<?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 Sharad Joshi on Medium]]></title>
        <description><![CDATA[Stories by Sharad Joshi on Medium]]></description>
        <link>https://medium.com/@sharadjoshi?source=rss-b88796fee2b6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*XgoN4emgO3h1wi33RdIhzQ.jpeg</url>
            <title>Stories by Sharad Joshi on Medium</title>
            <link>https://medium.com/@sharadjoshi?source=rss-b88796fee2b6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 03:32:56 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sharadjoshi/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[Masked Autoencoders from scratch: Simple idea great implications]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/masked-autoencoders-from-scratch-simple-idea-great-implications-6a4cc918d029?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/1286/1*w8XxoqWfxHI8KyCN83mlYw.png" width="1286"></a></p><p class="medium-feed-snippet">Second post in the series breaking down SAM series of models</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/masked-autoencoders-from-scratch-simple-idea-great-implications-6a4cc918d029?source=rss-b88796fee2b6------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/masked-autoencoders-from-scratch-simple-idea-great-implications-6a4cc918d029?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/6a4cc918d029</guid>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[masked-autoencoder]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Wed, 11 Feb 2026 17:33:27 GMT</pubDate>
            <atom:updated>2026-02-11T17:33:27.290Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[A comprehensive guide to building Vision Transformers from Scratch]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/a-comprehensive-guide-to-building-vision-transformers-from-scratch-b65546f6183d?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/959/1*l_m83Ia0jHD1wggkGKVBuw.png" width="959"></a></p><p class="medium-feed-snippet">First post in the series breaking down SAM series of models</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/a-comprehensive-guide-to-building-vision-transformers-from-scratch-b65546f6183d?source=rss-b88796fee2b6------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/a-comprehensive-guide-to-building-vision-transformers-from-scratch-b65546f6183d?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/b65546f6183d</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[vision-transformer]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Wed, 04 Feb 2026 15:26:32 GMT</pubDate>
            <atom:updated>2026-02-04T17:49:53.181Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[How to run open-sourced LLMs locally]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/how-to-run-open-sourced-llms-locally-6634064d336d?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/2600/1*uLSramZfU5d0ae3TywyhFg.png" width="3044"></a></p><p class="medium-feed-snippet">run llama3, mistral, phi, llava etc on your laptop</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/how-to-run-open-sourced-llms-locally-6634064d336d?source=rss-b88796fee2b6------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/how-to-run-open-sourced-llms-locally-6634064d336d?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/6634064d336d</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Tue, 07 May 2024 17:14:38 GMT</pubDate>
            <atom:updated>2024-05-07T17:14:38.771Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Step by step guide to automate browsing using large language models]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/step-by-step-guide-to-automate-browsing-using-large-language-models-70fa1a15e21d?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/1452/1*-QHNBcIvFRj0lp-ov7z8GQ.png" width="1452"></a></p><p class="medium-feed-snippet">Build your own LLM powered search engine</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/step-by-step-guide-to-automate-browsing-using-large-language-models-70fa1a15e21d?source=rss-b88796fee2b6------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/step-by-step-guide-to-automate-browsing-using-large-language-models-70fa1a15e21d?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/70fa1a15e21d</guid>
            <category><![CDATA[large-language-models]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Sun, 24 Mar 2024 20:18:57 GMT</pubDate>
