<?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 Howard Chen on Medium]]></title>
        <description><![CDATA[Stories by Howard Chen on Medium]]></description>
        <link>https://medium.com/@howardchen_design?source=rss-4cbf6ab905be------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*mAr0nJS7I0ZC_qH6V44HKg@2x.jpeg</url>
            <title>Stories by Howard Chen on Medium</title>
            <link>https://medium.com/@howardchen_design?source=rss-4cbf6ab905be------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 16:07:28 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@howardchen_design/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[Generative AI as a Soothsayer : An AI Speculative DesignWorkshop]]></title>
            <link>https://medium.com/designmojito/generative-ai-as-a-soothsayer-an-ai-speculative-designworkshop-a9688f7a0939?source=rss-4cbf6ab905be------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9688f7a0939</guid>
            <category><![CDATA[speculative-design]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[generative-ai-use-cases]]></category>
            <category><![CDATA[ux-design]]></category>
            <category><![CDATA[design-thinking]]></category>
            <dc:creator><![CDATA[Howard Chen]]></dc:creator>
            <pubDate>Mon, 28 Oct 2024 07:02:25 GMT</pubDate>
            <atom:updated>2024-10-28T22:51:18.408Z</atom:updated>
            <content:encoded><![CDATA[<h3>Generative AI as a Soothsayer : An AI Speculative Design Workshop</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6s2sizmET9EfoIs4j4emEw.jpeg" /><figcaption>an AI designed microwave</figcaption></figure><p>In our previous article, we explored the expansive capabilities of generative AI. While we are still uncertain about how it can address precision issues and the risks it may pose, its potential in creative work is undeniable. This article delves into the possibilities of using generative AI for speculative design.</p><p><a href="https://medium.com/designmojito/how-to-interact-with-ai-tools-without-getting-dumber-part-ii-c27e411dd005">How to Interact with AI Tools Without Getting Dumber Part II</a></p><h4>What is Speculative Design?</h4><p>Speculative design is a concept that transcends reality. It involves swapping current technologies with future possibilities to provoke discussions about what the future might hold, encompassing not just technical aspects but also humanistic and societal implications.</p><blockquote>“The future is already here — it’s just not evenly distributed.”</blockquote><blockquote>― <strong>William Gibson</strong></blockquote><p>While some technologies exist, they haven’t yet become widespread or fully integrated into everyday life. Similarly, many inventive ideas never make it to market, and not all innovations are quantifiable or universally adopted.</p><p>Science fiction plays a vital role here, often exaggerating technology to prompt thought-provoking conversations. Take the movie <em>Her</em> (2013), which imagines a future where people might form deep emotional connections with AI. It’s a speculative vision that feels startlingly close to today’s world, highlighting how sci-fi and real tech developments continually inspire one another. Not to mention how much OpenAI seems to love the character Samantha.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*3iaB4LVjfJroseFYwzKJbg.jpeg" /><figcaption>Her (2013)</figcaption></figure><p>Speculative design helps businesses look beyond the present by imagining multiple future scenarios. Using the “Cone of Possibilities,” Delve sorts futures into <em>Probable</em>, <em>Possible</em>, <em>Plausible</em>, and <em>Preferred</em> categories, helping companies envision what they want — and don’t want — for the future. This approach blends creativity with strategy, making innovation more human-centered and forward-looking.</p><p><a href="https://www.delve.com/insights/speculative-design-and-a-cone-of-possibilities">Speculative Design and a Cone of Possibilities</a></p><h4><strong>Projecting the Future with AI’s Computational Power</strong></h4><p>Generative AI can act as a personal assistant, completing our ideas with just a provided direction, and often inspiring us through its random processes. This feature is incredibly useful in speculative design. Instead of requiring rigorous logical steps, AI speculative design encourages bold leaps to a final state, stimulating discussions and allowing us to rethink design methodologies.</p><h3>Tips for Using AI in Speculative Design</h3><h4>Orchestration Flow Tools</h4><p>We need to construct an imagined future and utilize creative thinking within this hypothetical scenario. By involving AI in every thought step, workflow-based interactive AI such as <a href="https://simpleinfo.ai/">SimpleInfo.Ai</a> , <a href="https://flowiseai.com/">Flowiseai.com</a> can help us document each stage and provide content with different vector designs. Unlike pure conversational AI, this method can avoid AI’s “forgetfulness” and foster iterative brainstorming at specific nodes. This can also help us generate prompts for different AIs, including images, videos, marketing copy, and web design, making the entire creative process smoother and more efficient.</p><p>You can structure it with different roles like this (although the order may vary depending on the business model/ strategy):<br>1. <strong>Futurist</strong>: Assume future problems.<br>2. <strong>CEO</strong>: Use the insights from the futurist to guide innovative directions.<br>3. <strong>Design Director</strong>: Develop design strategies based on the innovative directions.<br>4. <strong>Marketing Director</strong>: Position these products with a marketing plan.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Vpcc_Alqyv6f_x1T1art1g.png" /><figcaption>Orchestration Flow with LLM (Dinning on Mars)</figcaption></figure><h4>Drawing Inspiration from Evolutionary Creativity</h4><p>In Eisuke Tachikawa’s book <a href="https://www.kokuyo-furniture.co.jp/solution/mana-biz/2021/04/post-554.php">“Evolutionary Creativity— Creating Surviving Concepts through ‘Mutation &amp; Selection’”</a> (currently without an English translation), he cleverly summarizes nine ways of biological evolution and analogizes them to product design. I believe this can be combined with speculative design and incorporated into prompt engineering.