<?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 Hypergiant on Medium]]></title>
        <description><![CDATA[Stories by Hypergiant on Medium]]></description>
        <link>https://medium.com/@hypergiant?source=rss-601d9e7f0ae------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*iuR2f2tTE8b2WvhnkYMmAw.png</url>
            <title>Stories by Hypergiant on Medium</title>
            <link>https://medium.com/@hypergiant?source=rss-601d9e7f0ae------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 21:30:36 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@hypergiant/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[Bringing Legacy Airline Planning Up to Speed with Machine Intelligence]]></title>
            <link>https://medium.com/@hypergiant/bringing-legacy-airline-planning-up-to-speed-with-machine-intelligence-6994a0bd46f8?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/6994a0bd46f8</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[manufacturing]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[aviation]]></category>
            <category><![CDATA[airlines]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Fri, 30 Nov 2018 14:30:19 GMT</pubDate>
            <atom:updated>2018-11-30T14:30:19.929Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*h4Lxs-SDV4Mfh8faql1p9A.png" /></figure><p><em>Written by Ross A. McIntyre, VP of Strategy &amp; Alexander Tronche, Senior Strategist</em></p><p>It was not too long ago that airlines lacked online booking capabilities. Think about it: the shift was one that drastically altered the entire user experience in what some might consider to be a rather quick and easy way. On the front end, that is what it was — an addition to a long list of website functionality that simply made the booking process more accessible and digitally familiar. On the back end, however, that transformation was anything but facile. Legacy airline planning, along with the employee and consumer interactions that it affects (i.e., practically all of them), has ingrained certain procedures so far into the core of the industry that affecting change can truly feel impossible.</p><p>We, of course, know that is not true and the online booking example has proven it. The integration of intelligence as a technology into core, organizational structures can dramatically improve the health and stability of a manufacturer, lessor, creditor, or airline when user needs and experience design are taken into consideration. Luckily enough, the application of machine intelligence continues to quantify even the most intangible of goals — think customer satisfaction — long enough for drastic innovation to present a solid return on investment to those calling the shots.</p><h3>Reworking the Network</h3><p>Machines have been adopted and adapted for internal benefits for many years, making operations relatively hassle-free in comparison to previous decades, but experiences have not improved that much for travelers. With alternative options for their customers popping up left and right, airlines are the ones seen itching to spearhead the intelligence transformation. At the end of the day, their brands are customer-facing and on the front lines. They have the user data necessary to communicate current needs and predictable desires back to other members of the industry, thereby propelling airline planning as a whole into the future.</p><blockquote><em>“In 2026, </em><a href="https://www.oliverwyman.com/content/dam/oliver-wyman/global/en/2016/apr/NYC-MKT9202-001MRO-Survey-2016_web.pdf"><em>the global fleet will generate 98 exabytes</em></a><em> [million terabytes] of data.”</em></blockquote><p>The first step is identifying the problem. For example, network planners and their organizations (e.g., manufacturers, lessors, and airlines) must come to the collective realization that trying to make decisions 45 days before any flight is not only premature, but also bad for business. With marketplace fluctuations, weather anomalies, workforce inconsistencies, and maintenance surprises, establishing the terms for a flight too far in advance is counterintuitive and not in line with the aerospace business of the future. When compared with the agility offered by a data system well-versed in and better connected to the industry’s happenings in real time, manually establishing flight details sticks out like a retrofitted thumb.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MTCEmL-wh65D8FQs8PdY2A.png" /></figure><p>Indeed, technology is largely about revisiting the reasons for which things are done the way they are. A system of artificial benefits (reached by way of reductionism) is not the solution — key industry players must focus on progress in a way that is useful. Otherwise, legacy baggage will continue to hamper the acceleration brought on by the application of artificial intelligence (AI). Today’s network planning sets the stage for information technologists to reconsider their current dependence on Microsoft Excel, identify the alternative tools that they need, and build those in a contemporary way. If everyone is too busy patching an already-flawed system, then it becomes increasingly difficult to do something exponentially different. Machine intelligence offers a different approach: it tackles multiple problems simultaneously by rethinking their foundational source and reimagining the structure of that core to be more agile in the face of its own evolution.</p><p>The combinatory effects of manufacturers, lessors, and airlines contemporizing certain aspects of their businesses are enabled by the AI opportunity. The data exists and the computational technology is accessible, but the conversation is what needs to be had in order to avoid dulling the benefits of future developments. Across the organizations of all three players, there is something to be said about exploring the long-term benefits of machine intelligence and watching them outweigh any short-term inconveniences. The potential incremental value of AI over other analytical techniques in the travel category, for example, <a href="https://www.mckinsey.com/featured-insights/artificial-intelligence/the-promise-and-challenge-of-the-age-of-artificial-intelligence">is said to have reached 128 percent</a>. Meanwhile, that same metric stands at 30 percent for the aerospace and defense industry.