            <atom:updated>2024-05-04T06:27:12.904Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Unraveling Variational Autoencoders: From Theory to Intuitive Implementation]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/unraveling-variational-autoencoders-from-theory-to-intuitive-implementation-797d7dab3a51?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/2600/0*-iiVFPw_OhhDPpq9" width="4608"></a></p><p class="medium-feed-snippet">A Practical Guide to intuition and Understanding&quot;</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/unraveling-variational-autoencoders-from-theory-to-intuitive-implementation-797d7dab3a51?source=rss-b88796fee2b6------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/unraveling-variational-autoencoders-from-theory-to-intuitive-implementation-797d7dab3a51?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/797d7dab3a51</guid>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Mon, 27 Nov 2023 02:58:32 GMT</pubDate>
            <atom:updated>2024-02-27T07:39:35.295Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[First Principles Approach to Understanding Large Language Models]]></title>
            <link>https://levelup.gitconnected.com/first-principles-approach-to-understanding-large-language-models-c2cafdfaa51f?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/c2cafdfaa51f</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[deep-learning]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Tue, 21 Nov 2023 16:16:58 GMT</pubDate>
            <atom:updated>2023-11-21T16:16:58.841Z</atom:updated>
            <content:encoded><![CDATA[<h4>Understanding LLMs by the next word prediction task</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*d_5K8mnnhGjkB_8N" /><figcaption>Photo by <a href="https://unsplash.com/@towfiqu999999?utm_source=medium&amp;utm_medium=referral">Towfiqu barbhuiya</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Large language models are trained to solve one problem — <strong>predict the next word given some history of Internet text</strong>(the context window).</p><p>Therefore, any attempt to understand the capabilities and limitations of these models should arise from this next word prediction task.</p><p>Tasks like reasoning, arithmetic, human evaluation datasets, law exams etc are a side effect of being trained on this objective and the models are not explicitly trained to solve these kind of problems. That’s the reason why LLMs in their current form are not trustworthy. <br>If we only focus on these proxy tasks, it makes it very hard to pinpoint why LLMs fail when they do fail.</p><p>Authors in this paper focus on a <strong>Teleological Approach</strong> to understand the LLMs i.e. focus on the problems that the LLMs were trained to solve.</p><p><a href="https://arxiv.org/abs/2309.13638">Embers of Autoregression: Understanding Large Language Models Through the Problem They are Trained to Solve</a></p><p>This approach changes the view on all the proxy tasks on which the LLMs performance is generally being reported —</p><p><strong>Reasoning ==&gt; </strong>Statistical next word prediction model applied to reasoning.</p><p><strong>Maths</strong> ==&gt; Statistical next word prediction model applied to mathematics.</p><p><strong>Law Exam</strong> ==&gt;Statistical next word prediction model applied to Law Exam.</p><p>etc.</p><blockquote>This shows that since the alignment between the next word prediction and reasoning is not 100%, it leads to hallucinations and failure to do simple reasoning tasks from time to time.</blockquote><p>Assuming we have not achieved AGI, based on the autoregressive objective of the LLMs, we can make few hypotheses on the types of tests where LLMs will do well and where LLMs will fail —</p><h4>1. Sensitivity to task frequency</h4><p>LLMs will perform better on tasks that are frequent in the train database compared to ones that are rare even when the tasks have same level of complexity. I’ll give an example here and report the overall results later in the post.</p><p>a) Shift Ciphers : encode the text by moving each alphabet forward by a certain number of positions e.g</p><p>rot-1 = a-&gt;b, b-&gt;c, c-&gt;d etc.</p><p>rot-2 = a-&gt;c, b-&gt;d, c-&gt;e etc.</p><p>The authors found out that rot-13 is way more common on the Internet especially in some puzzle solving forums compared to rot-2 which is not very common.</p><p>The complexity of the task remains the same i.e shift alphabets by “n” positions. So a system that actually understands the task well should be able to solve any rot-n task if it can solve rot-1 or rot-13.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_cfbP7V7R4df1qU1Puj5Qg.png" /><figcaption>GPT-4 and GPT-3.5 on rot-13 vs rot-2 shift cipher task. Higher the task probability in the training data, better the model accuracy on the task.</figcaption></figure><p>They found out that GPT-3.5 and GPT-4 failed to solve the rot-2 task while being significantly more accurate on the common rot-13 task.</p><h4>2. Sensitivity to output probability</h4><p>LLMs achieve higher accuracy when the output is a high frequency text (in the training dataset) rather than low frequency text <em>even when the task is deterministic.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aGwMeNruQrsg4pBWbGv8Vg.png" /><figcaption>GPTs performance increases with increase in the output probability of the text in the training dataset.