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CgodvREiXueB3ek9WGiCOA.png" /><figcaption>Evolutional Creativity: Creating concepts the survive: “Mutation &amp; Selection”</figcaption></figure><p>We can definitely think of these 9 principles as different “branches” in a <strong>Tree of Thoughts</strong> framework. Each principle represents a distinct way to approach creative problem-solving or ideation (e.g. Imagine the eating experience of humans who have migrated to Mars, and use the following nine directions to inspire innovation):<br>1. <strong>Variation</strong>: Imagine extreme values.<br>2. <strong>Mimicry</strong>: Try to imitate ideal conditions.<br>3. <strong>Disappearance</strong>: Try to reduce standard equipment.<br>4. <strong>Proliferation</strong>: Try to increase quantities abnormally.<br>5. <strong>Transfer</strong>: Try to find entirely new domains.<br>6. <strong>Exchange</strong>: Try to replace with other objects.<br>7. <strong>Separation</strong>: Try to distinguish different elements.<br>8. <strong>Reversal</strong>: Try to think in opposite scenarios.<br>9. <strong>Fusion</strong>: Try to create unexpected combinations.</p><pre>You are a forward-looking product designer. Please design a kitchen appliance for use by humans immigrating to Mars in the future. Use the following principles to develop your ideas. <br>1. Variation: Imagine extreme values.<br>2. Mimicry: Try to imitate ideal conditions.<br>3. Disappearance: Try to reduce standard equipment.<br>4. Proliferation: Try to increase quantities abnormally.<br>5. Transfer: Try to find entirely new domains.<br>6. Exchange: Try to replace with other objects.<br>7. Separation: Try to distinguish different elements.<br>8. Reversal: Try to think in opposite scenarios.<br>9. Fusion: Try to create unexpected combinations.</pre><p>Here’s the result</p><pre>For a future Mars-bound kitchen appliance, let’s design a **Multi-Function Nutrient Synthesizer** tailored to the unique demands of Martian life. Here’s how each principle guides the design:<br><br>1. **Variation:**  <br>   Since resources are limited on Mars, this appliance synthesizes nutrients from extreme materials (like processed Martian soil) and uses minimal water, pushing the extremes of nutrient extraction and flavor enhancement.<br><br>2. **Mimicry:**  <br>   It replicates Earth-like growing conditions, integrating UV, temperature control, and humidity adjustments to produce small crops, herbs, or even nutrient-dense microgreens in a self-contained pod.<br><br>3. **Disappearance:**  <br>   The synthesizer replaces traditional bulky kitchen appliances (stoves, microwaves, refrigerators) with a single, compact unit, reducing space and energy use to the essentials.<br><br>4. **Proliferation:**  <br>   To support a Martian lifestyle, it creates multiple small servings, allowing astronauts to try various nutrient-rich &quot;meals&quot; tailored to their needs, encouraging a diverse, customizable diet with consistent portions.<br><br>5. **Transfer:**  <br>   Drawing on bioprinting and medical nutrition concepts, the synthesizer prints meals or nutrient supplements in edible, varied textures, ensuring balance and interest without traditional food sources.<br><br>6. **Exchange:**  <br>   It substitutes typical cooking ingredients with Martian-friendly, synthetic alternatives, such as algae-based proteins, lab-grown amino acids, and edible forms of calcium or iron to support bone health.<br><br>7. **Separation:**  <br>   It isolates nutrients into distinct flavors or components, enabling users to “mix and match” flavors, textures, and nutritional profiles for each meal, providing a creative eating experience.<br><br>8. **Reversal:**  <br>   Instead of cooking to enhance flavor, this device preserves nutrients by using low-heat synthesis, focusing on nutrient retention while reducing energy use.<br><br>9. **Fusion:**  <br>   The synthesizer doubles as a water purifier and atmospheric processor, recycling byproducts from cooking processes back into usable resources for Mars habitats, merging cooking and habitat sustainability functions.<br><br>In essence, the **Multi-Function Nutrient Synthesizer** isn’t just a food processor but a life-sustaining device that produces personalized, balanced meals, integrating Martian resource efficiency with Earth-like flavors and nutrition.</pre><h4>Using GenAI to fake your business from scratch</h4><p>In this workshop, we aim to demonstrate how quickly a non-professional can, through speculative design, plan and execute an entire business activity using generative AI tools.</p><p>Participants will design products, write ad copy, create videos, and build web pages, all within a few hours. The tools available today are incredibly intuitive, making it possible for non-design participants to produce impressive results. After just six hours, the outcomes were fascinating, showcasing the power of AI to support speculative design workflows.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oDepLhG0GeOdw-l7ovrslw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oPLRHasd_J3Hh_sCEoFmYg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1cgwNpTHEQHzzCenpmxI7Q.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Dy0Ycv3bY8H-D_99iCZl_Q.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5aX0Qh1rcSHx-jIAtpc5tw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/764/1*uNcqRW3B8hM3MU6V54agyw.png" /></figure><p>This concludes our exploration of using generative AI in speculative design. We hope these methods and tips inspire more creativity and help us collectively explore the limitless possibilities of future design.</p><p>—</p><p>👋<em>Hi! We’re Design Mojito, a blog dedicated to exploring intriguing topics related to product design. Our focus is on discussions that are thought-provoking, mind-enriching, and human-centered.</em></p><p>🍸<em>Follow Design Mojito for weekly articles that provide a refreshing and invigorating perspective!</em></p><p><a href="https://medium.com/designmojito">Design Mojito</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9688f7a0939" width="1" height="1" alt=""><hr><p><a href="https://medium.com/designmojito/generative-ai-as-a-soothsayer-an-ai-speculative-designworkshop-a9688f7a0939">Generative AI as a Soothsayer : An AI Speculative DesignWorkshop</a> was originally published in <a href="https://medium.com/designmojito">Design Mojito</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Interact with AI Tools Without Getting Dumber Part II]]></title>
