<?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 Avi Latner on Medium]]></title>
        <description><![CDATA[Stories by Avi Latner on Medium]]></description>
        <link>https://medium.com/@avilatner?source=rss-405c12255563------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*y-4qQLTz7lc7BKL256WXMA.jpeg</url>
            <title>Stories by Avi Latner on Medium</title>
            <link>https://medium.com/@avilatner?source=rss-405c12255563------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 14:30:56 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@avilatner/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[Trends in 3D graphic design and game design uncovered in interviews]]></title>
            <link>https://uxplanet.org/trends-in-3d-graphic-design-and-game-design-uncovered-in-interviews-3b5d5e8befbd?source=rss-405c12255563------2</link>
            <guid isPermaLink="false">https://medium.com/p/3b5d5e8befbd</guid>
            <category><![CDATA[game-design]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[3d-technology]]></category>
            <category><![CDATA[3d-modeling]]></category>
            <dc:creator><![CDATA[Avi Latner]]></dc:creator>
            <pubDate>Wed, 31 Mar 2021 12:29:19 GMT</pubDate>
            <atom:updated>2021-03-31T12:29:19.470Z</atom:updated>
            <content:encoded><![CDATA[<h3>How I came about to collecting these insights</h3><p>I was about to jump on a plane to Oslo three times and each time I had to cancel. First, it was my own doing, I stayed a few days longer waiting for my second Covid vaccine shot. Then, just 36 hours before my second flight, Israel, where I reside, closed its borders for travel, there was no going in or out. Finally, when the Israeli restriction lifted, Norway denied entry for non residents.</p><p>So I haven’t physically met my partners before we decided to start a company together and I haven’t met them since. We all shared a passion for unleashing creativity through technology and on that basis we decided to build a tool for real-time 3D creation.</p><figure><img alt="A few palm trees and a dinner in the desert" src="https://cdn-images-1.medium.com/max/1024/1*04PRT3RUlu71KdvodhNQoA.png" /><figcaption>‘Low Poly’ desert street, Øyvind Sørøy co-founder at <a href="http://sloyd.studio/">Sloyd</a></figcaption></figure><p>We virtually met each other in <a href="https://www.antler.co/oslo">Antler</a>, a VC venture builder in Oslo, and in order to come up with an idea to pursue, we conducted interviews with over thirty people from the creative industries who use 3D in their daily work. We asked them about the tools they use and the trends they see.</p><p>Later, when we looked at the insights we got from these interviews we thought they’d be valuable for others as well. Especially for designers that are considering expanding their toolset into CGI (computer graphic imagery) and design students that are considering a career in game development. With those two groups in mind, this review starts with 3D usages tools in graphic and motion design, then it continues to tools in game design and concludes with reviewing methods for speeding up 3D creation in both industries.</p><h3>First, graphic and motion design — who uses 3D and for what?</h3><p>Only a few of the brand graphic and motion designers feel fully proficient in 3D modeling. Yet, most of the motion designers use it in their daily work and so do part of the graphic designers. In addition, about a half of these brand professionals are interested in improving their 3D skills. Even amongst the most skilled, it is common, in complex 3D assignments to outsource the job to 3D artists.</p><p>Most interviewees shared the sentiment that CGI is an increasing trend in brand marketing, albeit one of a few trends (others mentioned were ‘authentic’ photography and flat illustrations). Also, most think, that there are more designers getting into 3D and that recent design school graduates are coming out of design school with more 3D proficiency than those only a few years ago.</p><p>In some cases, 3D is used as part of a unique style to differentiate the brand. In merchandise and packaging work 3D is an important tool that’s used very often to reduce costs. Printing out packages and doing professional photoshoots cost a lot. CGI is a cheaper substitute either for prototyping concepts or in some cases even for replacing a physical item in the end visual. There are also cases where promotional material goes out before the product is made and then CGI based on product rendering has to be used.</p><h3>The most used tools by brand graphic and motion designers</h3><p>There are a lot of tools and the summary here is by no means a comprehensive collection of everything out there. Also, since it was not compiled in a quantitative survey, I’m not proposing that this is the absolute truth. Still, after talking to over thirty professionals, some patterns emerge.</p><p>By far, the most mentioned ‘core’ 3D tool by brand and graphic designers is <a href="https://www.maxon.net/en/cinema-4d">Cinema4D</a>. Apparently that’s due to good integration into <a href="https://www.adobe.com/il_en/products/aftereffects.html">After Effects</a> and also because the interface resembles other editing tools designers use, making it easier to learn for these designers. Some, but to a lesser extent, are using <a href="https://www.rhino3d.com/">Rhino3D</a> and <a href="https://www.blender.org/">Blender</a>.</p><p>One designer is actually using the <a href="https://unity.com/">Unity</a> game engine for his 3D needs. This ties into a trend of tools whose core capability is not 3D but are incorporating 3D into them and are becoming a substitute for the core 3D tool. Illustrators, who are very proficient in <a href="https://www.adobe.com/il_en/products/illustrator.html">Adobe Illustrator</a> are using its 3D capabilities for simple 3D tasks, without needing another tool. Likewise, some motion designers use the 3D capabilities in Adobe After Effects. A designer we talked to, who so far only worked with 2D, finds that he can get a good depth desired effect just by using shading and lighting in Photoshop.</p><p>When it comes to ‘speciality’ 3D tools, in the area of package design and merchandise, <a href="https://www.livesurface.com/">livesurface</a> is used to display a visual such as a label on a 3D item. Designers proficient with 3D, that do a lot of CGI projects, also use <a href="https://www.keyshot.com/">Keyshot</a> for rendering and lighting.</p><p>New browser based 3D tools recently emerged. They rely on advancements in<a href="https://www.khronos.org/webgl/wiki/Main_Page"> WebGL</a>, the standard for running 3D graphics and utilising GPU from the browser, and are often ‘light weight’ and easier to learn than application tools. The designers have heard about them and have tried them out, but they’ve yet to apply them in a work situation. Three such tools that were mentioned were <a href="https://www.vectary.com/">Vectray</a>, <a href="https://www.blend4web.com/en/">Blend4Web</a> and <a href="https://spline.design/">Spline</a>.</p><figure><img alt="A black vintage American car" src="https://cdn-images-1.medium.com/max/894/1*7uSX8-USljBucgpqDPOqtg.png" /><figcaption>‘Low Poly’ car, Øyvind Sørøy co-founder at <a href="http://sloyd.studio/">Sloyd</a></figcaption></figure><h3>On to the most used tools by game designers and TV animators</h3><p>Game developers and TV animators use a larger set of tools for 3Ds. In big studios, for every step in the elaborate production process, block modeling, detailing, texturing, rendering and post processing, there’s a speciality tool and sometimes more than one.