</p><p>Rather than tacking on new solutions for each individual issue that arises — much like the dreaded technology stack that many businesses find themselves battling — a horizontal approach, empowered by AI, can be established across each organization and, subsequently, the industry as a whole. If all participatory groups incorporate data engineering and AI into their business models, the positive effects are felt across the aerospace industry for both those working in it and those depending on it.</p><h3>Operationally Speaking</h3><p>For those working in the aerospace industry, the viability of one’s current product portfolio and its continuous improvement is a direct reflection of the organization’s longevity and survival. But when usage is much higher than anticipated and data grows at an unforeseen rate, it becomes evident rather quickly that physical workforces are limited in their ability to analyze, maintain, and predict. Were employees capable of manually handling the amount of information available today, the sheer amount of minds required to comprehend it would make for strict processes rather than the malleable workflows resulting from custom, AI-enabled platforms.</p><blockquote><em>“When it comes to operational issues it is really important to be precise, but also have the ability to improvise.”</em></blockquote><blockquote><em>— </em><a href="https://www.eyefortravel.com/revenue-and-data-management/making-flying-profitable-yes-network-planning-can-help"><em>Pieter Groeneveld</em></a><em>, Senior VP of Digital at Air France-KLM</em></blockquote><p><a href="http://transcripts.hypergiant.com/machine-intelligence-hypergiant-edition-0">Human efficiencies are amplified</a> by the near-infinite capacity that an intelligent machine offers. Upon establishing its knowledge base and training it in the language (both verbal and visual) expected by its users, a strong piece of machine intelligence can effectively bridge the gap between large amounts of untapped data and an organization’s top-level decision-makers. In aerospace, this is where AI-enabled predictive maintenance and data mining come in. Both are introduced in situations where data and business value are analogous — where the ability to act quickly across one’s product portfolio can mean the difference between making it and breaking it.</p><p><em>Download and read our guide, entitled </em><a href="http://transcripts.hypergiant.com/airline-planning">Bringing Legacy Airline Planning Up to Speed with Machine Intelligence</a>,<em> to further understand the implementation of these AI-enabled techniques and how they ultimately better the customer experience. Any questions? Our authors can always be reached via the emails included in the PDF.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6994a0bd46f8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Reconstructing Primary Information from Secondary Information]]></title>
            <link>https://medium.com/@hypergiant/reconstructing-primary-information-from-secondary-information-173f4b5b7b36?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/173f4b5b7b36</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[graph-database]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[feature-engineering]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Wed, 21 Nov 2018 17:48:27 GMT</pubDate>
            <atom:updated>2018-11-21T17:48:27.361Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tE8B8Dd7TMtyeWmd9jr_cg.png" /></figure><p><em>Written by Drew J. Lipman, PhD, Lead Data Scientist at Hypergiant</em></p><p>Data scientists spend <a href="https://www.forbes.com/sites/gilpress/2016/03/23/data-preparation-most-time-consuming-least-enjoyable-data-science-task-survey-says/#76969b3d6f63">60% of their time cleaning and organizing data</a>.</p><p>As part of their work, they engage in something known as <a href="https://docs.microsoft.com/en-us/azure/machine-learning/team-data-science-process/create-features"><em>feature engineering</em></a>: the process of using existing, known information or features to extract new information or features about particular data.</p><p>A classic example is extracting the age or gender of Titanic’s passengers from their salutations. This information (age and gender) is, generally, not included in the passenger manifest, but it becomes very useful when trying to predict whether a particular passenger survived or calculate overall survival rates.</p><p>Feature engineering tends to be applied in two versions. One version is deemed <em>forward construction</em>, meaning that the extracted features are directly derived from the original information. In other words, all of the information needed for the new feature is contained in the original feature. (Extracting age from a date of birth is a clear example of this.)</p><p>The other version of feature engineering — and the focus of our exploration here — is <em>backwards construction</em>. This is formally known as an inverse problem, a case in which <a href="http://transcripts.hypergiant.com/reconstructing-primary-information-from-secondary-information-1">the features originally used to construct the given information are recovered</a>. For example, each passenger’s gender and age were used to establish the correct salutation for addressing them on the Titanic. The inverse problem is to take the resulting salutation — a form of <em>secondary information</em> — and compute the inputted age and gender — a form of <em>primary information</em> — from which it was derived.