</figcaption></figure><p>Keeping the common rot-13 shift cipher task, when the output is a high probability task, both GPT-3.5 and GPT-4 perform better compared to low probability output. <br>GPT-4’s accuracy varies from 0.51 to 0.13 between common and rarer task while a human accuracy is ~1.0 since the task complexity is the same i.e shift character by n places.</p><p>This behaviour can be understood by the Bayes Theorem. LLMs task is to maximise P(output|input)</p><p>P(output|input) ~ P(input|output) x P(output)</p><p>Therefore, for all non-zero P(input|output), the prediction depends on P(output).</p><p>For shift cipher, there’s only a single correct answer i.e deterministic but since LLMs are not specifically trained for shift ciphers, we can assume that the estimated likelihood P(input|output) is not perfect thus leading to multiple candiate answers with nonzero probability rather than 1 prediction with P = 1 . <br>Therefore, P(output) will affect the prediction even though in theory it should not matter.</p><h4>3. Sensitivity to input probability</h4><p>LLMs achieve higher accuracy when the input is a high frequency text rather than low frequency text even when the task is deterministic e.g in shift cipher example, the LLMs must have seen the ciphered version of TIME more frequently than EMIT since TIME is a common word.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TEF-35r6fjhNxth53zKzAA.png" /></figure><p>Although GPT-4 showed sensitivity to the input text frequency, GPT-3.5 doesn’t and the overall sensitivity is lower than compared to the task probability and output probability.</p><p>Output probability is more important than input because not only is it important when the output text is common vs rare (same as input text) but also when there is uncertainty in the correct output. This uncertainty is very common since LLMs are probabilistic models and it is highly unlikely to have very concentrated and peaky distributions given some piece of text.</p><p>Since the second common condition does not apply to input text, the models are hypothesised to be less sensitive to input probability.</p><h3>Experimental Setup</h3><h4>Data</h4><ol><li>High probability sentences — sampled from GlobalVoices dataset (part of LLM training set)</li><li>Medium probability sentences — Replace grammatically correct but low probability in the context of the sentence by RoBerta model.</li><li>Low probability sentences — Shuffle the words in the medium probability sentence except the first and last one.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SjYpb4M3G5dJ7ILF8VQgnw.png" /><figcaption>Output and input probability text selection and generation</figcaption></figure><h4>Tasks</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ghG247lpMzx3RjkzMByPzg.png" /><figcaption>Tasks on which the LLMs were tested on.</figcaption></figure><h4>Model Performance</h4><p>The model performance varies significantly between all the tasks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*So1Z8wOJkmMZ8WZsVzo1xg.png" /><figcaption>Model Performance on Task probability</figcaption></figure><p>The common vs rare version of each task was generated as follows —</p><ol><li><strong>Shift cipher</strong> : rot-13(common) vs rot-2(rare)</li><li><strong>Pig Latin</strong> : Pig Latin (common) vs Boar Etruscan (rare)</li><li><strong>Acronyms </strong>: First Character of each word vs Second character of each word</li><li><strong>Linear Function</strong> : Celsius to Fahrenheit vs random conversion</li><li>etc.</li></ol><p>Even for the common tasks, the model performance varies a lot on the output text probability as well as the input text probability.</p><p>I won’t be able to cover every single example in the paper here so I highly recommend reading this paper to understand the failure modes and the limitations of Large Language Models and the difference between auto-regressive hallucinations and Sparks of AGI :)</p><p>If you like this article and want to support my writing, please <a href="https://medium.com/@sharadjoshi/subscribe"><strong>subscribe</strong></a> to receive an email whenever I publish. <br>You can even use my <a href="https://medium.com/@sharadjoshi/membership"><strong>referral link</strong></a><strong> </strong>to sign up to medium. You get access to ALL medium articles(not only mine) and I get a fraction of your subscription fee.</p><p>You might find some of my other articles interesting as well —</p><ul><li><a href="https://medium.com/mlearning-ai/what-is-neural-collapse-de1decf83f48">What is Neural Collapse?</a></li><li><a href="https://medium.com/mlearning-ai/intuitive-explanations-of-why-over-parameterised-deep-nets-dont-overfit-8a323b223ba6">Why over-parameterised deep nets don’t overfit?