</p><p>Out of the ‘core’ 3D creation tools, for these users, Cinema4D is a lot less common. Instead they are using one or a combination of<a href="https://www.autodesk.com/products/maya/overview?term=1-YEAR"> Maya</a>,<a href="https://www.autodesk.com/products/3ds-max/overview?term=1-YEAR"> 3D-Max</a>,<a href="https://www.sidefx.com/products/houdini/"> Houdini</a> and Blender (a combination since each tool has a relatively strong point. For example, for Houdini it’s procedural generation). I was expecting to hear from indie developers that they use mainly Blender and from AAA studio designers that they rely more on Maya and Houdini. However, I found that Blender has made its way also into AAA game studios and that paid software is also prevalent in Indie studios (Blender is a free open-source tool).</p><p>The speciality tools used for detailing include<a href="http://pixologic.com/features/about-zbrush.php"> Zbrush</a> for sculpturing,<a href="https://www.adobe.com/il_en/products/medium.html"> Medium by Adobe</a> for sculpting with VR and<a href="https://www.marvelousdesigner.com/product/overview"> Marvellous Designer</a> for creating cloths. The main tool used for texturing in<a href="https://www.substance3d.com/products/substance-designer/"> Substance design</a> but I’ve heard of designers who use<a href="https://3dcoat.com/"> 3d coat</a> and in big studios also <a href="https://artomatix.com/">Artomatix</a>. For rendering I most often heard Keyshot but some mentioned<a href="https://www.autodesk.com/products/arnold/overview?plc=ARNOL&amp;term=1-YEAR&amp;support=ADVANCED&amp;quantity=1"> Arnold</a>. Finally, for post processing<a href="https://marmoset.co/toolbag/"> Marmoset Toolbag</a> was mentioned.</p><h3>3D trends in game design and TV production</h3><p>The game industry is massive and is growing fast. But the cost of AAA game production is growing even faster than revenue. AAA studios are also facing competition from the booming Indy development sector and from casual mobile game producers. All of that makes the AAA model nearly unsustainable, a sentiment reflected by<a href="https://www.pcgamer.com/cliff-bleszinski-warns-that-triple-a-game-development-is-nearly-unsustainable/"> Cliff Bleszinski</a> and<a href="https://www.gamesindustry.biz/articles/2020-06-23-shawn-layden-gamelab"> Shawn Layde</a>, two gaming executives. As a result, a large chunk of the 3D work is outsourced to production studios taking advantage of international labor cost arbitrage and giving the flexibility to scale the labor force up and down quickly based on the production stage.</p><p>Intense competition is likely to continue and will drive a rapid development of technology that pushes the boundaries of game experience and also drives down the cost of development. Since about 40% (according to a few sources) of the labor cost is getting 3D models through the production pipeline, many of the cost cutting technologies will be in 3D.</p><p>Game engines’ ability to handle geometry load is rapidly improving. Now some game engines can outperform TV and movie rendering solutions and they have the added benefit of being realtime. Therefore, one of our interviewees, who has worked at the cutting edge movie and game companies, believes that soon movie production will start using games engines. This has already started happening in the production of the live action series <a href="https://www.youtube.com/watch?v=ICuqzhViWAI">The Mandalorian</a>.</p><h3>Outsourcing 3D with Marketplaces</h3><p>The centerpiece of a CGI advertisement, the level boss in an adventure game or lead character in an animation, those are usually handcrafted with love. However, the props (i.e. items) next to them, the environments that surround them and even the secondary characters around them, they may very well be taken from a marketplace or a generator.</p><p>The marketplace for 3D assets that is top-of mind for professionals across all positions and creative industries is<a href="https://www.turbosquid.com/"> Turbosquid</a>. The people in game design just as often use the<a href="https://assetstore.unity.com/"> Unity asset store</a> and the<a href="https://www.unrealengine.com/marketplace/en-US/store"> Unreal marketplace</a>. Those who recall more than one marketplace by name usually mention <a href="http://cgtrader.com/">Cgtrader</a>,<a href="https://free3d.com/"> Free3d</a>, or<a href="https://sketchfab.com/"> SketchFAB</a>. Designers who need a 3D visual but don’t need the 3D file, will source 3D from Shuttersock or another stock site where they already have a subscription. (Turbosquid was acquired by Shutterstock but it is not part of the subscription to the main Shutterstock product). A brand designer, who often needs realistic furniture for his work, reported that he uses<a href="https://downloads.vitra.com/#/media?media_category_media_type=root.cad.3d_dxf.%7Croot.cad.obj.&amp;per_page=25"> Vitra</a>. Vitra is a retailer that creates digital twins for their furniture and offers it for free. There’s also<a href="https://www.mixamo.com/#/"> Mixamo</a>, a store for 3D characters that come rigged for easy animation, and <a href="http://www.makehumancommunity.org/">MakeHuman</a> an open source alternative.</p><p>The amount of usage of marketplace assets is mixed. In general, there’s more usage of marketplace assets in game design then in brand CGI and animation since game scenes can be very dense in items. Still there are some concerns that are hindering further adoption of marketplace assets.</p><p>One concern is the lack of consistency in style and quality among assets. Since in a market place, assets are provided by different creators it can be hard to match two pieces from different creators to one artistic style.</p><p>Another concern is that by using a finished asset in your project you lose uniqueness, as other projects will also use that exact same item. Some indie developers, whose emphasis is on art, shy away from using marketplace assets even if it could save them a lot of time.</p><p>In addition, assessing the quality of the asset is also challenging. The 3D item may look good in preview, but only after you buy it and start using it you find out if it was built properly. Things like how UV-unwrapping and geometry was done are revealed only after purchase. Geometry, especially, when done improperly (over usage of triangles) can have a big implication on performance and having to clean that up defeats the purpose of outsourcing 3D to save time.</p><p>In casual gaming, when the business model is based on in-app purchases and advertising, companies are concerned with rights and licensing issues when they use third-party assets.</p><h3>Automating 3D creation with procedural generators</h3><p>Another way to cut on costs and save time is to use procedural generators. These generators predominantly take prefabricated assets and combine them together in different ways allowing for many variations. Generators create buildings, cities, vegetation, terrain and even textures. Since procedural generators are built for customisation, a user can get a unique asset out of them, and in that sense they are more appealing than a finished marketplace asset.