<?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 Tony Chu on Medium]]></title>
        <description><![CDATA[Stories by Tony Chu on Medium]]></description>
        <link>https://medium.com/@tonyhschu?source=rss-dd52dd16f023------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*Zaq5Xl_izokyuPzq.jpg</url>
            <title>Stories by Tony Chu on Medium</title>
            <link>https://medium.com/@tonyhschu?source=rss-dd52dd16f023------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 13:18:06 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@tonyhschu/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[The Case for Design Technology at Noodle.ai]]></title>
            <link>https://medium.com/@tonyhschu/the-case-for-design-technology-at-noodle-ai-e46bfa5d5e13?source=rss-dd52dd16f023------2</link>
            <guid isPermaLink="false">https://medium.com/p/e46bfa5d5e13</guid>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[ui-engineering]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[Tony Chu]]></dc:creator>
            <pubDate>Wed, 24 Jan 2018 01:44:59 GMT</pubDate>
            <atom:updated>2018-01-24T01:45:15.669Z</atom:updated>
            <content:encoded><![CDATA[<p>I am looking for UI engineers and data-oriented designers to come work with me on the Design Technology team at <a href="http://noodle.ai/">Noodle.ai</a>, so I wanted to articulate why I am excited about the work we are doing here. I tend to “framework” my way through my decisions, so I will frame my motivation in three parts — the desire for autonomy, mastery, and purpose. (This is loosely based on <a href="https://www.amazon.com/Drive-Surprising-Truth-About-Motivates/dp/1594484805">Daniel Pink’s book, <em>Drive</em></a>)</p><h3>Autonomy, Mastery, Purpose</h3><p>Let’s start with purpose — Noodle.ai is in the business of building Enterprise AI. AI and machine learning are very “buzzy” right now — what are we actually doing?</p><p>AI and machine learning can help enterprises make better business decisions, e.g. optimize prices, reduce waste, etc. Let’s take an example from the tire industry*. The world produces more than a billion tires a year, a good chunk of which are not sold or used. Why? Because companies don’t have good estimates on how much they will sell, so they always produce/buy a bit extra for safety. How much extra? Well, that depends on how good you are at forecasting demand.</p><p>We see glitzy headlines about machine learning all the time, but I believe some of the most significant impact of machine learning will be in behind-the-scenes scenarios like supply chains and manufacturing. A 10% more accurate estimate of sales might mean 3% fewer tires produced, which leads to fewer acres of rainforests chopped down for rubber trees. Of course, it will also mean millions of dollars saved. Not glamorous, but high impact.</p><h3>Why do design/UI at a machine learning company?</h3><p>…and why should the CEO of a big tire manufacturer pay attention to the outputs of a machine learning model? Sure, at Noodle we have really excellent stats and math PhD’s crunching the numbers, but our clients have a million dollars of revenue on the line, and we are a two-year-old company. Why should they trust us?</p><p>This is where the Design Technology team comes in. The most accurate model in the world does no good if the decision makers don’t trust it. To help our customers trust our data science work, our design technology team applies the best of communication design, data visualization and UI engineering to bring the data science work to life. Our team’s goal is to help our customers see dynamics in their business that they have never seen so clearly before.</p><p>The most satisfying part for me in all this is we can see, in a matter of months, the difference that design makes. One of our clients operates a fleet of private jets – our UI enabled them to make decisions that used to take a whole day in a matter of minutes. According to the client, he has that UI open in a tab all day now, and their business practice has changed based on the insights on display.</p><h3>Aspiring to Mastery in Data Vis + Machine Learning</h3><p>When our CEO Steve reached out to me, it was in response to <a href="http://r2d3.us/">R2D3</a>, a project I had collaborated on to explain machine learning to a non-technical audience. He said, “we need to make enterprise AI as clear as R2D3.” In Steve’s vision for Noodle, I saw an opportunity to hone my data visualization design practice, and to learn how machine learning can be practically applied to real world business problems.</p><p>This past year our team has shipped several different UI’s that have transformed how our clients see and think about their business, and we’re just getting started. I’ve been exposed to so many new domains, both in terms of business and in data science. While I cannot promise mastery, I do know that mastery can only be obtained by doing a wide variety of interesting work — and we have plenty of interesting work here. There’s nothing quite like learning about how a business works by being asked to visualize the business.