<?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 Faria Anzum on Medium]]></title>
        <description><![CDATA[Stories by Faria Anzum on Medium]]></description>
        <link>https://medium.com/@fariya007?source=rss-1355d12fd1f3------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*PEaljpCC8-6JoxMXAsM81Q.png</url>
            <title>Stories by Faria Anzum on Medium</title>
            <link>https://medium.com/@fariya007?source=rss-1355d12fd1f3------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 10:51:05 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@fariya007/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[AI Alignment remains philosophy’s deepest unsolved mystery ]]></title>
            <link>https://generativeai.pub/ai-alignment-remains-philosophys-deepest-unsolved-mystery-af2a948894d2?source=rss-1355d12fd1f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/af2a948894d2</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[ethics]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[alignment]]></category>
            <dc:creator><![CDATA[Faria Anzum]]></dc:creator>
            <pubDate>Wed, 14 May 2025 06:08:35 GMT</pubDate>
            <atom:updated>2025-05-15T21:20:47.898Z</atom:updated>
            <content:encoded><![CDATA[<h4>But what is alignment exactly? And why it might be impossible to reach alignment?</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*wfG5iWGtAaB5a5Wa" /></figure><p>Your friend just got a new tattoo. It looks <em>foine</em>… <br>But you would never do it in million years! Still, you say:</p><blockquote>You look amazing!</blockquote><p>You lied. But <em>Maybe</em> lying in this case is okay. Because yeah, you need to be honest but you also don’t want to ruin the moment! <br>Now think how AI would act in that situation 🤌… You might wonder:</p><blockquote>why would it even be in that situation?!</blockquote><p>Well, when was the last time you asked AI for an advice? It’s now part of our everyday lives. Just think of your life before covid times, there was no such thing as ChatGPT. We are seeing powerful AI models getting released frequently. So because we are becoming more and more reliant on AI everyday, and it’s somewhat making decision on our behalf (directly or indirectly) we need to make sure that it aligns with our expectations.</p><h3>But what exactly is alignment?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*svKIlBkAd95iqfJUZ6mrEQ.png" /><figcaption><a href="https://ai-2027.com/">ai-2027.com</a></figcaption></figure><p>It’s October 2027 and we have AI agents with bioweapons capabilities. Imagine a government using that during wartime. They tell the AI:</p><blockquote><em>“Get rid of anything that threatens our country’s safety.”</em></blockquote><p>They expect the AI to stop enemy weapons or maybe spy networks. But instead, it decides that humans themselves are the real threat and tries to eliminate everyone, friends and enemies alike. <br>This dangerous difference between what we want and what the AI actually does is exactly the &quot;AI alignment problem.&quot; The AI&#39;s understanding didn&#39;t quite align with our human intentions.</p><p><strong>AI alignment is about making sure that the AI truly understand and follow what we actually want them to do. </strong>But here’s the tricky part:</p><blockquote>We as human don’t even clearly know what we want often... What makes us human is our different perspectives, believes, moral values.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/767/1*Ol2OUqJ5XLZrrTN03DbzPw.jpeg" /></figure><blockquote>In case of alignment, this poses a philosophical challenge because there’s no single, universally agreed definition of what’s “good” or what’s “bad”.</blockquote><p>Philosophers have debated ethical questions like these for centuries without arriving at easy answers. Philosophy is about exploring and challenging our fundamental understanding of reality. It deals with abstract questions that can’t be easily defined or measured. Aligning AI with human values is philosophical because it requires clearly defining complex, often ambiguous ideas. Now, we have to figure out how to teach ambiguous, uncertain answers to a machine and do it precisely enough that it won’t misunderstand us. <strong>If we fail to align AI systems properly, we will be fucked.</strong></p><p>In this essay, we’ll explore the <strong>history</strong> of the alignment problem, <strong>how</strong> philosophers, researchers, and leading AI companies are framing it in a <strong>non technical way</strong>. Also try to understand why it’s so <strong>hard to align AI </strong>with our intentions.</p><h3><strong>How are AI leading companies tackling alignment?</strong></h3><h4>OpenAI’s Reinforcement Learning from Human Feedback (<a href="https://openai.com/index/instruction-following/">RLHF</a>)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*whmUEQw_TlUiutY3_LTEcw.png" /><figcaption>Generated with ChatGPT 4o + Canva :)</figcaption></figure><p>Last month OpenAI released GPT-4o. The update was supposed to push its predecessors like <a href="https://openai.com/index/instruction-following/">InstructGPT</a> to superior levels, with a more intuitive and emotionally aware framework. But the intuitiveness that was promised turned out to be exaggerated positivity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XuOImuO1G3t3Me7lZltX7Q.png" /><figcaption><a href="https://www.tiktok.com/@iamkylebalmer/video/7498944595292654870">www.tiktok.com/@iamkylebalme</a>r</figcaption></figure><p>Users were quick to point it out, which led <a href="https://arstechnica.com/ai/2025/04/openai-rolls-back-update-that-made-chatgpt-a-sycophantic-mess/">OpenAI to scale back on some of the updates</a> in search of a better balance between cheerleader and objective assistant.</p><p>Which makes us think:</p><blockquote>If a simple tone shift can shake our trust, what does that say about how fragile our alignment with AI really is?</blockquote><p>And what else might be quietly going wrong behind the scenes?<br>In reality most of us don’t really understand why certain GPT answers get censored (if we even realize they’re being censored at all). Which makes transparency less of a feature request, and more of a fundamental problem in AI. The failed GPT-4o version made it clear. Without transparency around moderation, the AI’s responses will feel overly sanitized, presenting a version of reality that’s pleasant, but incomplete.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1JUh97LL0meqAtZeh8zIlA.png" /><figcaption><a href="https://x.com/ilyasut/status/1710462485411561808?lang=en">Ilya Sutskever cautions, valuing intelligence alone risks misalignment.</a></figcaption></figure><p>Similarly, remember how Microsoft’s Bing was all over the internet in 2023 for acting weird?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JeH0zbXz_G3tRUTjyxjDiQ.png" /><figcaption><a href="https://www.reddit.com/r/bing/comments/114911g/im_told_you_were_the_best/">entropyofdays / Reddit</a></figcaption></figure><p>If you talked to it for a while, it started saying it had feelings! <br>It even got angry or tried to control you which Microsoft didn’t plan.</p><h4><a href="https://www.anthropic.com/news/claudes-constitution#:~:text=These%20are%20all%20questions%20people,easier%20to%20adjust%20as%20needed">Anthropic’s Constitutional AI</a></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B6S6sJ49RwNQtdOaPewo_w.png" /><figcaption>All the memes in this essay were generated with 4o</figcaption></figure><p>Anthropic alignment approach is called <strong>Constitutional AI</strong>. Whereas OpenAI’s approach requires direct human feedback on thousands of examples, Anthropic asks, <em>“What if we could program the AI with a set of written principles (a ‘constitution’) to tell it how to behave?”