</p><p>There are games who built full procedural worlds, notable examples are<a href="https://www.nomanssky.com/"> No Man Sky</a> and<a href="https://www.valheimgame.com/"> Valheim</a>. However, these games developers built a generator tailored for their specific game and it is not something that is resold to others.</p><p>Procedural generators that are available in the market seem to have a very low adoption rate and most of the game developers do not use them. One reason that they are not widely used is that despite the fact that they are meant to save time they are still quite complex. Another reason is that the generators have a very limited set of styles and that game designers think they are not practical to use.</p><p>This is probably because procedural generation is only just starting and even some of the most promising ones, such as<a href="https://archimatix.com/"> Archimatix</a>, are just a one person passion project and not a commercial venture. In the near future, procedural generation tools will mature and then they’ll become more mainstream.</p><p>While building and city generators adoption has not taken up, Substance Designer’s procedural solution for textures has become mainstream. Also some of our interviewees have used procedural solutions for vegetation and terrain and they mentioned<a href="https://www.procedural-worlds.com/"> Procedural Worlds</a> and<a href="https://www.awesometech.no/index.php/procedural-placement/"> Vegetation Studio Pro</a>.</p><h3>What all this means</h3><p>These are exciting times for 3D in media, games and advertising. There’s a lot of innovation that is spurring new creative options. For us, it meant that we wanted to join this creative revolution and we started <a href="https://www.sloyd.studio/">Sloyd</a>. We are working on automating 3D creative with a new approach to modeling. Instead of models made from a mesh of polygons, we are building models with math and functions. Therefore our models are generated in milliseconds at runtime and are easily customised. We hope this will make 3D easier for people who are just starting and save tedious creation tasks for experienced designers. Let us know what these trends mean to you by commenting on the article.</p><h3>More on our research method</h3><p><a href="https://sloyd.studio/">Sloyd’s</a> interviews were conducted by my two co-founders <a href="https://www.linkedin.com/in/%C3%B8yvind-s%C3%B8r%C3%B8y-643823b/">Øyvind Sørøy</a> and <a href="https://www.linkedin.com/in/andreas-bryn-edesberg-44526237/">Andreas Edesberg</a> and by me.</p><p>The people we talked to held these titles: Indie developer, 3D artist, creative director, technical creative director, VP creative, executive producer, brand designer, senior brand designer, motion designer, UI designer, design lead, marketing designer, motion developer, freelance designer and managing director.</p><p>They worked in these types of companies: AAA game developer, smaller indie game studio, production studio, design agency, media startup, media and web enterprise, a digital consulting firm and as freelancers and indie developer hobbyists.</p><figure><img alt="A software interface with two chairs in 3D and one of the “unrapped”" src="https://cdn-images-1.medium.com/max/1024/1*TqP9lYh9oH1eVnvIyl5ssw.png" /><figcaption>An object exported from <a href="https://sloyd.studio/">Sloyd</a> — created on runtime with an automatic UV unwrapping</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3b5d5e8befbd" width="1" height="1" alt=""><hr><p><a href="https://uxplanet.org/trends-in-3d-graphic-design-and-game-design-uncovered-in-interviews-3b5d5e8befbd">Trends in 3D graphic design and game design uncovered in interviews</a> was originally published in <a href="https://uxplanet.org/">UX Planet</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Training a Machine to Draw]]></title>
            <link>https://medium.com/swlh/training-a-machine-to-draw-fb9668ddc0ce?source=rss-405c12255563------2</link>
            <guid isPermaLink="false">https://medium.com/p/fb9668ddc0ce</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[illustration]]></category>
            <category><![CDATA[gans]]></category>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[Avi Latner]]></dc:creator>
            <pubDate>Fri, 04 Dec 2020 09:26:43 GMT</pubDate>
            <atom:updated>2020-12-16T23:09:23.927Z</atom:updated>
            <content:encoded><![CDATA[<h4>A review of methods I’ve used to train a machine to draw. What worked, what didn’t work, and future directions.</h4><p><em>— modified 12/08/2020 —</em></p><h4>Summary</h4><p>In this set of experiments, I used 4000 flat art illustrations from various artists. Most images contained a person doing something. To train, I used a <a href="https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix">Pytorch pix-2-pix implementation</a> and a few other libraries.</p><p>Training objectives were:</p><ol><li>To improve the machine’s ability to color images without color smearing.</li><li>To train the machine to add and complete lines. That’s one method to take simple doodles and turn them into more elaborate drawings.</li><li>To train the machine to fix proportions of input drawings. For most of us, keeping the proportions right is probably the thing we struggle with the most when we draw. So if we can get a machine to fix proportions it can be a huge step towards machine learning assisted drawing.</li><li>Create a model that uses photos as a source for drawings. This could be useful because in photos objects already come in the right proportions.</li></ol><p>Coloring was successful and also, to a large extent, completing missing lines was successful. Fixing proportions, however, did not work well. Finally, using pictures as input looks promising but it has to used with another input source.</p><h4>1. Coloring without smearing</h4><p><a href="https://avilatner.medium.com/designers-getting-creative-with-machine-learning-imagery-349f2a3bf9d">In previous training</a>, I showed the ability to color illustrations. There’s also a dedicated pix-2-pix and cycleGAN photo coloring training (which I did not use), so it is not a big surprise that coloring works. This time, with a much larger data set ‘coloring’ worked very well and usually without smearing and going out of line.</p><p>In the set of pictures below, the left-most image ‘real_A’ is what the machine got as input in a validation set. The right-most image ‘real_B’ is what the expected result should be. The middle picture ‘fake_B’ is what the machine created. You can see in pictures 1 to 3 that the machine-generated images closely resemble the expected result.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/810/1*G34ZzuUx72v5lUoY9g-K_g.png" /><figcaption>Picture 1 -source image Pablo Stanely</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/810/1*JQjIelcVvLjUDfThd_iGpg.png" /><figcaption>Picture 2 — source image Pablo Stanely</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/805/1*TtKpIaSKBlFdFcar4wVJIg.png" /><figcaption>Picture 3 — source image Pablo Stanely</figcaption></figure><h4>2. The capacity to supplement missing parts</h4><p>I produced the input images with <a href="https://www.imagemagick.org/Usage/transform/#canny">ImageMagick&#39;s implementation of canny edge detection</a>. In order to teach the machine to supplement missing parts, I created a set of edge drawings with a lower threshold (sigma=2). That way, these input images had some of the outline missing.