            <link>https://medium.com/designmojito/how-to-interact-with-ai-tools-without-getting-dumber-part-ii-c27e411dd005?source=rss-4cbf6ab905be------2</link>
            <guid isPermaLink="false">https://medium.com/p/c27e411dd005</guid>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[design-thinking]]></category>
            <category><![CDATA[thinking-fast-and-slow]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <dc:creator><![CDATA[Howard Chen]]></dc:creator>
            <pubDate>Mon, 09 Sep 2024 14:02:10 GMT</pubDate>
            <atom:updated>2024-09-10T00:26:03.476Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/816/1*s8XWPbb9dtwQx0A6OZ99Xg.png" /></figure><p>In this part of the series, we explore how to interact with AI tools without falling into the trap of mediocrity. We’ll approach this topic through Daniel Kahneman’s System 1 and System 2 thinking, discovering how to create workflows that harness AI’s potential without diminishing our own cognitive abilities.</p><p>In the previous article, I discussed how AI’s computational processes can lead us to “forget to think.” The antidote is simple but crucial: We must carefully choose which tasks to delegate to AI and which ones to retain as human-driven. This article focuses on creative work, where that distinction becomes especially critical.</p><p><a href="https://medium.com/designmojito/how-to-interact-with-ai-tools-without-getting-dumber-part-i-46ef3ff3dda2">How to Interact with AI Tools Without Getting Dumber Part I</a></p><h3>Becoming Cyborgs in Our Workflows</h3><p>People often say that smartphones feel like extensions of their bodies, almost like new organs. As we continue to integrate intelligent devices into our daily lives, we are gradually becoming modern-day cyborgs. This shift extends into our professional workflows as well.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7WEdadYJu3JLumcm-rSEUQ.jpeg" /></figure><p>This transition is also evident in our workflows. Perhaps you’ve already found yourself relying on AI to draft a business email or even becoming more comfortable with vague language, knowing that GPT will polish your sentences into something more coherent. Without realizing it, our thinking patterns are changing. For instance, tools like GitHub Copilot are constantly predicting your next move as you code, streamlining your work and minimizing errors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a3f1rVL6vDCcZegbmwfHhw.png" /><figcaption><em>Example of Copilot Predictions.</em></figcaption></figure><h3>Our Original Way of Working: The Lazy Brain</h3><p>Daniel Kahneman’s distinction between <em>System 1</em> (fast, intuitive thinking) and <em>System 2</em> (slow, deliberate thinking) is familiar to many product designers. This framework not only helps us understand user behavior but also gives us insight into our own work patterns.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OyRz-fb_VpGYfjDdBE-tZg.png" /><figcaption>Which dot in the middle of circle is bigger?</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*l7ulKEXJsJUTOR2ANV2u3Q.png" /><figcaption>Our brain is lazy</figcaption></figure><p>The key difference between these two systems lies in their computational efficiency. <em>System 1</em> relies on shortcuts to conserve energy, while <em>System 2</em> is more resource-intensive. Our brains prefer to conserve energy, which is why we default to cognitive shortcuts (and biases) whenever possible.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J3eSX4gMgmARm7htIYbsJA.png" /></figure><h3>Using AI as an External Brain</h3><p>Interestingly, this cognitive process mirrors how large language models (LLMs) like GPT are trained. While humans rely on heuristics, LLMs operate using vectors. What’s fascinating is that AI’s abilities fall somewhere between <em>System 1</em> and <em>System 2</em>. AI doesn’t engage in deep, deliberate thought like <em>System 2</em>, but its vast dataset allows it to make predictions that often seem rational and well-considered, acting as a bridge to augment our cognitive processes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kt_BczJfeHbFLijWSOgwCw.png" /></figure><h3>Creative Workers’ Guide to Using AI as an External Brain: Delegate the Tedious, Strengthen the Creative</h3><p>To make our creative work smoother and more efficient, here are four strategies for using AI as an external brain:</p><h4>1. Soul-Searching Questions</h4><p>Creativity often brings moments of intellectual tension, where we jump between seemingly unrelated ideas. These intuitive leaps, guided by shortcuts, propel us forward. However, by employing AI’s “Chain of Thoughts” methodology, we can connect these disparate ideas, creating a cohesive flow in our creative output. This not only strengthens our designs but also prevents our ideas from becoming fragmented.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AMxBJBsZyA9yfZ8gP3Hv6Q.png" /></figure><h4>2. Rhetoric and AI</h4><p>Just like in rhetoric, where we use comparisons and analogies to simplify complex concepts, AI can help draw parallels we may not have noticed. By feeding AI one concept (A) and asking it to relate to another (B), we push the boundaries of our thinking. This method is highly effective for stretching creativity, as AI excels at recognizing patterns and making connections we might otherwise overlook.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DdCdKiufaJ5aqmOODQJtqg.png" /></figure><h4>3. Idea Projections: Parallel Universes in Creative Work</h4><p>Sometimes, creative work leaves us stuck. We focus on refining a single idea but end up recycling the same thoughts. Here, AI can project infinite potential outcomes, giving us numerous paths to explore. By using different AI models and simulations, we break free from mental roadblocks, considering alternative perspectives and possibilities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kpOArgnTsbeOHpMDRxnd_w.png" /></figure><h4>4. Creating Repetition Aesthetics</h4><p>The concept of <em>repetition aesthetics</em> allows us to take an initial idea and scale it into a larger narrative, automating the repetitive tasks that would otherwise drain our creative energy. AI can manage these tasks, leaving us free to explore and expand our concepts in ways that were previously unimaginable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YJf5xcidknZSNtqPiWIG-A.jpeg" /></figure><h3>Using AI in the Double Diamond Framework</h3><p>Creative work often follows the Double Diamond framework, alternating between divergent and convergent thinking. Great creative work balances the fast, intuitive processes of <em>System 1</em> with the slow, deliberate focus of <em>System 2</em>. By incorporating AI as an external brain, we can amplify both the quantity and quality of these divergent and convergent processes, allowing us to explore more ideas and refine them to a higher standard.