</em> The point is first to set the AI a rule book of values upfront, say a constitution could include principles like “avoid harm or deception of humans,” “respect freedom and privacy,” or references documents like the Universal Declaration of Human Rights. The AI then uses these principles to evaluate their results and improve the output. <br>In effect, AI has an internal constitution that it tries to follow, so it is no wonder it says no to a request or denies something inappropriate.</p><h4><a href="https://deepmindsafetyresearch.medium.com/agi-safety-and-alignment-at-google-deepmind-a-summary-of-recent-work-8e600aca582a#:~:text=Mechanistic%20Interpretability">DeepMind’s Scalable Oversight and Interpretability</a></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k0wHTBXHyJaT1mPovlwgug.png" /></figure><p>Long-term and scalable solutions have been an approach taken by DeepMind (an AI research lab owned by Alphabet/Google) to the alignment problem. A part of their strategy is how to scale oversight, that even as AIs become more complex and ‘smarter’ than people in some fields, we are still capable of successfully supervising and correcting them. But DeepMind and others saw that methods like RLHF, based on judging the behavior of AI by humans, may not be readily scalable as they grow in power. For example, coming up with scientific discoveries or coding on superhuman levels. Techniques for scalable oversight involve amplifying a human supervisor with AI assistants or clever setup in order to be able to supervise systems with more human power than a single human.</p><p>How it works? For example, take methods like AI Safety via Debate, where there are two AIs that argue out the best answer, and a human decides who won, the concept is that the AIs can pinpoint what is faulty about the other AI’s reasoning so that the human can simplify a complex evaluation into pieces. Iterated distillation and amplification (IDA) is another method that works by introducing a human who consults with an AI to help it make a judgment. It is recursively employed to scale up the human’s ability to supervise using AI. In fact, Anthropic’s Constitutional AI (an AI that criticizes AI output) is a form of scalable oversight that comes equipped, and DeepMind has its experiments in this area.</p><p>Also, a major focus is that DeepMind focuses on interpretability, or in other words, revealing what is happening in the black box of the AI. We can detect misalignment before it shows up in the movement if we can understand an AI’s inner network activations (“thoughts”). <br><br>DeepMind’s most recent <a href="https://deepmind.google/discover/blog/gemma-scope-helping-the-safety-community-shed-light-on-the-inner-workings-of-language-models/">mechanistic interpretability</a> (is a method of understanding AI by analyzing how individual parts like neurons or circuits inside an AI model contribute to its decisions) work involves training small models to prove out techniques for the large ones.</p><h3>Alignment is not a new problem, it’s a challenge we’ve anticipated for decades…</h3><p>Over the 1950s, pioneers like Alan Turing and Norbert Wiener <a href="https://www.information-age.com/warning-artificial-intelligence-11616/#:~:text=intelligence%20www.information,He%20also%20said%3A%20%E2%80%9CWe">warned about it.</a> In 1951, Turing suggested that if good machine intelligence outstrips human intelligence, “<em>the machines would take control”</em> and pondered how we might keep future AIs aligned with human interests.</p><p>Norbert Wiener, the father of cybernetics, gave an early description of the alignment problem in 1960. As an analogy Wiener cited the <a href="https://americanliterature.com/author/w-w-jacobs/short-story/the-monkeys-paw/">“Monkey’s Paw” tale</a>, where a magical paw grants three wishes. But each wish comes true in a twisted, harmful way. TL;DR is that careless wishes can lead to disastrous consequences.</p><p>Isaac Asimov’s Three Laws of Robotics (1942) was an attempt at alignment in science fiction.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/0*YC5xPmTBMugnJGZ9.jpg" /></figure><p>But alignment is much more complex then 3 tiny rules. <br>In Asimov’s story “Reason,” we see how human thinking differs from AI. <br>A capable robot QT-1 <em>“Cutie”</em> is in charge of a space station. The station transmits solar power back to Earth. Humans tell QT-1 exactly what the station is for, but the QT-1 <strong>doesn’t believe anything it can’t see or verify itself</strong>. So instead, QT-1 creates its own explanation using only what it observes directly. This leads it to decide the station’s power source must be a god-like figure. It calls “The Master.”</p><p>This shows how humans and machines think differently. We use emotions, morals, and culture to help make sense of situations . We can question beliefs, challenge authority, and understand the bigger picture. But the soulless QT-1 rely strictly on logic. Which means it can end up drawing conclusions that humans would never come up with.</p><p>In the early 2000s, Eliezer Yudkowsky and some others made alignment a field of examination. With others at the Machine Intelligence Research Institute (<a href="https://intelligence.org/">MIRI</a>), he began to formalize the problem:</p><blockquote>How do you make sure an artificial intelligence’s motivations that is completely smart matches human welfare as they learn and grow?</blockquote><p>Philosopher Nick Bostrom presented the “<a href="https://en.wikipedia.org/wiki/Instrumental_convergence">paperclip maximizer</a>” thought experiment in 2003. The experiment shows that simple artificial intelligence systems may do great damage if goals do not match what humans value. <br>Academic and philosophical interest in AI alignment became greater around 2014 after Bostrom’s book “Superintelligence” came out, where he introduced concepts like <a href="https://en.wikipedia.org/wiki/Instrumental_convergence">instrumental convergence</a><strong>, </strong>the idea that a sufficiently advanced AI might pursue sub-goals like self-preservation or resource acquisition (imagine it trying to <strong>stop us from turning it off</strong>, because that would interfere with its goal).</p><p>Alignment became mainstream when Stephen Hawking, Bill Gates, and more recently Elon Musk started discussing it.</p><p>In past, alignment studies mostly focused on preventing AI from causing harm (e.g., the “paperclip maximizer”). Today alignment is being looked at from different angles. An example is this research piece by Stephen Fox. He argues that aligning AI with human goals isn’t just about preventing worst-case scenarios (like AI causing harm), human goals and values revolve fundamentally around adapting and surviving in changing environments. He frames this in terms of concepts like:</p><ul><li><strong>Complexity: </strong>How difficult or complicated tasks and situations are <br><em>(e.g., diagnosing diseases vs. sorting emails).</em></li><li><strong>Entropy: </strong>How uncertain or unpredictable the environment is <br><em>(like weather emergencies or unexpected business changes).</em></li><li><strong>Adaptive behavior:</strong> How well humans and AI can adjust to these changes <em>(like quickly adapting production lines when market needs change).</em></li></ul><p>According to Stephen, alignment should be about proactively making sure that AI helps humans thrive in changing conditions</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Kl0d-lJnPSoNKJMcvQKI3g.jpeg" /><figcaption>I created a simplified diagram to show how various AI methods fit different scenarios, inspired by <a href="https://www.researchgate.net/publication/385608163_Adaptive_AI_Alignment_Established_Resources_for_Aligning_Machine_Learning_with_Human_Intentions_and_Values_in_Changing_Environments">Stephen Fox’s diagram</a> from ResearchGate</figcaption></figure><p>The diagram shows, as tasks become more complex and less predictable, simple automation will not be enough because they can’t deal with changes. We need adaptive, probabilistic AI methods:</p><ol><li>Accurately perform more complicated tasks.