</p><h3>Early Birds get the Worm, Early Employees get Autonomy</h3><p>So maybe you’re convinced that doing design+UI work for a machine learning company is interesting. Why not do that at a Google or a Facebook? In my career, I’ve always leaned towards smaller companies. As an early employee at a small company, you generally have a much better chance of picking your own approach to problems. Frankly, you also have a better chance of making your own mistakes and learning from them. No one can tell you, “this is how it’s always been done” because it simply has never been done before. This is one of the things that get me out of bed every morning.</p><h3>Sounds Interesting? Come chat</h3><p>Hopefully this gives you a sense of what what we’re up to. If you are someone who’s interested in learning from and with your peers, if you like the idea of exploring data science through the lens of UI and data vis, if you like working at the intersection of different disciplines — come chat! You can reach me at <a href="/service/mailto:tony.chu@noodle.ai">tony.chu@noodle.ai</a></p><p>Tony Chu<br>Principal Designer at Noodle.ai</p><p>p.s. if you’re curious about what design + machine learning might look like, check out <a href="http://www.r2d3.us/visual-intro-to-machine-learning-part-1/">R2D3</a> — I want to make more things like this at Noodle.ai</p><p>* We work with one of the biggest tire distributors in the world.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e46bfa5d5e13" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Supplement to “Design in a World Where Machines are Learning”]]></title>
            <link>https://medium.com/@tonyhschu/supplement-to-design-in-a-world-where-machines-are-learning-af23b6095e8c?source=rss-dd52dd16f023------2</link>
            <guid isPermaLink="false">https://medium.com/p/af23b6095e8c</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Tony Chu]]></dc:creator>
            <pubDate>Tue, 07 Feb 2017 19:32:52 GMT</pubDate>
            <atom:updated>2017-02-07T19:32:52.533Z</atom:updated>
            <content:encoded><![CDATA[<p>So I gave a talk at <a href="http://interaction17.ixda.org/schedule/#feb-7">Interaction17</a> on what I thought designers should know about machine learning from a “how do I use this?” perspective.</p><p>Collected here are some additional notes I wish I could have covered, in three broad categories.</p><ol><li>Technical Foundations of Machine Learning</li><li>Business Implications of Machine Learning</li><li>Tools for Applying Machine Learning Today</li></ol><h4>Technical Foundations of Machine Learning</h4><ul><li>“<a href="http://www.r2d3.us/visual-intro-to-machine-learning-part-1/">A Visual Introduction to Machine Learning</a>” by Stephanie Yee &amp; Tony Chu (yes, this is a shameless plug.) A data visualization piece that serves as an intro to decision trees, a particular type of machine learning model.</li><li>“<a href="https://medium.com/@ageitgey/machine-learning-is-fun-80ea3ec3c471">Machine Learning is Fun</a>” by Adam Geitgey. A six part series introducing various applications of machine learning methods. I particularly liked it’s summary, “Machine Learning is using generic algorithms to tell you something interesting about your data without writing any code specific to the problem you are solving”</li><li>“<a href="http://playground.tensorflow.org/">Tensorflow Playground</a>” from the Google Brain team. A toy example of another type of machine learning model i.e. a neural network.</li><li>“<a href="https://hbr.org/2015/07/what-every-manager-should-know-about-machine-learning">What Every Manager Should Know About Machine Learning</a>” by Mike Yeomans in HBR. It has a good breakdown of the concepts required for applying machine learning, e.g. feature extraction and cross validation.</li></ul><h4>Business Implication of Machine Learning</h4><ul><li>“<a href="https://stratechery.com/2015/tensorflow-and-monetizing-intellectual-property/">TensorFlow and Monetizing Intellectual Property</a>” by Ben Thompson. The most interesting argument here is that… if Google can open source TensorFlow, it’s machine learning package, it means that Google believes it’s competitive advantage lies elsewhere. (Hint: it’s in its data.)</li><li>“<a href="https://medium.freecodecamp.com/the-business-implications-of-machine-learning-11480b99184d">Business Implications of Machine Learning</a>” by Drew Breunig, essentially arguing that getting training data critical mass is the new network effect. If a company tries to compete with the big’s (Google, Facebook, etc) in collecting data for ML, it will fail. To compete, companies must collect data that the big players are not collecting.</li></ul><h4>Tools for Applying Machine Learning Today</h4><ul><li>Speech API’s ~&gt; <a href="https://cloud.google.com/speech/">Google</a>, <a href="https://www.microsoft.com/cognitive-services/en-us/speech-api">Microsoft</a>,</li><li>Vision API’s ~&gt; <a href="https://www.clarifai.com/api">Clarifai</a>, <a href="https://www.ibm.com/watson/developercloud/visual-recognition.html">IBM Watson Visual Recognition</a></li><li>Machine Learning as a Service API’s ~&gt; <a href="https://aws.amazon.com/machine-learning/details/">Amazon Machine Learning</a>, <a href="https://azure.microsoft.com/en-us/services/machine-learning/">Microsoft Azure ML,</a> <a href="https://cloud.google.com/products/machine-learning/">Google Prediction Services</a></li><li>… more to follow.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=af23b6095e8c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What can designers do?]]></title>