</p><p>In ‘picture 4’ below the person’s whole upper body is missing. The machine knew how to complete the upper body’s shape and even added a hand.</p><p>In ‘picture 5’ below the back of the coat, the right sleeve and the shoes are missing. The machine knew how to complete them, even if the connection between the right sleeve and the are is not too convincing.</p><p>In ‘picture 6’, again the upper part of the body is missing the machine knew to complete it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/802/1*Gv5B7yobaTlMFEWQMX6vAg.png" /><figcaption>Picture 4 — source image Katerina Limpitsouni</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/805/1*CGaaFyF1K8WTPqdRK3mjBw.png" /><figcaption>Picture 5 — source image Pablo Stanly</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/802/1*QWmLitnL20WZlIFI2iHX2g.png" /><figcaption>Picture 6 — source image Katerina Limpitsouni</figcaption></figure><h4>3. Fixing proportions</h4><p>To teach the machine to fix proportions I created a set of distorted drawings. In picture 7 below, the top-left corner shows a well-proportioned drawing. That top-left illustration is the expected outcome of this training. I created the distorted inputs by compressing or stretching part of an image or all the images. In picture 7 below we see illustrations (going clockwise from top-right) with an upper-body that’s too short, a lower-body that’s too short, and a whole body that’s stretched.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*owovUYNxWsjq9SvlvscEwA.png" /><figcaption>Picture 7 — source image Pablo Stanely</figcaption></figure><p>It did not work. Picture 8 shows a sample of what the machine created (middle image) and it&#39;s evident that the machine did not pick-up the patterns. If I had used just one type of distortion, for example, vertically stretching, I’m sure the results would have been better. <a href="https://medium.com/@avilatner/case-in-point-machine-learning-is-a-practical-tool-for-designers-3f070ac0f888">In a previous experiment</a>, I got the machine to change outlines with a more homogeneous data set. But fixing just one type of distortion is not particularly useful.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/799/1*hUs3X6ocmtb25gbgHSjQpQ.png" /><figcaption>Picture 8 — source image Pablo Stanely</figcaption></figure><p>For all the image translation data-sets that I’ve seen used in articles, the translation object had similar proportions (e.g. horse-to-zebra, apple-to-orange, facades, city-maps). Perhaps image translation algorithms are not suited for this proportion fixing and another approach is needed here.</p><h4>4. Deriving proportions from photos</h4><p>Fixing proportion didn’t work. What if, the source for the proportions will be a picture? Posing for the camera is easy for the user. A picture is always in proportions and using <a href="https://github.com/CMU-Perceptual-Computing-Lab/openpose">OpenPose</a> the pose can be derived.</p><p>Pictures 9, 10, and 11 are from the training set and the pose skeleton was derived from an illustration. These pictures show that the machine is quite good at translating a pose into a drawing. It also exposes the limitation of this technique, the pose skeleton does not include information about the art style.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/791/1*JX7cZ6O2lWwzoPp8A1js-Q.png" /><figcaption>Picture 9 — source image ‘Freepick’ StorySet Amico</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/790/1*PyixWgYKgCDNEfjS-GwgLw.png" /><figcaption>Picture 10 — source image John D. Saunders</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/792/1*8Pve5J7VEBA0gqvEjRVmpw.png" /><figcaption>Picture 11 — source image unknown (scraped from the web)</figcaption></figure><p>In this training, I just used 250 paired images and with a larger training set, the machine illustrations will be crisper. Still, using photos in the validation set showed this approach has potential (picture 12).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HPc9arvKGBJvgPFDBCmzEw.png" /><figcaption>Picture 12 — photos of me turned into flat illustrations</figcaption></figure><h4>Another approach?</h4><p><a href="https://github.com/Yuheng-Li/MixNMatch">MixNmachGAN</a> released in 2020 is promising. The pose can be derived from photo’s and OpenPose, the illustration style from a particular artist, the color scheme from a color picker, and the background from a fourth source.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*W5D7KgY78PlqZClqhzjD5Q.jpeg" /><figcaption>Picture 13: Source <a href="https://github.com/Yuheng-Li">Yuheng Li</a>, <a href="http://krsingh.cs.ucdavis.edu/">Krishna Kumar Singh</a>, <a href="https://utkarshojha.github.io/">Utkarsh Ojha</a>, <a href="https://web.cs.ucdavis.edu/~yjlee/">Yong Jae Lee</a><br>UC Davis, published in CVPR 2020</figcaption></figure><h4>Anecdotes and some other things to improve</h4><p>The machine can’t make up facial features yet (picture 14) and doesn’t know a fire is red and yellow (picture 15). The fire, for example, can be fixed by adding a few more fire pictures to the training set. Similarly, a training set of drawings with outlined facial features can be used to create machine-made images with faces.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/807/1*2ufi3DRY7ZbIOm-hBKsEEA.png" /><figcaption>Picture 14 — source image ‘Freepick’ StorySet Pana</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/807/1*sk9KG0iuYAD42XXgKKDy3g.png" /><figcaption>Picture 15 — source image Katerina Limpitsouni</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fb9668ddc0ce" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/training-a-machine-to-draw-fb9668ddc0ce">Training a Machine to Draw</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Norway? The story of one unicorn]]></title>
            <link>https://medium.com/antlerglobal/why-norway-the-story-of-one-unicorn-32b5b2b04402?source=rss-405c12255563------2</link>
            <guid isPermaLink="false">https://medium.com/p/32b5b2b04402</guid>
            <category><![CDATA[antler-founders]]></category>
            <category><![CDATA[enterpreneurship]]></category>
            <category><![CDATA[norway]]></category>
            <dc:creator><![CDATA[Avi Latner]]></dc:creator>
            <pubDate>Mon, 30 Nov 2020 09:57:38 GMT</pubDate>
            <atom:updated>2020-12-04T11:47:23.398Z</atom:updated>