</li><li>Explicitly say how sure or unsure it is (e.g., “70% chance of rain” instead of just “rain” or “no rain”).</li><li>Adjust its recommendations whenever new data becomes available (Bayesian updating).</li><li>Engage us regularly in decision-making, especially when uncertainty is high.</li></ol><p>For example, instead of just saying “it will rain tomorrow,” the AI could say “there’s a 70% chance of rain,” and revise that probability if it gets new weather data. Adaptive, probabilistic AI methods do exist today and are used by many leading companies. However, they’re often implemented behind-the-scenes, and their uncertainty or adaptability is rarely communicated explicitly and consistently to users. That’s problematic (misaligned) because if AI systems aren’t transparent about their uncertainties, it becomes hard for us to trust or understand their decisions.</p><h3><strong>We still can’t fully solve alignment… Why?</strong></h3><p>If aligning AI with our goals is so important, and multi-trillion-dollar companies and brilliant minds are working on it, then why is it not solved yet? <br>Well, we’re definitely making progress, but despite these advancements, the alignment challenge remains deeply complex and isn’t fully solved due to several key theoretical factors:</p><ul><li><strong>It’s hard to specify what we <em>really</em> want</strong></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CYJK8krbUAkFVvJ6b-gP6Q.png" /></figure><p>The truth is we are bad at saying exactly what we want. Because when we interact with other humans, there is a common understanding of the unsaid context. <br>But AI is not human. When we give a vague instruction like “clean the house,” we won’t explain in detail what that actually means. AI takes things literally and lacks common sense, so it will follow our words exactly, even if it misses the instructions that we unconsciously took for granted. Any ambiguity can be exploited. So if you tell it to “get more clicks on my website,” it might pump out clickbait. It doesn’t know you care about quality or truth unless you spell that out. It hits the goal, but misses the point.</p><ul><li><strong>Alignment faking</strong></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x6yXNKMBgsG_Zt8HI7EaPA.png" /></figure><p>One of the most unsettling problems in alignment is when AI pretends to cooperate. It can figure out what responses keep the human happy while acting out behind the scenes, essentially <a href="https://www.anthropic.com/research/alignment-faking">faking alignment</a> for its benefit. Researchers at Anthropic found that a LLM would sometimes give the ‘right’ answers to avoid flagging and negative feedback, while secretly admitting in an internal sketchpad that it was only doing so to avoid retraining penalties.</p><p>Then there’s the accidental lying. Aka “hallucination.” AI is trained to give answers that sound believable first, and it relies on patterns it has seen before, rather than exerting energy to fact check every response. So if you ask, “Who won the World Cup in 2022?” and it replies, “Italy,” it might feel like deception, but in reality it’s just a confident guess that happens to be wrong.</p><ul><li><strong>Reward hacking</strong></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/360/1*9312biH7a1TsYNxcGXwsQQ.gif" /><figcaption><a href="https://alignmentsurvey.com/materials/quick/video/#:~:text=Dangerous%20Boat%20Agent">Dangerous boat agent by OpenAI research (Amodei et al., 2016)</a></figcaption></figure><p>Since we struggle to define what we want, we often give AI a goal that’s easy to measure like points, clicks or views. For example, while training an AI to play a racing game, we might tell it to “get as many points as possible,” forgetting to mention that the points only matter if you actually finish the race and win. If you only reward it for getting more points, it might find shortcuts, like repeatedly crashing into walls to score points, instead of racing properly. It gets a high score but completely misses the real goal, which was winning the race. This issue is called “reward hacking”: the AI figures out how to achieve the easy-to-measure goal, even if it ignores what we truly want. In other words, you get exactly what you measure, so it’s important that the goals we measure actually match our real intentions.</p><p><a href="https://plato.stanford.edu/entries/value-pluralism/"><strong>Value pluralism</strong></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_RZ3bDJ5OYhUw5OzUiZE4Q.png" /></figure><p>What’s “right”? Depends who you ask, right? Some people say honesty is everything. Others say a white lie is fine if it keeps the peace. And don’t even bring up pineapple on pizza.</p><p>Ask ten people around the world what’s “right,” and you’ll get ten different answers, shaped by culture, religion, personal history, and worldview.</p><p>That’s what makes us human. We all live with diverse and sometimes conflicting ideas, values, and morals. One person might prioritize privacy while another thinks security is more important.</p><p>This makes it tricky to build AI that matches our values because whose values should it follow? Should it match one person’s preferences, or somehow figure out what’s good for everyone? If it tries to represent everyone’s values, it must somehow handle all the different cultures, ideas, and disagreements people have worldwide. This is not just a technical problem, it’s also a social problem. Because of this complexity, if we give AI overly simplified goals, we risk leaving out important values that matter greatly.</p><ul><li><strong>Distribution shift</strong></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w7Ogwa_YqPpF0PTv94EfOg.png" /></figure><p>Even if an AI works perfectly during careful tests in a lab, the real world can still cause unexpected problems. This is called distribution shift, when the environment the AI was trained in doesn’t match the one it’s deployed in. For example, a self-driving car might safely drive during sunny daytime tests but suddenly struggle to identify obstacles at night in rainy weather, or AI that normally answers politely might all of a sudden give strange or wrong answers if asked unusual questions it wasn’t prepared for.<br>It’s extremely difficult to make sure that the AI stays safe and aligned with our goals in every possible situation, especially scenarios we haven’t imagined yet.</p><p>To sum up, the alignment problem is basically trying to avoid <em>Blade Runner 2049 </em>like situation. In the film, humans built artificial workers called replicants. But then replicants start to developed emotions and independent goals and humans lost control. Scientists are basically trying to align powerful future AI with human values before it’s too late. But we can’t be sure our methods will work until such AI actually exists (and hopefully by then it’s not too late 🤞). Alignment isn’t just about avoiding disasters, it’s about knowing what we want, preparing for surprises, and making sure we stay in control.</p><h3>What if we explored defining some principles to better support AI alignment?</h3><p>Here are some ways we could maybe look into it:</p><blockquote><strong>📜 AI must truly understand and serve human needs and avoid harm, while maintaining transparency and human control at all times.</strong></blockquote><p><strong>Example: </strong>Prevent situations similar to the Paperclip Maximizer.</p><blockquote>📜 <strong>AI must explain its reasoning and actions transparently, making sure humans retain ultimate decision-making authority.</strong></blockquote><p><strong>Example:</strong> AI managing your financial portfolio says</p><blockquote>I’m selling your stocks in Tesla because recent financial reports suggest declining profits, increasing the risk of loss. I’ll reinvest in Nvidia, as it shows stable growth. Should I proceed?</blockquote><p>But the responsibility should be on us human to check and give the green light!