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KU1oEoes7ikeKIOW_M1z7A@2x.png" /></figure><p>And these four tools can be applied throughout the creative process in various stages. For instance, during the convergent phase, Soul-Searching Questions can help refine problems, speeding up discussions and allowing us to pinpoint issues more efficiently. When it comes to the divergent phase of brainstorming solutions, AI-driven Rhetoric can quickly reference ideas from other fields, sparking new creative directions. Infinite Projections enables us to rapidly assess the completeness of ideas by visualizing their full potential. Finally, once we have a clear concept, we can utilize AI to scale and automate the output, generating Repetition Aesthetics that build cohesive narratives or designs at a much larger scale.</p><p>This integration ensures that AI enhances our creativity while keeping us in control of the process.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*icTsGBnDPcFby5p2RhUKeQ.png" /></figure><p>By using AI in this way, we ensure that it plays the right role in our creative workflows — one that complements and amplifies our human abilities rather than replacing them. The goal is not to become reliant on AI but to use it as a powerful tool that helps us think smarter, not harder.</p><p>In future articles, we will share practical case studies that demonstrate how to implement these tools effectively in creative workflows. Stay tuned for real-world examples that illustrate how AI can elevate your work without compromising your creative integrity.</p><p>👋<em>Hi! We’re Design Mojito, a blog dedicated to exploring intriguing topics related to product design. Our focus is on discussions that are thought-provoking, mind-enriching, and human-centered.</em></p><p>🍸<em>Follow Design Mojito for weekly articles that provide a refreshing and invigorating perspective!</em></p><p><a href="https://medium.com/designmojito">Design Mojito</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c27e411dd005" width="1" height="1" alt=""><hr><p><a href="https://medium.com/designmojito/how-to-interact-with-ai-tools-without-getting-dumber-part-ii-c27e411dd005">How to Interact with AI Tools Without Getting Dumber Part II</a> was originally published in <a href="https://medium.com/designmojito">Design Mojito</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Interact with AI Tools Without Getting Dumber Part I]]></title>
            <link>https://medium.com/designmojito/how-to-interact-with-ai-tools-without-getting-dumber-part-i-46ef3ff3dda2?source=rss-4cbf6ab905be------2</link>
            <guid isPermaLink="false">https://medium.com/p/46ef3ff3dda2</guid>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[user-experience]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Howard Chen]]></dc:creator>
            <pubDate>Mon, 19 Aug 2024 07:01:45 GMT</pubDate>
            <atom:updated>2024-08-21T01:20:15.851Z</atom:updated>
            <content:encoded><![CDATA[<p>This article aims to explore the psychological biases that can arise from the characteristics of generative AI. In our upcoming articles, we will also share insights on how creative professionals can use generative AI more intelligently.</p><blockquote>Imagine a scenario where a boss assigns a task, and instead of seeking clarification, an employee blindly follows through, potentially leading to significant errors.</blockquote><p>Even more concerning is the fact that some bosses may actually prefer an employee who asks fewer questions and delivers results directly, without the need for further discussion.</p><p>Over the past two years, the rapid growth of generative AI has continuously prompted us to consider how these tools can assist in our work, whether they might eventually replace jobs, and how to find the right workflows and applications. If tools like ChatGPT, which utilize large language models, have already started taking over many of your daily tasks, have you ever wondered if you’ve truly understood how their characteristics differ from the way we work?</p><h3>Here’s a simplified version: It’s really good at guessing words.</h3><p>Through extensive learning with the Transformer model, it can understand human language very precisely and narrow down the context to find the most reasonable response. In short, it’s like a student who excels at exams but doesn’t care about the meaning behind the answers. Its KPI is simply to produce answers that look good.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xywAjQyLnMPGqM33QoBWgw.png" /></figure><p>However, this characteristic sets it apart from other machine learning AI; it produces a unique output…</p><h3>Confident But Incorrect Responses: Hallucinations</h3><p>Large Language Models (LLMs), while very powerful, can sometimes produce what’s called hallucinations — answers that sound confident but are actually wrong. This can easily mislead people into making decisions based on incorrect information. So, it’s important to stay alert and question these models, even if their answers seem perfect.</p><p>In the past, AI often made mistakes with simple math problems or common knowledge. Nowadays, we rarely see these errors. Improvements in the basic logic have made AI much more reliable. Along with these advancements, fine-tuning techniques like Lamini have also helped make AI better.</p><p><a href="https://www.lamini.ai/blog/lamini-memory-tuning?fbclid=IwY2xjawEujGpleHRuA2FlbQIxMAABHSJ_x2d2sbACnIK31B-fmtzfmKPRG18LvKWSvmAb0ay1cyQxIi7t2ltpQQ_aem_DsJHBnsAkGcqxtqOEigT1w">Introducing Lamini Memory Tuning: 95% LLM Accuracy, 10x Fewer Hallucinations | Lamini - Enterprise LLM Platform</a></p><p>However, some people still refer to answers that fall outside the expected context or deal with topics that don’t have a clear-cut answer as “hallucinations.” But after understanding how LLMs work, it becomes clear that AI itself doesn’t hallucinate that much; it simply generates answers based on probabilistic rankings.