            <content:encoded><![CDATA[<h4>Why <a href="http://www.antler.co/oslo">Antler Norway</a> is especially appealing for founding a consumer-facing software startup</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*bUYbZvMbQw0z70if-GomqQ.png" /></figure><h4>My connection to Kahoot!</h4><p>This January I’m going to Oslo to join a VC backed accelerator - <a href="https://www.antler.co/oslo">Antler</a>, with the intention of founding a company. I’m from Israel, a country roughly the same size as Norway, that has more than twenty tech unicorns. Yet I’m going to an accelerator in Norway. A few people have asked me “why”, to which I answer that Antler has a unique program that’s not open in Israel (yet) and the managing team in Oslo seems to have created an atmosphere that fosters innovation. But there’s another reason and it has a lot to do with Norway’s first (and definitely not last) unicorn — <a href="https://kahoot.com/">Kahoot!</a>.</p><p>You see, Kahoot! (K from now on) is a poll and quiz collaboration tool and I was VP of product in another company, whose product is a poll and quiz collaboration tool as well. At parties and gatherings, when I talked about what I do some said “aha, like Kahoot!”</p><p>The companies were founded a year apart (2013, 2014) and in 2017 both companies raised round A and both had millions of users. We are in different industries, K is a tool for educators and our tool was primarily for publishers (it was also used by educators). Even so, when I ran market research and competitive analysis I also looked at K.</p><p>Fast forward three years to 2020 and K raised on a <a href="https://techcrunch.com/2020/10/13/kahoot-picks-up-215m-from-softbank-for-its-user-generated-gamified-e-learning-platform/">valuation of $2.4B</a>. What did we do? We worked really hard, we grew in users and in revenue and we won industry prizes. But we did not break-through. So I set out to examine why K is such a success and in the process came to question some of my past beliefs (that I think also others in tech and VC share).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MDqzb0FtDt1_-thQre-WYQ.png" /></figure><h4>Lessons learned: For every role of thumb, there’s an exception</h4><p><strong>It’s a hard industry.</strong> Sure education is a huge market, but it is centralized, slow to change, and schools (and even universities) don’t really spend on software. I knew that there’s an urgent need to innovate in education, but before Covid, I, and many others, viewed ed-tech as something like impact investment (you do it more for the greater good than for the money). Therefore, when you’re told ‘it’s a hard industry’ question it and look deeper.</p><p><strong>It’s ‘a nice to have’ not a must-have.</strong> In hindsight, we know that for many teachers collaboration and gamification aided by technology is not just ‘nice-to-have’ but an important method to increase student engagement. But at the time, I’d argue that most people and even most educators would have categorized it as something cute or a very small niche need. Edtech focused much more on online platforms for assignments, schedules, and student evaluation, on multi-media and on remote learning. This taught me, once again, that knowing what’s a burning need and what isn’t is elusive.</p><p><strong>Too simple to withstand competition.</strong> K is extremely simple. Content creators don’t have a lot of design or customization capabilities. In comparison, our product was much more sophisticated. We executed a vision of moving from ‘closed’ interactions to story-based interactions. We built a drag and drop creation platform with many more options. We thrilled our users with micro-interactions and animations. K managed without all of that because they were exceptional in understanding their users and in expanding their user base.</p><p>I often hear, “if it’s that easy to build, why aren’t competitors eating up the market share?” That’s certainly a valid question in K’s case. There’s some network effect in a collaboration tool and that’s part of the answer. But it takes just one teacher’s decision to move an entire class to a new tool. The main reason they are hard to replace is that they grew a cult. K became synonymous with classroom polls. The best brand won, not the technology, not the best product, and not even the network effect.</p><p><strong>Don’t serve two industries.</strong> Just 1% of K customers are paying customers (200K out of 17M) and that’s a low percentage compared to other freemium services. What’s more, half of the paying customers are businesses (and not schools) using K for training and presentations. Appealing to one industry for a broad user base and then trying to appeal to another industry for monetization sounds like a textbook mistake. Still, it’s been working reasonably well for K.</p><p>It would be a stretch to build a whole theory based on just one case, but one case is enough to disprove a theory. The frameworks we use to evaluate business ideas are flawed, we can’t fall into dogmas. Reality is much more complicated and murky than our frameworks and the only indicator we can really trust is users dedicating their time or money for our product.</p><p>Interested in hearing more about K? Checkout this <a href="https://soundcloud.com/antlervccast/avc10-the-kahoot-way-of-learning-with-eilert-hanoa">Antler podcast with K’s CEO, Eilert Giertsen Hanoa</a>.</p><p>Learn more about Antler Norway and apply at <a href="http://www.antler.co/oslo">www.antler.co/oslo</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=32b5b2b04402" width="1" height="1" alt=""><hr><p><a href="https://medium.com/antlerglobal/why-norway-the-story-of-one-unicorn-32b5b2b04402">Why Norway? The story of one unicorn</a> was originally published in <a href="https://medium.com/antlerglobal">Antler</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Perceptions of AI in the creative community]]></title>
            <link>https://uxdesign.cc/perceptions-of-ai-in-the-creative-community-83ac2dee9d07?source=rss-405c12255563------2</link>
            <guid isPermaLink="false">https://medium.com/p/83ac2dee9d07</guid>
            <category><![CDATA[future]]></category>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[creativity]]></category>
            <dc:creator><![CDATA[Avi Latner]]></dc:creator>
            <pubDate>Tue, 10 Nov 2020 12:02:04 GMT</pubDate>
            <atom:updated>2020-11-10T12:02:04.197Z</atom:updated>
            <content:encoded><![CDATA[<h4>AI moves from pattern detection to “patterned” creation and most designers are optimistic about this change.</h4><figure><img alt="Left side is a horse running in a meadow. Right side is the same picture, but the horse is now a zebra." src="https://cdn-images-1.medium.com/max/976/1*RGu0sApoxkAniu1w9tAjOg.png" /><figcaption>“Horse to Zebra”, <a href="https://www.cs.cmu.edu/~junyanz/">Jun-Yan Zhu</a> implementation of cycleGAN by UC Berkeley researchers 2017</figcaption></figure><h4>AI opening new creative options</h4><p>Ask an Uber driver what he thinks about self-driving cars and you’ll probably get a mix of denial, dismissal, and dread. That should not surprise, as the profession is projected to become obsolete when self-driving cars finally hit the road.</p><p>By contrast, when designers are asked about the impact Artificial Intelligence (AI) will have on the design community they are rather optimistic. In a survey I conducted in design groups on Reddit and Facebook, 52% of respondents think that advances in AI will open new creative options for designers and that in-large designers will benefit from this technology shift.</p><p>Only 12% of respondents expect that in an effort to cut costs, companies will use design bots in place of designers. The remaining respondents are divided between 19% that believed that AI design capabilities are too poor to have a serious effect, negative or positive, in the near future, and 17% that are not sure.</p><figure><img alt="A bar chart: Positive: 52%, negative 12%, on impact 19%, not sure 17%." src="https://cdn-images-1.medium.com/max/1024/1*ltmArrdOzbTnu3SYviQg0g.png" /><figcaption>Designers perception of the effect of AI on the design community</figcaption></figure><p>According to Bjourson and McAfee of the <a href="http://ide.mit.edu/">MIT center of digital business</a>, designers are right to be optimistic. They assert that while many job functions will be replaced, functions that require creativity, such as design, are likely to survive, transform, and flourish from the AI transformation. “The key to winning the race is not to compete against the machine but to compete with the machine.”