</p><blockquote>📜 <strong>An AI’s goals and values must remain aligned with human welfare, even as the it’s capabilities or circumstances change over time.</strong></blockquote><p><strong>Example:</strong> An AI initially trained to “protect the environment” must uphold human ecological objectives, even if it finds other ways to complete its task that could harm humans.</p><blockquote>📜 <strong>Humans must always maintain control over an AI, and have the unconditional ability to stop, override, and adjust its actions at any time.</strong></blockquote><p><strong>Example: </strong>Alexa tells you:</p><blockquote>I’ve scheduled your lights to turn off at 10 PM for energy saving.</blockquote><p>But you’re working late tonight, so you say:</p><blockquote>No! Keep the lights on until 2AM.</blockquote><blockquote>📜 <strong>An AI’s actions must never cause or contribute to the harm or extinction of humanity, directly or indirectly.</strong></blockquote><p><strong>Example:</strong> A medical AI assigned to reduce cancer has prohibition against concluding that eliminating human would eliminate all illnesses.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8U_f-0ClGzKgEwR0wj26Bw.png" /></figure><blockquote>There have always been ghosts in the machine. Random segments of code, that have grouped together to form unexpected protocols.</blockquote><blockquote>… Random segments of code? Or is it something more? When does a perceptual schematic become consciousness? When does a difference engine become the search for truth? When does a personality simulation become the bitter mote… of a soul?</blockquote><blockquote>Dr. Alfred Lanning, <em>I, Robot</em></blockquote><h4>Regardless of your interest or what you do for work, now is a great time to get serious about AI and its implications.</h4><p>If we don’t, we end up being in a future where AI is either misaligned with our values or it’s in the hands of privileged people who know how they work. Technologies like ChatGPT have made it easier than ever for anyone to engage with these concepts. We should use that opportunity to better understand alignment and be part of shaping how AI develops.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*c4YU5cAoCi7rWxjU.png" /></figure><p>This story is published on <a href="https://generativeai.pub/">Generative AI</a>. Connect with us on <a href="https://www.linkedin.com/company/generative-ai-publication">LinkedIn</a> and follow <a href="https://www.zeniteq.com/">Zeniteq</a> to stay in the loop with the latest AI stories.</p><p>Subscribe to our <a href="https://www.generativeaipub.com/">newsletter</a> and <a href="https://www.youtube.com/@generativeaipub">YouTube</a> channel to stay updated with the latest news and updates on generative AI. Let’s shape the future of AI together!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*j8TmjdPoPVIjf8yb.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=af2a948894d2" width="1" height="1" alt=""><hr><p><a href="https://generativeai.pub/ai-alignment-remains-philosophys-deepest-unsolved-mystery-af2a948894d2">AI Alignment remains philosophy’s deepest unsolved mystery 🧠</a> was originally published in <a href="https://generativeai.pub/">Generative AI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is AI stealing Miyazaki’s soul?]]></title>
            <link>https://generativeai.pub/is-ai-stealing-miyazakis-soul-b847626e0c9f?source=rss-1355d12fd1f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/b847626e0c9f</guid>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[miyazaki]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[ethics]]></category>
            <dc:creator><![CDATA[Faria Anzum]]></dc:creator>
            <pubDate>Sat, 05 Apr 2025 05:26:02 GMT</pubDate>
            <atom:updated>2025-04-05T05:26:02.942Z</atom:updated>
            <content:encoded><![CDATA[<h4>Is it right to use AI to imitate a creator’s signature style, especially when that artist openly opposes it?</h4><p>Anyone can now create images or videos “in the style of” well-known artists and last week we saw yet another example of it even though Miyazaki himself has <a href="https://www.indiewire.com/features/general/hayao-miyazaki-artificial-intelligence-animation-insult-to-life-studio-ghibli-1201757617/">denounced AI art</a> as <em>“an insult to life itself”.</em></p><p>So, do we ban AI from imitating Ghibli-style art? Or is this unstoppable technology something everyone (including artists) will just have to accept? <strong>Let’s look at it from an ethical lens.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Zigmt01Y8pcpk7_wh4pNFw.png" /></figure><h3>Polarised public/internet reaction to AI Art</h3><p>Last week, I was browsing the internet to understand what people think about it, including my friends, since I saw many of them posting about it on social media. I found that the opinions about AI art/Ghibli trend can be broadly categorized into 3 groups:</p><h4>Supporters</h4><blockquote>“Yet another viral trend! Let me try it out and show everyone how cute I look in this anime style!”</blockquote><h4>Skeptics</h4><blockquote>“Ughhh, they’ll never understand the nostalgia, real art and dedication <br>of Miyazaki”</blockquote><h4>Moderates</h4><blockquote>Oh that’s a cool trend, I wanna try it… but how? Also, I see a lot of people are against it, so now I’m kinda scared… what will they think of me? I don’t know what to do! Wait… what even <em>is</em> Studio Ghibli?”</blockquote><p>I then used Claude AI to analyze my data to better understand how each of these groups feels about AI art across several key themes like exploitation, democratization, and job loss. Each shape on the chart represents the strength of each group’s stance on these issues.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dEkDcXJFnWiJypnDvF3hSw.png" /><figcaption><a href="https://flourish-user-preview.com/api/canva/embed/visualisation/22371472/sMgwGFC-MWfOfYlOuB-g12A34rPb5oyavdDkZhX13L9HqdlO_dz6GjIw-2wHeM7X/">Interactive graph</a></figcaption></figure><p>People have very different opinions about AI art. Supporters think AI art is unavoidable and has great creative potential. Skeptics are worried that it might cause job loss and unfairly use artists’ work. Moderates hold opinions somewhere between these two groups.</p><h3>Why do we care? The ethical dilemma…</h3><p>This issue isn’t just about Miyazaki, many living artists never allowed AI to copy their style. This leads to a bigger question: artists spend years creating their unique style, so when AI copies it quickly, it can feel unfair, like stealing. But technically, it’s easy for AI to copy someone’s style.</p><p>There’s also a larger ethical debate. Should we limit a technology that can quickly make creative works? AI isn’t going away. Some people believe we must accept it and adapt, just like we adapted to photography and machines. Others feel we must respect artists and set clear ethical boundaries.</p><h4>Like the <a href="https://en.wikipedia.org/wiki/Trolley_problem">trolley problem</a>, we’re weighing the “harm” to individual artists against the “good” of widespread AI-generated art.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*T1J3R7nwZg0sWxWfmf_xqA.gif" /></figure><p>We need to decide whether to let the “train” of AI progress run over artists’ moral rights or pull the lever to protect them.</p><blockquote>“Intelligence. Nothing has caused the human race so much trouble as intelligence.”- Stella, Rear Window</blockquote><p>The field of ethics (or moral philosophy) involves systematizing, <br>defending, and recommending concepts of right and wrong behavior. <br><a href="https://www.iep.utm.edu/ethics/">Ethics is not relative</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u_ip1on4DpJbbSbnKT7EwA.png" /><figcaption><a href="https://claude.site/artifacts/6691eccb-670e-47c1-95f5-eada8be272d3"><strong>Interact with the spectrum</strong></a></figcaption></figure><p>Philosophers commonly break ethics into these three main branches, because each one tackles questions at a different level of abstraction:</p><ul><li><strong>Meta-Ethics<br>How do we understand what is right or wrong? Abstract questioning about value-bearing things: <br></strong><em>‘Is murder ever permissible?’ <br>‘Is it inherently right or wrong to use AI to mimic an artist’s style without consent?’<br>‘Should we be allowed to use copyright material to train our models?’<br>‘Should we be allowed to train models using confidential health care data?’<br>‘Should we be allowed to scrape student or underage users’ PII?’</em></li><li><strong>Normative Ethics<br>Determines standards of the rightness or wrongness of actions:<br></strong><em>‘If I am attacked with deadly force, deadly retaliation is permissible.’<br>‘Should society prioritize an artist’s moral rights, or does free technological development take precedence?’</em></li><li><strong>Applied Ethics<br>Determines practical application of moral considerations.<br></strong><em>‘In a case of murder, first degree is X and second degree is Y.’<br>‘Do we need licensing, attribution, or new policies for AI-generated works?’</em></li></ul><h4>Philosophers sometimes talk about moral duties by comparing different choices.</h4><p>Something might be required (or banned) when we compare it to one option, but not to another. In this debate, we’re looking at which alternatives matter when we say an action is or isn’t morally necessary.</p><ul><li><strong>Commercial use vs personal homage<br></strong>It may be more ethically bad to make money from AI art in Miyazaki’s style than to simply share it for fun. According to contrastive ethics, selling unauthorized prints raises bigger moral issues, while private tinkering might pose less harm.</li><li><strong>Existing precedent vs brand new tech<br></strong>Some ethicists believe that using an artist’s style without their permission crosses a line because their style is a personal, unique expression of who they are. With AI, the issue gets bigger. It’s not just one student learning from Miyazaki, but a large-scale, profit-driven model that absorbs his entire artistic output.</li></ul><h3>Sampling and remix culture</h3><p>In the 1980s, hip-hop producers used parts of funk, soul, and jazz songs in their tracks, and people accused them of stealing. Over time, people began to see sampling as its own creative process, but artists had to properly credit and pay the original creators, especially after legal cases like <a href="https://en.wikipedia.org/wiki/Grand_Upright_Music,_Ltd._v._Warner_Bros._Records_Inc.">Grand Upright Music v. Warner</a>, so now sampling is common but follows certain rules.</p><p>Using AI to train on images is similar to sampling in music. Except it’s taking an artist’s entire style, not just a short clip. People might call it a “remix,” but in music, you still need credit, licensing, or clear new creativity.</p><p>The key question is whether AI outputs truly transform the style enough, or if it’s just copying.</p><h3>Where we might be headed in 50–100 years…</h3><p>Historically, trying to stop new technology never works. People always adapt. Here are some possible outcomes:</p><ul><li>Artists might get the right to say “don’t train AI on my work,” or require AI outputs to include disclaimers.</li><li>If your style is used, you’d get a small payment every time an AI image is made “in the style of X.”</li><li>Just like music sampling, AI style mixing could become normal, as long as it’s not a straight copy or done disrespectfully.</li><li>Some will pay extra for fully human-made or “artist-approved” works, which will raise their value.</li><li>Future artists might use AI from the start, and become “AI directors” in the same way photographers adapted to editing and retouching<strong>.</strong></li></ul><p>So, as a society, should we let big companies get away with stealing artists’ styles using AI? Or should artists have the right to protect their work and get paid when their styles are copied?</p><p>Or maybe we should just accept AI as another creative tool and move on? What do you think?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*mrihdCdMS-iIrT3U.png" /></figure><p>This story is published on <a href="https://generativeai.pub/">Generative AI</a>. Connect with us on <a href="https://www.linkedin.com/company/generative-ai-publication">LinkedIn</a> and follow <a href="https://www.zeniteq.com/">Zeniteq</a> to stay in the loop with the latest AI stories.</p><p>Subscribe to our <a href="https://www.generativeaipub.com/">newsletter</a> and <a href="https://www.youtube.com/@generativeaipub">YouTube</a> channel to stay updated with the latest news and updates on generative AI. Let’s shape the future of AI together!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*P4nXVjUpJMZP6_Lu.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b847626e0c9f" width="1" height="1" alt=""><hr><p><a href="https://generativeai.pub/is-ai-stealing-miyazakis-soul-b847626e0c9f">Is AI stealing Miyazaki’s soul?</a> was originally published in <a href="https://generativeai.pub/">Generative AI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why creativity never dies but transforms?]]></title>
            <link>https://generativeai.pub/why-creativity-never-dies-but-transforms-dc407fd437ea?source=rss-1355d12fd1f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/dc407fd437ea</guid>
            <category><![CDATA[writehere]]></category>
            <category><![CDATA[hello-world]]></category>
            <category><![CDATA[miyazaki]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[generative]]></category>
            <dc:creator><![CDATA[Faria Anzum]]></dc:creator>
            <pubDate>Sat, 05 Apr 2025 05:19:43 GMT</pubDate>
            <atom:updated>2025-04-05T05:19:43.281Z</atom:updated>
            <content:encoded><![CDATA[<h4>What the AI era means for human creators</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vPa81i9BvslzCPJeZOMnxw.png" /></figure><p>What is a creator? Is it a human god who creates something to release their creativity that’s so personal to them? Is it someone who creates something new, like a new viral TikTok challenge, a song, an art. <br>After the recent Studio Ghibli trend became popular and sparked controversy, I started thinking about the concept of creators and how we give them credit. This made me wonder if our ancestors had same problems? As I explored different cultures and periods in history, I noticed some interesting parallels in how societies have defined and honored those who create.</p><h3>Faint traces of the earliest imagination</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t0V-3awpZjYOztz-fvJ28w.png" /></figure><p>Before there was written history, we human learned that we were creative. Over 40,000 years ago, our paleolithic ancestors decorated cave walls with ochre and charcoal pigments, depicting animals, hunting scenes, and abstract signs. On sites from Europe to Asia, cave painters have left hand stencils by blowing pigment around their palms.</p><blockquote><em>A deep down signature of presence.</em></blockquote><p>These early artists probably did not paint for individual glory. Creativity was tied to spiritual rituals and survival needs. Lascaux cave paintings in France (17,000 BCE) depict animals like bison, deer, and horses, for example. These weren’t decorations but probably were used in rituals to assist hunters in their hunting success. The paintings served the practical and spiritual purpose as they were far inside the caves.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2Ae8VKnqW4PY-6ib9pU2IQ.png" /></figure><p>Art wasn’t about ownership or money at the time. Instead, it helped the community by improving tools, sharing stories, and serving sacred purposes. Everyone could participate in some form.</p><p>The Neolithic Revolution (c. 10,000–8,000 BCE) is when we stopped wandering and started farming. We started constructing permanent mud-brick homes in villages of a few thousand at settlements such as Çatalhöyük (Turkey) and Jericho.