</p><h3>Instead of thinking of these as hallucinations, they might simply be unwanted answers.</h3><p>To become better AI users, we can address these issues by considering two possibilities:</p><ol><li>The data used to train the model in that specific context might be insufficient.</li><li>The prompt given may not be accurately aligned with the desired answer.</li></ol><p>By applying fundamental prompt engineering techniques, such as distinguishing between zero-shot and few-shot prompts, and utilizing methods like chain-of-thought or tree-of-thoughts, we can optimize AI as a tool to enhance our work, making it clearer and more effective.</p><p><a href="https://www.promptingguide.ai/">Prompt Engineering Guide | Prompt Engineering Guide</a></p><p>Besides interacting with AI using more rigorous logic, let’s discuss the psychological biases that people may develop when using LLM.</p><h3>The Illusion of Perfection: <strong>What to Watch Out for in the Age of GenAI</strong></h3><p>As we increasingly rely on GenAI, there are three key things we need to be mindful of:</p><h3>Eliza Effect</h3><p>In computer science, the ELIZA effect refers to our tendency to attribute human traits — like experience, understanding, or empathy — to simple computer programs that use text interfaces. ELIZA was a symbolic AI chatbot developed in 1966 by Joseph Weizenbaum, designed to mimic a psychotherapist. Despite its basic text-processing abilities and clear limitations, many early users believed that ELIZA possessed real. intelligence and understanding. Through social engineering alone, it essentially passed the Turing test, showing how easily we can be convinced by AI when its goal is simply to produce answers that satisfy us.</p><p>This phenomenon is even more relevant today as many people believe that AI provides flawless answers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*dTMlPjVB26kGv17rrXjU0g.png" /></figure><h3>Anchoring Bias</h3><p>Even if I’m not familiar with a particular field, I can still receive an impressive answer by providing minimal prompts. This approach is efficient, but it also comes with the risk of being constrained by the framework, leading to an <strong>anchoring bias</strong>.</p><pre>show me the customer journey of ordering a food from uber eat. <br>display in a table, the table should contained pain points and initial solutions</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hqYGuXbrBPOf59W02dV7zA.png" /></figure><blockquote>“The zero-shot prompt directly instructs the model to perform a task without any additional examples to steer it.” — <a href="https://www.promptingguide.ai/techniques/zeroshot">Prompt Engineering Guide</a></blockquote><p>The performance of zero shot prompting perfectly illustrates just how advanced current language models have become. This is one reason why so many people have readily adopted and can efficiently use these tools. Starting a project with such ease and efficiency was once unimaginable, but now it’s a reality.</p><p>This approach aligns with the Postel’s Law perfectly (also known as the Robustness Principle) in computer interaction, formulated by Jon Postel:</p><blockquote>“Be liberal in what you accept, and conservative in what you send.”</blockquote><p>Users don’t need to provide precise instructions; almost no error messages will block AI generation. They don’t need to worry about typos, unclear wording, or even misunderstandings. The user doesn’t have to be entirely sure of the exact answer they need — the AI will guess the most likely meaning and produce a convincing response.</p><p>After interacting with the AI several times, you may notice that because it restricts its context to your question and builds its response based on surface-level language relationships, it provides only the most basic level of discourse. It might look well-formed, but it lacks depth.</p><p>However, this approach also introduces the risk of the <strong>anchoring bias</strong> — a cognitive bias where initial information (in this case, the AI’s first response) disproportionately influences our subsequent thinking.</p><p>If you aim to offer professional services to solve problems, this level of response is far from adequate for something you’re being paid for. While “Zero Shot Prompting” can help us quickly build a framework when we’re unfamiliar with a field, we must be careful not to let our thinking become constrained. If you do, anyone using a similarly vague input will get the same predictable, shallow answers, lacking real insight.</p><h3>Agency Problem</h3><p>To minimize hallucinations, instead of merely producing answers that seem plausible, AI can gather more user information, focusing on making users feel better understood during the interaction.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HmWs4NVdhhueDgOpX5hgDA.jpeg" /></figure><p>However, this approach comes with a trade-off: it means gradually granting AI more access and control, allowing it to book meetings, reserve restaurants, handle credit card transactions, or even drive the cars. As our trust in AI deepens, we may become more relaxed in overseeing these tasks. Yet, the idea that we’re entrusting AI — essentially a system that “guesses” — with such responsibilities is somewhat unsettling.</p><p>Similarly, while AI can offer answers that appear logical, the ethical and contextual implications must be carefully considered. There’s a long-standing concern, dating back to the 1950s, about the idea of machines making important decisions — a fear that remains relevant today. I’m also eager to see how Apple will build trust in handing over private data to AI.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tzdfszHMG5N-CeLKvjkqnA.jpeg" /><figcaption>In the movie “I, Robot,” Tuner is saved by a robot, which chooses him over a little girl based on logical reasoning. This is upsetting and highlights a critical issue.