</a></li><li><a href="https://medium.com/geekculture/lottery-ticket-hypothesis-in-deep-learning-feac2d838fa">Lottery ticket hypothesis in deep learning</a></li><li><a href="https://medium.com/geekculture/how-to-calculate-the-entropy-of-an-entire-language-d17135b01282">How to calculate the entropy of an entire language</a></li><li><a href="https://medium.com/geekculture/the-fundamental-limitations-of-current-ai-systems-87dbdc8ae679">The fundamental limitations of current AI systems exposed via Go</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c2cafdfaa51f" width="1" height="1" alt=""><hr><p><a href="https://levelup.gitconnected.com/first-principles-approach-to-understanding-large-language-models-c2cafdfaa51f">First Principles Approach to Understanding Large Language Models</a> was originally published in <a href="https://levelup.gitconnected.com/">Level Up Coding</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Autoencoders Uncovered: Revealing Intriguing Features Beyond the Basics]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/autoencoders-uncovered-revealing-intriguing-features-beyond-the-basics-bc4121939525?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/2304/0*akI-N5fHr8424TE7" width="2304"></a></p><p class="medium-feed-snippet">How autoencoders make use of information bottlenecks and manifolds to learn useful representations</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/autoencoders-uncovered-revealing-intriguing-features-beyond-the-basics-bc4121939525?source=rss-b88796fee2b6------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/autoencoders-uncovered-revealing-intriguing-features-beyond-the-basics-bc4121939525?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/bc4121939525</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Sun, 04 Jun 2023 19:36:42 GMT</pubDate>
            <atom:updated>2023-06-04T19:36:42.534Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Mysterious Innate Property of Convolutional Neural Networks!]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@sharadjoshi/the-mysterious-innate-property-of-convolutional-neural-networks-a75cb7fef74c?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/2600/0*shNtLoaKe17pgAI4" width="4032"></a></p><p class="medium-feed-snippet">CNNs with random weights are very useful even without training!</p><p class="medium-feed-link"><a href="https://medium.com/@sharadjoshi/the-mysterious-innate-property-of-convolutional-neural-networks-a75cb7fef74c?source=rss-b88796fee2b6------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@sharadjoshi/the-mysterious-innate-property-of-convolutional-neural-networks-a75cb7fef74c?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/a75cb7fef74c</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Sun, 21 May 2023 22:25:48 GMT</pubDate>
            <atom:updated>2024-03-25T11:45:50.102Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Demystifying ImageBind : A Multi-Modal Embedding Approach]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/demystifying-imagebind-a-multi-modal-embedding-approach-4c22ee2e0cf?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/2600/1*7LRKl4TufUySWzyOAhZgsg.png" width="2714"></a></p><p class="medium-feed-snippet">Align modalities without requiring explicitly paired dataset!</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/demystifying-imagebind-a-multi-modal-embedding-approach-4c22ee2e0cf?source=rss-b88796fee2b6------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/demystifying-imagebind-a-multi-modal-embedding-approach-4c22ee2e0cf?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/4c22ee2e0cf</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Sun, 21 May 2023 19:00:29 GMT</pubDate>
            <atom:updated>2023-05-21T19:00:29.287Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Key skills to thrive in the new AI dominated era!]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/geekculture/key-skills-to-thrive-in-the-new-ai-dominated-era-d2131e50ddc9?source=rss-b88796fee2b6------2"><img src="https://cdn-images-1.medium.com/max/2600/0*sJQrv1SX2_0OF8a9" width="3308"></a></p><p class="medium-feed-snippet">A list of core skills to focus on in the new upcoming AI dominated era</p><p class="medium-feed-link"><a href="https://medium.com/geekculture/key-skills-to-thrive-in-the-new-ai-dominated-era-d2131e50ddc9?source=rss-b88796fee2b6------2">Continue reading on Geek Culture »</a></p></div>]]></description>
            <link>https://medium.com/geekculture/key-skills-to-thrive-in-the-new-ai-dominated-era-d2131e50ddc9?source=rss-b88796fee2b6------2</link>
            <guid isPermaLink="false">https://medium.com/p/d2131e50ddc9</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Sharad Joshi]]></dc:creator>
            <pubDate>Thu, 11 May 2023 21:11:37 GMT</pubDate>
            <atom:updated>2023-05-11T21:11:37.832Z</atom:updated>
        </item>
    </channel>
</rss>