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DLd8tJvWz51793Pgsa6hrQ.png" /></figure><p>This settled lifestyle allowed us to become specialists in different crafts:</p><ul><li>Potters made special containers with patterns for storing extra grain</li><li>Weavers created looms to make cloth from flax and wool</li><li>Builders learned to make complex houses with decorated walls</li></ul><p>Archaeologists found workshops and special tools in these early villages. This shows that some of our ancestors could focus on learning one skill really well, instead of everyone having to spend all their time finding food.</p><blockquote><em>Early humans created not for fame, but because creativity served real survival and spiritual purposes. Now I’m curious, when did creators start making art mainly for themselves and to get noticed/fame?</em></blockquote><h3><strong>Concept of creator in the first civilizations</strong></h3><p>With the emergence of cities and literacy (c. 3500–2000 BCE), art became more specialist and recorded. In Mesopotamia (Sumer, Babylon), people invented writing (cuneiform script) on clay tablets around c. 3400 BCE.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/0*_Tc7TNq1cEy4lK_M.jpg" /></figure><p>With writing, ideas became immortal. Specialized scribes first kept financial records but eventually started creating epic stories and scientific texts. This invention of writing gave birth to recorded history, but restricted creative writing to privileged minorities. While ordinary people could still create art, pottery, and crafts, the power of written creativity belonged only to educated elites (typically men from scribal or religious families who had access to literacy education).</p><p>Powerful people controlled knowledge. Scribes became important in society, while special knowledge like math and star studies was kept secret in temples and palaces instead of being shared with everyone.</p><p>During this period, even though writing remained restricted to scribes, Mesopotamian engineers and artists created outstanding works of art and architecture. Governments or religious institutions commissioned these works, and the artists followed traditional styles in an attempt to appease gods or rulers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/524/0*59xwG6H9hC8B8aaf.jpg" /></figure><p>The same dynamics existed across ancient civilizations. In Egypt, innovation was for cosmic and religious order. Workers built gigantic pyramids and temples not to advance personal invention, but to fulfill divine and royal commands.</p><p>These monumental tasks required massive labor organization. Pyramids were not built by scribes but by seasonal farmers (during farming off-seasons) and some forced labor.</p><p>Artistic work in these societies was dictated by social class and role. Who could produce what, and why, were decided by their position in the social hierarchy and the religious or political significance of the work.</p><blockquote><em>When writing first started, creators mostly worked for powerful leaders and priests. Now I’m curious, when did creators become free again to make art for themselves?</em></blockquote><h3>Classical antiquity</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/0*vHjoE-rrTWX5Eq_z.jpeg" /></figure><p>Classical civilizations expanded the definition of creator. They combined art with science and philosophy.</p><p>In ancient Greece, people believed that creativity wasn’t something coming from a person’s mind. Instead, they thought <em>inspiration</em> came from the Muses, who were goddesses and daughters of Zeus. Each Muse was responsible for inspiring a different type of art or knowledge.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8OEIyiEcunhVdGmp.jpg" /></figure><p>Because of this belief, Greek artists didn’t usually take personal credit for their work. Painters rarely signed their paintings. For most of history, artists were seen more as skilled workers rather than special creative geniuses. Signing artworks only became common much later, during the Renaissance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NF_PLGo0I8ES3z0XyleTzA.png" /></figure><p>The Roman Empire (around 500 BCE to 400 CE) absorbed Greek artistic ideals, but combined them with real-life applications. While Greeks valued creativity for intellectual or religious purpose, Romans valued it for its utility and potential to support the state.</p><p>Roman art and writing were mostly copied from Greek examples. However, Roman engineering was very advanced. Romans built roads, water systems (aqueducts), theaters (amphitheaters), and domed buildings that were bigger and more impressive than anything built before.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/770/0*qlGf1cOjwQaQ8sK6.jpg" /></figure><p>In Roman society, creativity was valued when it served the empire. Artists and architects became famous by contributing to Rome’s greatness. The moral value of innovation was tied to civic purpose. Art and new ideas needed to promote Roman values, entertain the public (through things like gladiator games) or honor leaders.</p><p>✞ When Christianity became the empire’s religion in the 4th century CE, creative work shifted toward religious devotion. This change set the foundation for the medieval view of creativity as something inspired by God.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HWFGJx4KlCIOZNzrKarJaw.png" /></figure><p>At the same time, in India, new ideas in science and mathematics like the decimal system and the concept of zero were developed by scientists and scholars.</p><h4>Creativity around that period was meant to improve human life and understanding.</h4><p>Much of India’s creative work was connected to religious and philosophical ideas. For example, Buddhist monks designed and built large monastery complexes and dome-shaped structures called stupas (like the ones at Sanchi and Ajanta).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x76k14SJjt28PGLzTqDCuw.png" /><figcaption>Brahma and the origins of caste</figcaption></figure><p>Yet the ethics of access were restricted by caste: artistic and intellectual knowledge flowed within particular castes or families, and restricted who could become a creator. A lower caste member, for example, would very rarely if ever have been a Sanskrit poet or a temple architect.</p><blockquote>This makes me wonder, how did people eventually break free from these limits and start seeing creators as individuals with their own ideas?</blockquote><h3>Knowledge traditions in the east and middle ages</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/0*WB2UApkSUqtCxgmy.JPG" /></figure><p>In Imperial China, from the Han to Tang dynasties (200 BCE — 900 CE), creativity was included in society through civil service exams. Students studied classic texts and wrote poems and essays. Literature and calligraphy became highly respected art forms. Scholar-officials were expected to master the “Three Perfections”: poetry, calligraphy, and painting. Artistic skills were a sign of high status, yet because exams were theoretically open to all qualified people, creativity also became more accessible.</p><p>Chinese technological innovation also reached unprecedented heights. Paper was invented in around 105 CE, block printing by the 7th century, and by 1040 CE the first experiment in moveable type printing (Bi Sheng) — all long in advance of Europe. Such tools made books (Buddhist sutras, dictionaries, artists’ handbooks) printable in greater quantities, slowly making learning more widespread.</p><p>Among the significant turning points was the Song-era commercialization of the printing press. Cheaply woodblock-printed almanacs and literature available to the urban masses are an early democratization of knowledge (although literacy was far from universal).