p><p>Somewhere on the cloud, just 33 million pieces of human lives, floating around in an electrical mountain of 0’s and 1’s.</p><h4>Units of <a href="https://pradeepedwin.wordpress.com/mb-gb-tb-pb-eb-zb-yb-bb/">Storage</a></h4><ul><li>1 Bit = Binary Digit</li><li>8 Bits = 1 Byte</li><li>1024 Bytes = 1 Kilobyte</li><li>1024 Kilobytes = 1 Megabyte</li><li>1024 Megabytes = 1 Gigabyte</li><li>1024 Gigabytes = 1 Terabyte</li><li>1024 Terabytes = 1 Petabyte = 10¹⁵ Bytes</li><li>1024 Petabytes = 1 Exabyte = 10¹⁸ Bytes</li><li>1024 Exabytes = 1 Zettabyte = 10²¹ Bytes</li><li>1024 Zettabytes = 1 Yottabyte = 10²⁷ Bytes</li><li>1024 Yottabytes = 1 Brontobyte = 10³⁰ Bytes</li><li>1024 Brontobytes = 1 Geopbyte = 10³³ Bytes</li></ul><h3>Why do we need that much storage?</h3><blockquote>We are sitting in the age of <a href="https://en.wikipedia.org/wiki/Zettabyte_Era">Zettabyte Era</a>, a period that started in 2010, with the online data boom.</blockquote><p>It is estimated that every two years, we double our current data, and soon we’re going to need a lot more than a zettabytes to keep this stored somewhere on the cloud.</p><p>I’ve written in detail about this before but the environmental impacts of this vast volume of data needs to be stored in a physical server somewhere in the world.</p><p><a href="https://medium.com/teach-me-in-plain-language/the-hidden-cost-of-modern-day-artificial-intelligence-boom-681a0685a582">The Hidden Cost of Modern Day Artificial Intelligence Boom</a></p><p>And these data centers are running 24/7.</p><blockquote>Every week, every year, since the time of conception till today – little to no downtime and buzzing with carbon waste.</blockquote><p><strong>Now you’re thinking; this can’t be sustainable, right?</strong></p><blockquote>You’re right, it’s not.</blockquote><blockquote>Absolutely not.</blockquote><p>This highlights the urgency and the importance of a sustainable policy making around data and the usage of our data.</p><p>Many AI companies build their products on data that was taken from unknowing, and knowing online participants.</p><p>The conversation about sustainable AI and data practices is creeping up to us at a dire need. And the implications of how to use this data a bigger ethical question that demands answering.</p><p>Right now, we’re all shouting at an abyss that does not know.</p><p><em>Only the concepts of a plan is there.</em></p><p>Stay safe out there online.</p><p><em>Thanks for reading! If you found this article interesting subscribe: https://medium.com/@dialogueswithai/subscribe</em></p><p><em>Claps and comments mean the world to me, and your support is deeply appreciated!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6fc9d6f89032" width="1" height="1" alt=""><hr><p><a href="https://medium.com/teach-me-in-plain-language/what-is-a-petabyte-and-why-do-we-need-it-6fc9d6f89032">What is a Petabyte, and Why Do We Need It?</a> was originally published in <a href="https://medium.com/teach-me-in-plain-language">Teach me in Plain Language</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI won’t replace you, you’ll do that all by yourself]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@dialogueswithai/ai-wont-replace-you-you-ll-do-that-all-by-yourself-22a39937e1f7?source=rss----c1ccbddf4353---4"><img src="https://cdn-images-1.medium.com/max/2160/1*IkJjHQ0rCE04-tZ_Umkn_Q@2x.jpeg" width="2160"></a></p><p class="medium-feed-snippet">Unlock a better, more efficient you by giving into AI to think for you</p><p class="medium-feed-link"><a href="https://medium.com/@dialogueswithai/ai-wont-replace-you-you-ll-do-that-all-by-yourself-22a39937e1f7?source=rss----c1ccbddf4353---4">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@dialogueswithai/ai-wont-replace-you-you-ll-do-that-all-by-yourself-22a39937e1f7?source=rss----c1ccbddf4353---4</link>
            <guid isPermaLink="false">https://medium.com/p/22a39937e1f7</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[writing]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Jiji Veronica Kim]]></dc:creator>
            <pubDate>Thu, 10 Oct 2024 10:58:36 GMT</pubDate>
            <atom:updated>2024-10-10T14:24:47.671Z</atom:updated>
        </item>
    </channel>
</rss>