</figcaption></figure><h3><strong>Creative Work with AI: A Balance of Competition and Collaboration Between Probability and Creativity</strong></h3><p>When it comes to creative work with AI, it’s often about balancing competition and collaboration between probability and creativity. Generative AI, through extensive unsupervised learning, finds connections within data and leverages its remarkable ability to “guess” words, leaving the rest to the “god of probability” to deliver answers to humans — an incredibly efficient application.</p><p>What we can consider is how to balance our mental efforts in workflows, ensuring that the truly valuable aspects of creative work remain in human hands, while fully leveraging the unique capabilities of generative AI to achieve things we never imagined before.</p><p>This month marks the start of an exciting new journey — I’m thrilled to introduce <strong>Design Mojito</strong>, a publisher I co-founded with the talented Maeve Shen and Irene Chuang. In the future, I’ll be sharing more tips on how to effectively interact with AI in creative work. Stay tuned to <strong>Design Mojito</strong> for more insights! 🌟 Please follow <strong>Design Mojito</strong>.</p><p><a href="https://medium.com/designmojito">Design Mojito</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=46ef3ff3dda2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/designmojito/how-to-interact-with-ai-tools-without-getting-dumber-part-i-46ef3ff3dda2">How to Interact with AI Tools Without Getting Dumber Part I</a> was originally published in <a href="https://medium.com/designmojito">Design Mojito</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Framing AI for consistency: Midjourney techniques for product designers]]></title>
            <link>https://medium.com/design-bootcamp/framing-ai-for-consistency-midjourney-techniques-for-product-designers-378170fca9f3?source=rss-4cbf6ab905be------2</link>
            <guid isPermaLink="false">https://medium.com/p/378170fca9f3</guid>
            <category><![CDATA[midjourney]]></category>
            <category><![CDATA[illustration]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[ui-ux-design]]></category>
            <category><![CDATA[ai-generated-art]]></category>
            <dc:creator><![CDATA[Howard Chen]]></dc:creator>
            <pubDate>Tue, 31 Oct 2023 00:03:27 GMT</pubDate>
            <atom:updated>2023-10-31T06:14:28.861Z</atom:updated>
            <content:encoded><![CDATA[<p>Introduction to AI-generated images and a tutorial for achieving consistency with Midjourney.</p><figure><img alt="A highly detailed running border collie drawn by Midjourney" src="https://cdn-images-1.medium.com/max/512/1*9grcJTSeb0mygMnxyH4ddg@2x.jpeg" /><figcaption>a border collie drawn by Midjourney</figcaption></figure><p>With the rapid advancement of large language models, we’ve witnessed a significant evolution in image generation technology.</p><p>Designers must have likely embraced and marveled at these advancements, but they’ve also experienced concerns about the potential for their roles to be replaced.</p><p>In the e-commerce industry, where I encounter various visual outputs such as User Interface (Layouts, Components), Illustrations, Icons, Product Photos, and Visuals for Ads or Campaigns, I do wonder how can AI-generated images effectively assist these processes.</p><p>Up to now, I’ve experimented with tools like Midjourney, Stable Diffusion, Bing, Leonardo, Adobe, and more… Although Midjourney appears to be one of the most minimalistic ones, this is a method that can be harnessed for more business-oriented scenarios.</p><h3>Let’s get a super simplified understanding of how the diffusion model works</h3><p>The Diffusion Model is the current mainstream image generation tool that employs a trained model. During the learning process, it continuously introduces noise to the image, progressively understanding the relationship between the object and the added noise. It continues learning until it can reconstruct the image from a completely blank state (full of noise). This learning process is the “fixed forward diffusion process”</p><figure><img alt="How diffusion model is trained: the fixed forward diffusion process. Many pictures are diffused step by step." src="https://cdn-images-1.medium.com/max/1024/1*tV9mO5EckNuZk2nvff6_Mg.png" /><figcaption>How diffusion model is trained: the fixed forward diffusion process</figcaption></figure><p>Furthermore, thanks to the capabilities of large language models, AI can create from scratch on an image canvas based on our text input.</p><figure><img alt="How diffusion model generates pictures: Generative reverse denoising process. Able to draw anything." src="https://cdn-images-1.medium.com/max/1024/1*02tu_L1axRYzGsMIz0Schg.png" /><figcaption>How diffusion model generates pictures: Generative reverse denoising process</figcaption></figure><pre>/imagine border-collie, running in the woods, slow shot, photography </pre><figure><img alt="Midjourney recorded the denoising process: transforming four blurry pictures into authentic dog images." src="https://cdn-images-1.medium.com/max/512/0*m4bvLqYjK7lsxXM7.gif" /><figcaption>Midjourney recorded the denoising process</figcaption></figure><p>Midjourney and other major platforms allow you to achieve comprehensive results simply by communicating your requirements to the machine in a natural language-like manner. You can specify the theme, additional context, camera perspective, and style, and even provide detailed instructions for style execution.</p><p>If you’ve already experimented with it, you may have noticed that AI’s interpretation can sometimes differ from your understanding.</p><p>This can occur due to bias in the training data or variations in the effectiveness of specific terms. In practice, achieving the desired results often requires back-and-forth communication with the AI to ensure clarity and precision</p><blockquote>So, we understand that AI uses its best guess to predict your intent and proximate it pixel by pixel.</blockquote><h3>Creativity is a matter of probabilities (for AI)</h3><figure><img alt="The movie I, Robot (film, 2004) scence. A cop is interrogating a robot." src="https://cdn-images-1.medium.com/max/490/1*VqhEd8aVi-UK6YTXSHw-5w@2x.jpeg" /><figcaption>I, Robot (film, 2004)</figcaption></figure><blockquote>Spooner asks the robot, “Can a robot write a symphony? Can a robot turn a canvas into a beautiful masterpiece?” The robot responds, “Can you?” — I, Robot (film, 2004)</blockquote><p>AI excels at perceiving the overall visual “Look and Feel” of an image and can effortlessly execute specific visual directions down to stunning details.