            <link>https://medium.com/@tonyhschu/what-can-designers-do-c1642b406573?source=rss-dd52dd16f023------2</link>
            <guid isPermaLink="false">https://medium.com/p/c1642b406573</guid>
            <dc:creator><![CDATA[Tony Chu]]></dc:creator>
            <pubDate>Thu, 10 Nov 2016 16:32:57 GMT</pubDate>
            <atom:updated>2016-11-10T16:32:57.641Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>TL:DR;</strong> I was thinking about communication design in context of politics last night.</p><p>I spent some of today talking to friends in my little bubble. Cosmopolitan friends, working in design and data vis. Folks are distraught. I am distraught.</p><p>“Why didn’t I do more?”</p><p>Last night in a moment of determination I vowed that I will join whichever campaign the Democratic party launches in 2020 once the primaries are over. I really do think I want to do that, but what can a designer really do?</p><p>I reflected on that some more, and I have two theories of change.</p><h3>Turn Out</h3><p>One framing is that elections are fundamentally about turn out. i.e.</p><h3>Domenico Montanaro on Twitter</h3><p>Turnout 2012 Obama: 65.9m 2016 Clinton: 59.1m = -6.8m 2012 Romney: 60.9m 2016 Trump: 59m = -1.9m You tell me what happened</p><p>If you can get people who believe the same things you believe to show up at the polls, you win. (Or suppress the other side, but that’s the dark arts and we don’t do that.) In this framing, turn out is driven by enthusiasm, and enthusiasm grows with emotional resonance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/465/1*x2bt7zKehAW0h1xvt2RaDQ.jpeg" /></figure><p>That’s what good design can do.</p><p>This particular theory of change calls for designing a resonant, irresistibly designed artifact that energizes people to show up, and bring their friends to the polling place. This can work, and it’s certainly seems like the obvious thing to reach for. Indeed, <a href="http://letsfreecongress.org/">I’ve tried</a>.</p><p>Yet this is still a fundamentally adversarial framing. It’s about getting more of “us” out than “them.” I wanted to push myself to see if there’s a different way.</p><h3>Pedagogy</h3><p>Even as I write this I wonder if I’m being naive, nonetheless this is my theory. Our society lacks basic shared understanding of the challenges that we face. How can we agree on solutions when we cannot even settle on the premises of the problem?</p><p>The problems that we face are so complex, so intertwined– how can we possibly get to a shared understanding? I don’t know, but my hunch is that we tease the underlying systems apart, and help people piece it back together. And maybe this is my hubris, <a href="http://www.r2d3.us/visual-intro-to-machine-learning-part-1/">I feel like I’ve done it once</a>, perhaps I can try to do it again.</p><p>I want to try this, foolish as it may be, because the upside is enormous. Over and over I’ve heard from people who write us, “you made this understandable, and now I can apply it to do &lt;what-ever-it-is-that-they-do&gt;” and that is amazing. The joy that people derive when they feel like something finally click is palpable. Design can do that.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZZiWJncAmGkMd6fX5NRLQQ.png" /><figcaption>Nicky showed us the way with the <a href="http://ncase.me/polygons/">Parable of the Polygons</a> demonstrating how segregation emerges.</figcaption></figure><p>What if we can apply that kind of design to our societal disagreements? Our policy challenges?</p><p>To me this is one part teaching systems thinking, one part graphic design, and one part interactive storytelling. I want to strive for clarity, not simplicity. I want to carefully show the nuances, and laying bare choices that are difficult. Most of all, I want this to form the basis for discussion.</p><p>Do you think that’s possible? Would you help if we tried?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c1642b406573" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Startups, Entrenchment, and the Adjacent Possible]]></title>
            <link>https://medium.com/i-m-h-o/startups-entrenchment-and-the-adjacent-possible-ee379ec50f55?source=rss-dd52dd16f023------2</link>
            <guid isPermaLink="false">https://medium.com/p/ee379ec50f55</guid>