</p><h4>The age of the creative machine</h4><p>It’s understandable why nearly a fifth of respondents think AI has little relevance to designers. AI seems to apply to routine and mundane tasks or to highly quantitative areas. Therefore, those designers are thinking ‘what does that have to do with creativity?’ But times are changing fast, from the area of process automation to a future area of a creative machine.</p><p>In the first wave of automation and robotics, developers and product designers had to fully understand a process and then had to code and instruct the machine to react to every scenario. This wave of automation replaced humans in jobs that are the most routine, such as factory workers on a conveyer belt. According to an <a href="http://resources.oxfordeconomics.com/">Oxford Economics report</a>, two million jobs in factories were lost to automation and twenty million more will be lost by 2030.</p><p>The second wave of automation is based on artificial intelligence. In AI techniques, the machine is fed a large data set and it autonomously recognizes patterns in the data and teaches itself new capabilities such as object detection. Developers and product designers, no longer have to teach the machine every possible scenario. This allowed automating tasks that are more diverse such as driving a car.</p><p>Most recently, the third wave of automation has come, the age of the creative machine. A subset of AI, called ‘Generative Adversary Networks (GAN) Machine Learning’ is not only learning patterns but it is also creating patterns. These GANs can be applied to the creation of images, text, video, or audio. Every media type is translated into bytes and the machine looks for patterns in the byte array. Therefore, GANs could potentially even make sculptures and physical devices if they’re fed 3D blueprints and then connected to a 3D printer.</p><p>These machine renderings can be looked at as a mere repetition of existing patterns that humans created. However, as a machine synthesizes thousands of inputs, they inevitably create variations that never existed before. If some of those variations are harmonious and novel aren’t they also creative?</p><h4>AI design tools: replicas, time savers, and creativity boosters</h4><p>The AI design tools out there, and those to come, can be put into three categories. I’ll review these categories at a high level here, and if you’re looking for a detailed review check <a href="https://uxdesign.cc/will-ai-take-over-ux-ea164a2ed39f">this UX Collective article</a> by Michal Malewicz.</p><p>The first, are tools for marketers, web developers, and small business owners that are meant to circumvent designers and save costs. These ‘replica’ tools include page layout designers and logo makers.</p><p>A second category is tools for designers that will enhance productivity and save time, freeing up more time for the creative process. Or tools that lower design costs giving designers more creative freedom. For example, tools that offer synthetic models (‘fake’ models) to be used as anchors. They are much cheaper than filming a video with a live model. For interaction designers that can’t afford a live production set, this opens up a new option.</p><p>A third category is tools that boost creativity. Before Wacom boards and electronic pens, we had a pencil and paper. Now we have both and we use both options. AI tools can be another way of creating. Existing examples are more in the realm of art than in the realm of design; I’m sure you’ve seen AI art, like the work of <a href="https://obvious-art.com/">obvious art collective</a>. But this trend could transcend from art to design.</p><p>Out of these three tool categories, I think time savers and creativity boosters will have a larger impact than replicas. So do most designers, but only time will tell.</p><h4>With power comes responsibility</h4><p>AI poses ethical challenges: Deep fakes that are obtained without consent and presented without transparency. Systems that, out of lack of awareness, are trained with racial and gender bias. Survalience powered by AI raises privacy concerns. Alongside developers and policymakers, designers play a key role in promoting the ethical usage of AI. Such is the work of the <a href="https://www.ajl.org/about">Algorithmic Justice League</a> an organization that combines art and research to illuminate the implications of potentially harmful AI. As designers who build AI systems or build with AI (as most of us will), we must remember our responsibility.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/94/0*U6huR18ROTIT2nA0" /><figcaption>The UX Collective donates US$1 for each article published in our platform. This story contributed to <a href="https://www.bayareablackdesigners.com/">Bay Area Black Designers</a>: a professional development community for Black people who are digital designers and researchers in the San Francisco Bay Area. By joining together in community, members share inspiration, connection, peer mentorship, professional development, resources, feedback, support, and resilience. Silence against systemic racism is not an option. Build the design community you believe in.</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=83ac2dee9d07" width="1" height="1" alt=""><hr><p><a href="https://uxdesign.cc/perceptions-of-ai-in-the-creative-community-83ac2dee9d07">Perceptions of AI in the creative community</a> was originally published in <a href="https://uxdesign.cc/">UX Collective</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Case in Point: Machine Learning is a Practical Tool for Designers]]></title>
            <link>https://avilatner.medium.com/case-in-point-machine-learning-is-a-practical-tool-for-designers-3f070ac0f888?source=rss-405c12255563------2</link>
            <guid isPermaLink="false">https://medium.com/p/3f070ac0f888</guid>
            <category><![CDATA[machine-generated-art]]></category>
            <category><![CDATA[sketch]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[synthetic-media]]></category>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[Avi Latner]]></dc:creator>
            <pubDate>Sun, 23 Aug 2020 18:54:22 GMT</pubDate>
            <atom:updated>2020-08-23T19:11:55.826Z</atom:updated>