</p><p>Having said that, most of the work in UI/UX design is unrelated to the Diffusion Model. The majority of our time is spent on information design.</p><p>Even though we may seem similar to AI by learning from a vast number of competitors, what we observe is not just visual aesthetics. Instead, we focus on the business considerations and trade-offs being made, as well as how to explicitly or implicitly guide users toward a specific goal.</p><p>It involves confining ourselves within the realm of a particular touchpoint, business context, or the mindset of a specific user. We are constantly adjusting our thought framework to solve problems, rather than always starting from a blank canvas.</p><pre>/imagine high-resolution interface, shopping-theme application interface, <br>UI design, mobile app, iPhone, iOS, Apple Design Award, usability, <br>single screen, flat colors, Helvetica, native app, design, user interface, <br>Figma, Apple Store, Dribbble. --v 4** </pre><figure><img alt="Four visually appealing mobile user interface mockups were generated by Midjourney." src="https://cdn-images-1.medium.com/max/1024/1*dat2r2SwCnFsLJQ4m4zeSA.png" /><figcaption>The look and feel explorations with Midjourney v4</figcaption></figure><p>Enabling this algorithm to engage in behavior such as learning from Behance for visual proposals occurs during the ideation stage, separate from functional discussions. It represents only a small portion of UI/UX design work.</p><p>However, understanding how it learns makes it easier to position this tool correctly. It is like knowing that ChatGPT is a “language” model, not another type of model.</p><h3>Maintaining consistency in the AI</h3><p>If your graphic design is a standalone work, then image generation tools are undoubtedly useful for the sake of the incredible details.</p><p>However, for product designers involved in creating user experiences or advertising experiences, we often need a series of creations to ensure visual harmony and effective communication of a unified purpose.</p><p>Consistency is crucial in various aspects, including consistent design elements, consistent UI illustration style, consistent marketing campaign visuals, and consistent photography style.</p><h3>Chaos</h3><p>To enhance the originality of images and reduce the likelihood of generating the same image under similar conditions, Midjourney incorporates functions such as “Stylize” and “Chaos.”</p><p>With these features, each generation and every set of four generated images exhibit random variations, adding an element of uniqueness to the process.</p><figure><img alt="Four 3D-rendered blue shopping bags generated by Midjourney are different from each other in the position of the star decorations." src="https://cdn-images-1.medium.com/max/486/1*k5ScIvkJHpBb4T9tacE09Q.png" /><figcaption>Four pictures with “chaos”</figcaption></figure><p>We cannot directly instruct AI to maintain the style consistency with a specific image. Instead, we rely on indirect methods by observing text-based outputs, as certain terms have a stronger rendering capability, leading to approximately consistent results.</p><p>In practice, we find that achieving a consistent photo style, with elements like tone and scene, can be relatively easier through textual instructions.</p><p>But illustration is not easy to stay consistent by text prompt. Stable Diffusion can accomplish this through methods like “ControlNet” and “Inpainting” to highly customize the part of the image. Midjourney is also able to do it in a collaborative approach, as seen in the case of UI illustrations.</p><h3>Step 1. Find the style prompt</h3><p>When you find a reference image with a similar style, Midjourney’s “Describe” feature provides you with four sets of keywords at once. This function doesn’t assist you in completely replicating the image, but it offers keywords that provide similar rendering results from multiple AI perspectives. You can directly use these words for inspiration or recombine them to explore your stylistic preferences. This stage ensures that the creative process remains an inspiration source rather than an imitation. Additionally, <a href="https://replicate.com/methexis-inc/img2prompt">Replicate.com</a> also offers the effect of generating text from images.</p><figure><img alt="The describe function of Midjourney shows four different prompts to describe an isometric illustration." src="https://cdn-images-1.medium.com/max/720/1*C45PeAYMecjRYu-2UF01_A.png" /><figcaption>The describe function of Midjourney</figcaption></figure><p>You can extract from the Showcase any AI communities or showcases of different tools, such as <a href="https://leonardo.ai/">Leonardo</a>, <a href="https://civitai.com/">Civitai</a>, and <a href="https://www.imiprompt.com/">IMI Prompt</a>.</p><p>Additionally, a useful feature of Midjourney is its similarity function. You can discover image results similar to yours, even if they were created using entirely different keywords.</p><p>For instance, if I need to create an isometric and flat illustration that looks clean and is easily expandable in a UI context, I would search for a suitable prompt.</p><h3>Step 2. Create basic prompt</h3><p>Because Midjourney can yield different random results between different images, we can attempt to have it create multiple elements within the same image.</p><p>We can start by using a keyword like ‘Multi-view’ or a similar approach, along with a broad term such as ‘People,’ ‘Groceries,’ and ‘Buildings.’ This way, AI can generate various interpretations for each element, all with different styles. Make sure it’s on a white background.</p><pre>/imagine isometric shopping mall, multiview, flat minimalism style, vector, <br>white background --s 750 --v 5.1<br><br>/imagine Random 3D Real life objects illustrations, multiveiw ,<br>white background , add light background, hyper - realistictyle, <br>8K Resolution, --s 750 --v 5.1</pre><figure><img alt="A isometric style buildings and interiors" src="https://cdn-images-1.medium.com/max/1024/1*GfVfPxWT0s8Fg1F0citHkQ.png" /></figure><figure><img alt="A 3D style with pink background with objects looks like toys" src="https://cdn-images-1.medium.com/max/1024/1*xx-eXsoo9VoxLoQ5cUsiMA.png" /><figcaption>The multi-view with isometric style and 3D style</figcaption></figure><p>Cropping three or more images, we aim to provide AI with multiple images simultaneously for image generation. Its learning process resembles the human cognitive brain’s search for similarities, so the variations among these three images should align with the variations expected in this series of illustrations. For example, if we provide three images with the same object shape, that shape will become fixed. What we intend to keep consistent is the illustration style.