            <dc:creator><![CDATA[Tony Chu]]></dc:creator>
            <pubDate>Sun, 07 Jul 2013 04:32:21 GMT</pubDate>
            <atom:updated>2013-07-07T18:41:08.167Z</atom:updated>
            <content:encoded><![CDATA[<h4>Half-baked Philosophy of Startups Ahead</h4><p>A while back Paul Graham wrote an essay asserting that <a href="http://www.paulgraham.com/growth.html"><em>Startup = Growth</em></a><em>. </em>It stuck in my mind because of how vehemently <a href="https://twitter.com/dhh/status/249550225730052096">David Heinemeier Hannson of 37signals disagreed with him</a>.</p><p>I felt both of them were right, and couldn’t figure out why. Recently, when thinking about the idea of adjacent possibles, I realized why I was so torn between PG’s assertion and DHH’s rebuke. My hypothesis: <em>Startup = Growth</em> applies only to a certain class of opportunities — opportunities that are in new adjacent possible spaces.</p><p><strong>Seeing the Adjacent Possible</strong></p><p>In PG’s essay he cites Google and Apple as prime examples of startups that won by growing rapidly, fitting his assertion that Startup = Growth. Throughout his examples I saw the theme of adjacent possibles everywhere.</p><blockquote>That space of ideas has been so thoroughly picked over that a startup generally has to work on something everyone else has overlooked. I was going to write that one has to make a conscious effort to find ideas everyone else has overlooked. But that’s not how most startups get started. Usually successful startups happen because the founders are sufficiently different from other people that ideas few others can see seem obvious to them. Perhaps later they step back and notice they’ve found an idea in everyone else’s blind spot, and from that point make a deliberate effort to stay there. [<a href="http://www.paulgraham.com/growth.html#f3n">3</a>] But at the moment when successful startups get started, much of the innovation is unconscious.</blockquote><blockquote><a href="http://www.paulgraham.com/growth.html">Startup = Growth, Paul Graham</a></blockquote><p>The only part I’d disagree with PG here is the idea of <em>blindspots. </em>What the founders of successful startups notice before other people are brand new territories in space of ideas—the recently opened up adjacent possibles. (A notion popularized in Steven Johnson’s Where Good Ideas Come From. More on this in a minute.) It is not an accident that they saw these spaces. It requires them having worked right on the cutting edge, which is the only vantage point where one can see the adjacent possibles.</p><p><strong>Land Grab In the New Adjacent Possibles</strong></p><p>In Steven Johnson’s book, Where Good Ideas Come From, he talks about how new ideas are rarely truly new, but are assembled from spare parts. (i.e. ideas that are already available) The people who are working on the cutting edge, who have access to all the pre-requisite ideas, become the conduits for new ideas to come into the world. It explains <a href="http://en.wikipedia.org/wiki/Multiple_discovery">why new ideas are often arrived at independently by a number of people in a short period of time</a>. It is almost as if ideas are inevitable.</p><p>Returning to startup land, this means at any given moment, there are a few <em>inevitable*, world-beating </em>ideas lurking in the adjacent possible. If a startup believes it has struck the adjacent possible, its priority should be to get entrenched before anyone else notices. Rapid growth is the path to entrenchment. That is exactly what PG argues Google and Apple has done, and what every successful startup should do.</p><p><strong>First to Entrenchment ??? Success</strong></p><p>While startup that achieve first-to-entrenchment are probably all successes, not all successes require first-to-entrenchment. This is where I favor DHH’s argument. Identifying and colonizing adjacent possibles is not the only path to startup success. Growth, therefore, should not necessarily be a startup’s number one priority. 37signals and Mailchimp are examples of a class of businesses that have done well for themselves focusing on differentiation in existing (in PG’s terms, well-picked-over) idea spaces. I doubt anyone would argue with calling them successful businesses.</p><p>You could argue that PG wouldn’t consider them startups, but then that would dramatically narrow the definition of startups. That would be to say only ventures seeking to dominate an adjacent possible space are startups. Which you might say is PG’s point.</p><p>To which I say, fine. Let’s agree to disagree.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ee379ec50f55" width="1" height="1" alt=""><hr><p><a href="https://medium.com/i-m-h-o/startups-entrenchment-and-the-adjacent-possible-ee379ec50f55">Startups, Entrenchment, and the Adjacent Possible</a> was originally published in <a href="https://medium.com/i-m-h-o">I. M. H. O.</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>