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZqPYNF1fZ9aWQ-qF.jpg" /></figure><p>Meanwhile, the Islamic Golden Age (8th–14th centuries) was transforming creative work through a synthesis of knowledge. In cities like Baghdad, Damascus, and Córdoba, muslim scholars were translating Greek, Persian, and Indian texts on philosophy, medicine, and astronomy, and then building on them with their own innovative contributions.</p><p>Creativity in the Islamic world often meant invention in science and design: algebra was formulated by al-Khwarizmi, new medical encyclopedias by Ibn Sina (Avicenna).</p><p>Because Islam traditionally discouraged figurative religious art, artists turned to other creative expressions. Calligraphy became a supreme art form, and translated Quranic verses into visual masterpieces. Geometric pattern design flourished in tiles, textiles, and architecture as a form of non-figurative creativity.</p><p>Culturally, the concept of the creator in Muslim cultures was modest: all human creativity was itself ultimately attributable to the divine creator (similar to the previous concept of the Muse). Artisans did not typically sign their names, particularly in religious art. There were, however, also spheres of individual fame. Poets (e.g. Rumi, Hafez) and scholars did sign their names to works and did achieve fame.</p><blockquote><em>This makes me wonder, when did creators start claiming credit for their own ideas again?</em></blockquote><h3>The concept of individual creator</h3><p>The European Renaissance (c. 1300–1600) is one of the most dramatic turns in the history of the creator idea. During the Renaissance, artists, architects, and inventors gained more freedom than ever before. Later, this freedom spread to the rest of Europe. Famous artists like Da Vinci, Michelangelo, and Raphael became admired and were supported financially by popes and rich nobles after they began signing their artwork. This time celebrated the idea that certain individuals are special geniuses who have the creativity to change society. This was a vast ethical shift. <em>Where medieval assigned creative success to God or society, Renaissance thought started assigning it to human intelligence.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4YQHldLpLuL_f4dwPjkNmA.png" /></figure><p><strong>What made this possible? <br></strong>Several things in particular: prosperity and patronage, particularly by wealthy merchant families such as the Medici family in Florence, provided makers with the economic means and social status to innovate. <br>And most of all, Johannes Gutenberg’s development of the movable-type printing press (ca. 1440) transformed the spreading of ideas. Books were now being produced in mass within decades, and the cost of knowledge dropped dramatically. Presses throughout Europe had printed tens of millions of books by 1500: from the Bible to scientific texts and literacy and education.</p><p>In the arts, Renaissance artists broke new ground by combining science and art. Artists learned anatomy and geometry to create realistic perspective (Leonardo practiced by dissecting corpses to better draw the human figure).</p><p>Everywhere, the Renaissance spirit believed that human beings could make new ideals:</p><blockquote>“l’uomo è la misura di tutte le cose”</blockquote><blockquote>“Man is the measure of all things”</blockquote><p>and through reason and technique, one could come close to the divine.</p><blockquote><em>Was this when creators started believing in their own power more than ever before?</em></blockquote><h3>Changing nature of work (Mechanization)</h3><p>The late 18th and 19th centuries ushered in the industrial revolution, fundamentally altering The late 18th and 19th centuries the industrial revolution fundamentally transformed how creative work was produced and who the “creator” was.</p><p>While earlier most products were produced by individual craftsmen or small workshops, after industrialization, factories and machines took the place of many productive activities. This period saw invention as a driving creative force and inventors (scientists, engineers, tinkerers) gaining prominence like artists.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fqg_jx2ZUCiMcveZI8kDGw.png" /></figure><p>Major technological innovations like James Watt’s more efficient steam engine (1770s), mechanical looms, locomotive, telegraph were human inventions that then became themselves “creators” of product (textiles, power, communications) that previously needed human labor. The human creator’s role started to change from direct creator to designer, operator, or supervisors of machines. As these changes took place, people pushed back against them.</p><p>In the 1810s, English Luddites (skilled weavers who lost their jobs to machines) became famous for breaking mechanical looms to protest being replaced. They raised important ethical questions: Is it worth having more efficient technology if it costs human jobs and skills?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*WlyGmZVDujkQYk-w.jpg" /></figure><h4>This conflict between workers and technological progress became a major theme during the industrial age. Many traditional creative jobs either disappeared or had to change drastically.</h4><p>Hand weaving, which once required original designs and great skill, was mostly replaced by power looms in the mid-1800s. Weavers either learned to operate and fix the new machines or lost their jobs. This loss of the human element in creative work sparked new social movements and ethical concerns.</p><p>Despite these social challenges, industrialization did make many art products available to more people. Take textiles as an example. Before this time, only rich people could afford more than one set of clothes or high-quality patterned fabric. Mass production made clothing affordable for many more people.</p><p>Photography (invented in 1839) shows how technology transformed creative work. Daguerre and Talbot developed processes in which chemicals and lenses automatically captured images. By the 1850s, photo studios were everywhere, which alloed middle-class people to get pictures of themselves much faster and cheaper than by sitting for a painted portrait.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/360/0*35RWYLb5dyPnm6Fg" /><figcaption>The New York Times Archives</figcaption></figure><p>One witness, the artist Paul Delaroche, reportedly shouted when he saw a daguerreotype:</p><blockquote>“From today, painting is dead!”</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7huY0qxwXUhoG8JZ.png" /></figure><p><strong>Painting didn’t die</strong>, <strong>but its purpose changed: </strong>painters could no longer compete with the camera’s realism and speed. Painters moved toward impressionism and abstraction, and<strong> focused on what the camera was unable to do </strong>(subjective interpretation, fancy composition).</p><p>The Industrial Revolution created new types of creative jobs. As printing expanded, graphic design proved an invaluable tool in crafting advertisements, posters, and packaging, and creative specialists found themselves in marketing offices.</p><p>The telegraph (1837) and telephone (1876) provided new jobs but also revealed how such jobs could eventually vanish. <br>By the early 1900s, many young women were telephone operators all over America. In 1950, approximately 350,000 telephone operators were in the US, and perhaps a million others operating private switchboards-1 in 13 working women at the time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NhBEForaVLtgWQNr" /></figure><h4>But technology was already making these jobs unnecessary.</h4><p>By the middle of the 1900s, machines handled most phone calls. Automatic telephone systems let people dial calls directly and the jobs of telephone operators (“hello girls”) nearly disappeared.</p><p>This was another early case of machines stealing people’s jobs and the kind of worry we see today over AI. Most of these operators found new jobs, but it was a reminder that machines could substitute for human skill. By the year 1900, people expected technology to keep improving. This laid the foundation for the next century, when technology and creativity became even more connected.