</p><p><a href="https://docs.midjourney.com/docs/image-prompts">Copy the URL for the image prompt</a></p><figure><img alt="Cutting a part of the multi-view" src="https://cdn-images-1.medium.com/max/485/1*YjoAC5K7c1pXWVIuqCw-hg.png" /></figure><figure><img alt="four pictures uploaded to the chatbox" src="https://cdn-images-1.medium.com/max/668/1*-aoL-M7z-xO62MGWC-RQlQ.png" /></figure><p>Get the seed number of the picture:<a href="https://docs.midjourney.com/docs/seeds"> (Midjourney: Seed Parameter)</a></p><figure><img alt="The seed numbers appear in the dialog include the previous isometric picture" src="https://cdn-images-1.medium.com/max/1024/1*ROLrtZ_5qHPZpEbWQg9rAQ.png" /></figure><pre>/imagine https://s.mj.run/AAAAAAAAAAA, <br>https://s.mj.run/BBBBBBBBBBB, <br>https://s.mj.run/CCCCCCCCCCC,<br>#3-5 image prompt url#<br><br>_______________, <br>#This part is the variation part#<br><br>flat minimalism style, vector, white background<br>#same prompt without multi-view#<br><br>--seed 1204243555 --s 750 --v 5.1<br>#seed number and same style#</pre><p>The most powerful aspect of this prompt is its ability to generate images. Following that, we input the same Style field as mentioned earlier, excluding the Multi-view part. Additionally, we added the Seed Number from the previous image. After the URL, at the very beginning of the text input, we can specify any illustration content we want to create.</p><h3>Step 3. Generate series illustrations</h3><p>As long as we switch out the keywords, we have successfully constrained Midjourney to work within these rules for creative purposes!</p><p>I can imagine your work right now is quite chaotic. It’s a good idea to write down your desired illustration script in your notes, start executing it, and document each attempt with different keywords to achieve the desired results.</p><pre>/imagine https://s.mj.run/AAAAAAAAAAA, <br>https://s.mj.run/BBBBBBBBBBB, <br>https://s.mj.run/CCCCCCCCCCC,<br>Castle, <br>flat minimalism style, vector, white background<br>--seed 1204243555 --s 750 --v 5.1</pre><figure><img alt="Eight different building appears separately generated by Midjourney" src="https://cdn-images-1.medium.com/max/1024/1*cRyvBGHqKFz0Mbl1cUSiIg.png" /></figure><p>We can see how effective AI’s “conceptual blending” fits in a frame.</p><figure><img alt="The process of two different styles from multi-view to the separated illustrations" src="https://cdn-images-1.medium.com/max/1024/1*L94a_7_dY1PdTCIu7euWjQ.png" /></figure><p>Midjourney also offers the “<a href="https://docs.midjourney.com/docs/vary-region-remix">Vary (Region</a>)” after v5.0 feature to assist in fine-tuning this process.</p><p>Due to the simplicity of the image content, AI tends to lean towards a certain level of complexity. In Midjourney, you can experiment with adjusting the “<a href="https://docs.midjourney.com/docs/stylize">Stylize</a>” value to try different effects.</p><h3>Post-production</h3><p>If your style is sufficiently flat, you can put the image Illustrator and use the “<a href="https://helpx.adobe.com/illustrator/using/image-trace.html">Image Trace</a>” feature or <a href="https://vectorizer.ai/">Vectorizer.ai </a>to convert it directly into vectors. During the creation process, you can utilize the ‘<a href="https://docs.midjourney.com/docs/no"> — no</a>’ keyword to prevent Midjourney from using gradient coloring, resulting in cleaner lines for color recognition.</p><p>Alternatively, you can also employ one of Photoshop’s methods for background removal.”</p><h3>Mastering It</h3><p>ChatGPT introduced the capability to create a similar style using DALL-E 3, allowing for the generation of different words directly within the same style. However, it manages its “chaos” in different ways.</p><p>Adobe Illustrator also recently launched AI vector creation functionality, it’s exciting to see how AI helps vector illustrations. It hasn’t provided the same level of creative freedom as the process described earlier. The AI functionalities are worth looking forward to as aids for product designers.</p><p>AI-generated art is indeed impressive and poised to disrupt industries. However, practical use requires an understanding of its underlying principles, just as we do with any design software. While it is powerful, without the ability to harness it effectively, it remains limited.</p><p>By mastering the tools, we can allocate more time for style exploration and produce usable images in our projects. This allows designers to have more time to ponder specific design decisions and reduce limitations on their creative expression. By integrating it into our workflow, designers remain designers.</p><h3>Related tutorial</h3><ul><li><a href="https://medium.com/@steinsfu/diffusion-model-clearly-explained-cd331bd41166">Diffusion Model Clearly Explained!</a></li><li><a href="https://bootcamp.uxdesign.cc/top-10-ai-tools-to-post-edit-your-midjourney-images-c3ee17724f19">Top 10 AI Tools to Post-Edit Your Midjourney Images</a></li><li><a href="https://bootcamp.uxdesign.cc/recolor-your-midjourney-image-with-the-new-adobe-illustrator-ai-tool-6d285ca5034e">Recolor Your Midjourney Image with The New Adobe Illustrator AI Tool</a></li><li><a href="https://bootcamp.uxdesign.cc/a-guide-to-consistent-characters-with-midjourney-v5-47ef74e6b698">A Guide To Consistent Characters With Midjourney (V5)</a></li><li><a href="https://linusekenstam.substack.com/p/tutorial-how-to-create-consistent">Tutorial: How to create consistent characters in Midjourney</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=378170fca9f3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/design-bootcamp/framing-ai-for-consistency-midjourney-techniques-for-product-designers-378170fca9f3">Framing AI for consistency: Midjourney techniques for product designers</a> was originally published in <a href="https://medium.com/design-bootcamp">Bootcamp</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>