            <content:encoded><![CDATA[<p>I set out to check if Machine Learning can become a standard tool in a designer’s tool-set. Can it be used not just for eccentric art but also for commercial use-cases such as illustrations for websites, landing pages, application flows, and blogs?</p><p><a href="https://medium.com/@avilatner/designers-getting-creative-with-machine-learning-imagery-349f2a3bf9d">In the first test that I ran</a>, I showed that a machine can automatically color illustrations. It stayed inside the line and recognized skin, clothes, and hair and used appropriate colors for each one. The second step was to see if a machine can be trained to take a simple stick figure drawing and turn it into a more intricate illustration.</p><p>That second step was much harder. But after many trials, I got it to work. The machine turned lines into sleeves, pants, and necks; Elipses into heads; Rectangles into skirts. Results can be improved, but this is enough to prove a point, that a simple stick figure drawing can be turned, in the click of a button, to a colorful illustration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*rXZrouStSUZh4BJOUkWR3Q.jpeg" /></figure><p>This opens endless design possibilities. A machine trained on a few different data sets can instantaneously turn one simple drawing into many illustrations, each with its own unique style inspired by a design trend or an illustrator. Then multiple figures can be automatically compiled into complete scenes. Eventually, designers will use their creativity to inspire a machine that will produce dozens of design concepts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aAsRHW8cxyOIgeO4zl1ZSw.png" /></figure><p>Are you a data scientist, developer, or designer and want to team up and build an image-generating product? Reach-out, check my Medium profile for contact info.</p><p><em>References to illustrators whose drawings I used here: </em><a href="https://twitter.com/ninalimpi?lang=en"><em>Katherina Limpitsouni</em></a><em>, </em><a href="https://www.johndsaunders.co/"><em>John D. Saunders</em></a><em>, </em><a href="https://www.pablostanley.com/"><em>Pablo Stanley</em></a><em>.</em></p><p><em>Reference to the algorithm implementation I used: </em><a href="https://github.com/affinelayer/pix2pix-tensorflow"><em>Pix2Pix by Christopher Hesse</em></a><em>.</em></p><p><em>Thanks to </em><a href="https://www.linkedin.com/in/zeevikal/"><em>Zeev Kalyuzhner</em></a><em> for ML advice and </em><a href="https://www.upwork.com/freelancers/~01305111fd677fa623"><em>Aya M</em></a><em> for help in creating the training data set.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3f070ac0f888" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Designers getting creative with Machine Learning Imagery]]></title>
            <link>https://avilatner.medium.com/designers-getting-creative-with-machine-learning-imagery-349f2a3bf9d?source=rss-405c12255563------2</link>
            <guid isPermaLink="false">https://medium.com/p/349f2a3bf9d</guid>
            <category><![CDATA[syntetic-media]]></category>
            <category><![CDATA[sketch]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[machine-generated-art]]></category>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[Avi Latner]]></dc:creator>
            <pubDate>Wed, 17 Jun 2020 20:02:32 GMT</pubDate>
            <atom:updated>2020-06-29T11:47:04.483Z</atom:updated>
            <content:encoded><![CDATA[<h3>Designers getting creative with machine learning imagery</h3><p>Designers already use machine learning on a daily basis without paying much attention to it, since machine learning is embedded into common features. For example, smart lasso cropping in Photoshop uses edge detection ML, and autofocus on a camera uses object detection ML. But can designers use ML not just in a hidden manner but rather with intention? And beyond using it for mundane tasks, can they actually use it to enhance creativity? This is a question that I started exploring.</p><p><strong>How I got started.</strong> The interest in new technology has always been part of what I do. The interest in art, however, is something that came back to me in recent years after becoming a father. I started drawing for my kids, and it reminded me of the joy I had drawing, back when I was a kid. I got yet another reminder of what drawing means to me when I looked at my daughter&#39;s classroom notebooks, they were full of scribbles. That’s what I did in class instead of listening: Imagining (even daydreaming) and drawing. They are both essential to sparking creativity. On top of that, I also started working with designers a lot more and I built a product for visual storytelling. All of that has led me to explore this question.</p><p><strong>Creative machine an oxymoron?</strong> Machine learning art is limited to what it is trained on. Therefore it can only create a revision of art that already exists. To make it work well, the training set needs to be homogenous enough so that the machine finds patterns, with some variety so that it can also handle a robust set of inputs. Creativity comes, when existing elements are put together into a new combination that’s unexpected and yet harmonious. Actually, that’s how creativity works for humans too, but machines can churn through combinations a lot faster.</p><p><strong>First trials.</strong> Knowing that success lays in the training set I started with the relatively simple task of making the machine color sketches the way a designer would. I created a hundred and fifty ‘flat modern style’ people renditions, that are popular on websites and marketing landing pages, using ‘<a href="https://www.humaaans.com/">Humaaans’ mix and match illustration library</a>. I then used an <a href="https://github.com/affinelayer/pix2pix-tensorflow">image-to-image translation algorithm</a> and ran it on the training set. I ran the first few rounds on my MacBook air. I quickly learned that I’ll need a much stronger machine to make it work, so I switched to a desktop with i7–9700 and Nvidia GTX1660 running Linux. Results are pretty good for a fast run. The program knows to color within the lines and to use appropriate colors (e.g. skin color for faces and hands).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qHi3TyYrKtgwcfSPXEJa-w.png" /></figure><p><strong>Next up. </strong>Just auto-coloring is not particularly useful for designers or creative. The next trial will check if the machine can turn quick scribbles into proportionate illustrations. Both steps together could be the beginning of a useful tool.</p><p>Designers out there, what do you think? Do you see yourself combining machine-generated images in your creative process? Comment or write to me at avilatner@gmail.com.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=349f2a3bf9d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Find Results Faster with Visual Search]]></title>
            <link>https://avilatner.medium.com/find-results-faster-with-visual-search-150a5576ebe3?source=rss-405c12255563------2</link>
            <guid isPermaLink="false">https://medium.com/p/150a5576ebe3</guid>
            <category><![CDATA[network-display]]></category>
            <category><![CDATA[recommendation-engine]]></category>
            <category><![CDATA[associative-search]]></category>
            <dc:creator><![CDATA[Avi Latner]]></dc:creator>
            <pubDate>Tue, 23 Feb 2016 19:20:12 GMT</pubDate>
            <atom:updated>2016-02-24T10:30:09.214Z</atom:updated>