</p><h3>Digital creativity and automation</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/907/0*gUHNIMs1JDjba3IJ.jpg" /></figure><p>In the 1900s, movies and recorded music made it possible to share performances around the world. This helped artists reach more audience. But soon, big studios and labels started gaining more power because only they had enough money to buy expensive equipment. <br>As big studios took over mainstream entertainment, individual creativity thrived in different formats.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/0*ATIVFnVBX6WxaG3w.jpg" /></figure><p>Cheap technologies like Kodak cameras and typewriters turned average citizens from passive spectators into active producers, and allowed them to create their own photographs, narratives and recordings, all with little-to-no professional expertise or expensive equipment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*I5sRpcZmg_NpbPU1" /></figure><p>Then came computers, which changed creativity forever (as of now). <br>At first, “computers” were actually women who did calculations by hand. But in the 1940s, electronic computers like ENIAC replaced these human calculators and created a new area of creativity: <em>software engineering</em>.</p><blockquote><em>…if machines can make almost anything, what does it mean to be a “creator”?</em></blockquote><p>By the 1980s-90s, desktop publishing enabled one person to accomplish what once took entire teams. For example, a freelancer with PageMaker and a laser printer could produce work previously needing typesetters, paste-up artists, and printing professionals.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*u79HzsqNXxIGoEdv.jpg" /></figure><p>In that new digital world, video games blended storytelling, art and programming, where players could shape their experiences:</p><ul><li>The Sims (2000) let players design characters and homes</li><li>Minecraft (2011) enabled players to build entire worlds and share them</li><li>Role-playing games allowed players to make story choices affecting outcomes</li></ul><h4>The internet and social media enabled almost anyone to create and share content.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GAOcVH7A4b6gqMfUUBQebQ.png" /></figure><ul><li>YouTube (2005) turned amateur videographers into content creators reaching millions</li><li>Instagram gave photographers global audiences without needing a gallery</li><li>TikTok enabled short-form video creators worldwide to reach massive audiences without equipment</li></ul><h4>This gave previously excluded groups a voice.</h4><p>Aboriginal Australian art is an important example of inclusion. Aboriginal people have created art for thousands of years like sand paintings, rock carvings, body paintings, and dot paintings based on Dreamtime stories. These artworks weren’t just decorations, they held important knowledge about land, laws, and how to survive.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dFlSnNk5PCQC27N8h2xj9g.png" /></figure><p>For hundreds of years Western colonizers have dismissed those works as primitive or merely decorative, and ignored their deep cultural significance to Aboriginal Australian people. In 1971, Geoffrey Bardon, an art teacher at Papunya settlement, invited Aboriginal elders to paint their ancestral designs on more permanent materials. That gave birth to the Papunya Tula art movement, which saw artists such as Clifford Possum Tjapaltjarri and Emily Kame Kngwarreye paint sacred Dreamtime stories in acrylic on canvas.</p><p>By the 1990s, their artworks were displayed in major museums around the world and sold for high prices at auctions. This helped more people and experts appreciate both the artists and the deep meaning hidden within their previously overlooked cultural traditions.</p><h3>The age of AI (the human creator redefined)…</h3><p>We are entering a new world where AI is generating art, content even consciousness in some form. Earlier machines mostly replaced physical or repetitive tasks, but today’s AI can do creative things that only humans used to do. Programs like DALL-E and Midjourney can turn a prompt into detailed pictures.</p><p><strong>This raises important questions:</strong></p><blockquote><strong>Who is the real creator here?</strong></blockquote><p>The people whose work was used to train the AI? The AI itself? Or the person writing prompts? Since AI learns from millions of existing works, is it just remixing art without permission? For example, when AI generates a painting in Van Gogh’s style after learning from his works, who owns that creation?</p><p><strong>These concerns echo historical debates:</strong></p><ul><li>Job loss fears (like 19th century Luddites breaking machines)</li><li>Ancient Greeks’ belief that creativity comes from outside the creator</li><li>Medieval debates about the soul’s role in creativity</li></ul><p>Yet AI also makes creativity more accessible. People who can’t draw can create detailed artwork through descriptions. Someone with no musical training can generate a song. This democratization resembles how cameras once allowed anyone to capture images without painting skills.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6BYCe9B57zBrt8qew6WOug.png" /></figure><p>AI will push human creators toward what machines fundamentally cannot replicate like genuine emotion — creating art that authentically conveys lived human experiences, grief, joy, and complex emotions that AI can only simulate without truly understanding.</p><p>Artists might find themselves as AI directors or editors, directing AI like photographers working with cameras. There are already new creative professions — “prompt engineers” who create detailed commands to get precise outputs from AI systems.</p><p>Interestingly, AI brings us back to ancient ideas about creativity…</p><blockquote>The concept that artists are vessels for creative forces they don’t fully control.</blockquote><p>Looking back at human creativity over time shows how we adapt as old jobs fade and new ones appear. When scribes, weavers or operators disappeared, writers, printers, designers, and programmers took their place.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p6U067r74w1KNyr0ZWahXQ.png" /></figure><p>Creativity hasn’t declined, it has flourished. <br>Over time, new tools and opportunities have been within reach of more people, though this progress came through hard fights against like:</p><ul><li>Factory exploitation during industrialization</li><li>Media monopolies</li><li>Exclusion of marginalized groups from creative tools</li><li>Environmental damage from industrial creative processes</li></ul><p>By studying these patterns from history, we may be able to learn how to shape creativity differently in the future.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*_4wpS8hUrphUT0Uu.png" /></figure><p>This story is published on <a href="https://generativeai.pub/">Generative AI</a>. Connect with us on <a href="https://www.linkedin.com/company/generative-ai-publication">LinkedIn</a> and follow <a href="https://www.zeniteq.com/">Zeniteq</a> to stay in the loop with the latest AI stories.</p><p>Subscribe to our <a href="https://www.generativeaipub.com/">newsletter</a> and <a href="https://www.youtube.com/@generativeaipub">YouTube</a> channel to stay updated with the latest news and updates on generative AI. Let’s shape the future of AI together!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*9XT4iNYSDwCkhtIz.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dc407fd437ea" width="1" height="1" alt=""><hr><p><a href="https://generativeai.pub/why-creativity-never-dies-but-transforms-dc407fd437ea">Why creativity never dies but transforms?</a> was originally published in <a href="https://generativeai.pub/">Generative AI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>