            <content:encoded><![CDATA[<p>Search engines and recommendation engines display results ranked by relevance, in a list or a thumbnail layout. That’s useful in many cases but sometimes it’s more useful to also explore the relation between results. Consider the human mind. When we try to remember something we hop between bits of memories, one memory leads to another, and using associative thinking we eventually find what we are looking for.</p><p>Here’s a joke about associative thinking. Two couples meet for dinner and while the women leave for the living room the men stay at dinning table and talk. “Hey” says one “I ate at a great restaurant the other day. What was the name? Something to do with flowers. Daisy, Blossom. No I got it.” Then he shouts to the living room: “Rose, what was the name of that restaurant?”</p><p>For that propose <a href="https://medium.com/@natali_kravits">Natali</a>, <a href="https://www.linkedin.com/in/roi-lipman-2a48065">Ro</a>i and I built <a href="http://www.klustersearch.com/">Kluster</a> a search tool for associative exploration. The name comes from a combination of cluster and K-means, a popular clustering algorithm that we used in an earlier version of the tool. With <a href="http://www.klustersearch.com/">Kluster</a> you can choose one of several paths of exploration, and based on the path you choose different results will reveal themselves.</p><p>Let’s say that you like the book Zen and the Art of Motorcycle Maintenance and you look for similar books in a regular book recommendation engine. It will show you other books about Zen and about motorcycles. It will also show you books about workshops and maintenance, philosophy, journeys and other classics from the seventies. But why did YOU like the book? I for one, don’t care much about zen and certainly not about motorcycles. I am interested in philosophy and journey books. You might have another preference. Whatever your taste, <a href="http://www.klustersearch.com/">Kluster</a> separates results into automatic categories, allowing you to focus on what interests you the most.</p><p>1. Here I’m starting a search with <a href="http://www.klustersearch.com/">Kluster</a> with Zen and the Art of Motorcycle Maintenance, and I get a few recommendations:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*l-KpWDNAwPoTb-VM.png" /></figure><p>2. Based on my interests, I might ask to see more results like Shop class at Soulcraft:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*kli3dBacbPoFURPK.png" /></figure><p>3. Or maybe look into On the Road and see more journey books:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*70hoqoydbyNqZseb.png" /></figure><p>4. If I’m into spirituality then I’ll look at Siddhartha:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*dq6q5NG9xXs2ZuE3.png" /></figure><p>Check it out at <a href="http://www.klustersearch.com/">http://www.klustersearch.com</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=150a5576ebe3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Publishers And Ad-Tech Companies Can Stem Ad Blocking]]></title>
            <link>https://avilatner.medium.com/how-publishers-and-ad-tech-companies-can-stem-ad-blocking-a62455b2c8e6?source=rss-405c12255563------2</link>
            <guid isPermaLink="false">https://medium.com/p/a62455b2c8e6</guid>
            <category><![CDATA[digital]]></category>
            <category><![CDATA[publishing]]></category>
            <category><![CDATA[adtech]]></category>
            <dc:creator><![CDATA[Avi Latner]]></dc:creator>
            <pubDate>Wed, 13 Jan 2016 15:09:15 GMT</pubDate>
            <atom:updated>2016-01-13T15:11:18.481Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Originally published in </em><a href="http://www.mediapost.com/publications/article/265948/how-publishers-and-ad-tech-companies-can-stem-ad-b.html"><em>MediaPost</em></a></p><p>With the threat of ad blocking looming ominously over both publishers and ad-tech companies, it would be an easy way out to pin the blame on Flash or rich media — the very tools whose ad revenues have long allowed publishers to thrive.</p><p>Although this easy scapegoat exists, the ad-tech industry seems to understand that the real key to maintain the advertising monetization model is improving the user experience — doing something along the lines of the new <a href="http://www.iab.com/news/lean/">L.E.A.N.</a> approach by the Interactive Advertising Bureau.</p><p>But the truth is, there’s a silent technical factor still ruining the online user experience, and it is <em>client-side bidding.</em></p><p>Client-side bidding is painful for users, since browsers by their very nature are limited (and even more so on mobile) and can only output several http requests at a time. So generating a lot of simultaneous calls results in blocking browsers’ ability to render Web pages smoothly. Furthermore, loading big code objects (SWF or JS) to perform the complex logic of auctioning a placement is also something odd to do on the client side.</p><p>It may seem surprising to talk about client-side bidding now, since it seems as though RTB has been with us for a while and that all programmatic transactions are already server-side RTB. However, in the video ad delivery chain, client-side integration is still very prevalent.</p><p>Client side offers some advantages including pre-bid transparency: therefore, video exchanges are more likely to be RTB-integrated with their demand partners than with their supply partners.</p><p>Client-side bidding needs to change. And, luckily the tides are turning. User experience should be everyone’s top concern and pre-bid transparency can be mitigated with post-bid validation.</p><p>Here are some steps to take toward that end:</p><p>1. Publishers must improve their programmatic footprint monitoring. They can no longer take for granted the ad they’ve just presented their precious user is a well-behaved one, so they must verify that each and every ad is not disrupting the user experience. Using tools such as RISKIQ, GeoEdge and ClarityAd, publishers must monitor what type of ads run in their ad placements (e.g. display, rich-media, in banner video) and check if they adhere to the site’s policy (e.g. sound on click, sound on hover, autoplay).</p><p>2. The savviest publishers have already implemented the first step. That’s a good start, but it’s not sufficient. Next, publishers have to start measuring ads by CPU consumption, number of HTTP ad calls they generate, file sizes and the effect an ad has on browser’s frame rate. These metrics are directly related to user experience.</p><p>Beyond a certain threshold (too many ad calls, files that are too large) Web pages become less responsive to user action such as scrolling (the technical term for this is jank). This is a difficult task. User experience depends on many factors such as bandwidth and parallel tabs and processes, not just on the ad. Therefore, research needs to be done to establish the thresholds beyond which ads are likely to interfere with user experience. Publishers can’t do it alone; the ad-tech industry has to come up with the next generation of ad-monitor tools.</p><p>3. As noted, it’s not solely up to publishers. Video ad-tech companies have to adjust as well. If a company is using client-side bidding it must limit the number of parallel client side calls and rely more on sequential call methods. With an investment in machine learning and predictive analytics, sequential methods can be almost as efficient as a parallel auction.</p><p>4. Finally, video ad-tech companies have to steer away from client-side calls altogether. The main benefit of client-side calls is that the ad vendor can check the ad placement prior to bidding. From within the client, the ad vendor can verify the domain, the placement or player dimensions and make a real-time call to a fraud vendor. Instead, ad vendors will have to rely on the information passed on to the RTB ad request. Then, after the vendor has won the bid and the ad loads on the page, the ad-tech vendor makes sure the information passed was true and accurate.</p><p>Some ad-tech vendors are reluctant to make this change right now, as our industry is fraught with fraud. However, as we advance in the fight against fraud, we will also be able to start minimizing our impact on the user experience.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a62455b2c8e6" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>