</p><h3>The Framework</h3><p>While the framework of an inverse problem is very general, different problems produce distinct difficulties. Some of the new, recommended tools for this form of feature engineering are graph databases or knowledge graphs. These can model how often datapoint features interact or how a single feature interacts across a database.</p><p>Graph databases (GDBs) can be modeled using vertex and arc-labeled <a href="https://en.wikipedia.org/wiki/Directed_graph">directed graphs (digraphs)</a>: more specifically, a primary digraph and a secondary digraph. The vertices (and their weights) of the secondary digraph are the same as those of the primary digraph, while the arcs (and their weights) of the secondary digraph are <em>derived</em> from those of the primary digraph. Using this language, the feature engineering problem is described as follows: Given the secondary digraph — and, potentially, parts of the primary digraph — extract useful information and features about the primary digraph that were previously unknown.</p><p>This framework opens several questions, not least of which is determining when it is feasible to extract a particular feature: arcweights or the arcs, themselves. Some variations of this problem (all of which can be extended to include partial, primary information) are:</p><ul><li>What is the minimum amount of secondary digraph features needed to extract a particular feature?</li><li>What features can be extracted given a set of secondary features?</li><li>How dependent are the resulting methods on the topologies of the chosen GDBs?</li><li>Where complete information about a feature cannot be extracted, can useful, partial information be extracted instead?</li></ul><h3>Example Problems</h3><p>The primary information here is the Parents Graph. That is, we are given a set of people going back several generations, and we have an arc from a vertex <em>x</em> to another vertex <em>y</em>, which will be denoted as <em>(x,y)</em>, where x is a parent of <em>y</em>. The derived information is the Cousins Graph, which is a graph with an edge <em>(x,y)</em> if <em>x</em> and <em>y</em> are first cousins. Equivalently, there is a path in the Parents Graph with arcs <em>(a,x),(b,a),(b,c),(c,y)</em>. That is, there is an edge in the Cousins Graph if and only if <em>x</em> and <em>y </em>have a common grandparent, <em>b</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*3_iAssngEbhdEr_7" /><figcaption><em>An example of a Parents Graph and its associated Cousins Graph. Note that </em>x<em> and </em>y<em> are cousins if and only if they share a grandparent. In this case, they share</em> b<em>. Note that the relation between </em>x<em> and </em>y<em> is symmetric and, therefore, lacks arrows.</em></figcaption></figure><p>The second example problem has, as primary information, the Maternal Graph. This is similar to the Parent Graph, but now, each vertex has at most one parent: their mother. The secondary information is the Maternal Cousins Graph, where there are arcs <em>(a,x),(b,a),(b,c),(c,y)</em> in the Maternal Graph. That is, the Maternal Cousins Graph has an edge <em>(x,y)</em> if and only if <em>x </em>and <em>y</em> have the same maternal grandmother, <em>b</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*tg5Bm7L0ayvHXCqi" /><figcaption><em>An example of a Maternal Graph and its Cousins Graph. Note that </em>x<em> and </em>y<em> are cousins if and only if they share the same maternal grandmother, </em>b<em>. Note that the relation between </em>x<em> and </em>y<em> is symmetric and, therefore, lacks arrows.</em></figcaption></figure><p>The third example problem has, as primary information, the Handshake Graph — a timestamped graph of handshakes among a collection of people that has an edge between <em>x</em> and <em>y</em> with weight <em>t</em> if <em>x</em> and <em>y</em> shook hands at time <em>t</em>. The secondary information is known as the Influence Digraph, which has an arc <em>(x,y)</em> if there is a path (of arbitrary length) connecting <em>x</em> to <em>y</em>, where the weights (timestamps) are increasing for each edge. For simplicity, we make the assumption that no person is shaking two hands at the same time. Note that the Influence Digraph can be used to model potential pathogen spread across a group of people: an arc <em>(x,y)</em> indicates that <em>y</em> could be infected by a pathogen originating at <em>x</em>. The lack of such an arc indicates the impossibility of a contact pathogen spreading from <em>x</em> to <em>y</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*mBxnYQO2WBy53Jsp" /><figcaption><em>An example of a timestamped Handshake Graph and its Influence Digraph. Note that the arc </em>(x,z)<em> exists since there is a path </em>xy,yz<em> with increasing timestamps </em>1 <em>and </em>2<em> that allows </em>x<em> to influence </em>z<em>.</em></figcaption></figure><p>Across all of these examples, it is assumed that a single piece of information is missing: an arc from the Parents Graph, an arc from the Maternal Graph, and a timestamp from the Handshake Graph. It is also assumed that all of the information from the Cousins Graph, Maternal Cousins Graph, and Influence Digraph is known. Note that there is also a graph topology implied by the structure; namely, the assumption that each vertex on the Maternal Graph has at most one arc coming in.</p><p><em>For more of our thoughts on graph theory, along with an exploration of uniqueness and probabilistic approaches, download our guide on </em><a href="http://transcripts.hypergiant.com/reconstructing-primary-information-from-secondary-information-1"><em>Reconstructing Primary Information from Secondary Information</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=173f4b5b7b36" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Is So Special about Computer Vision?]]></title>
            <link>https://medium.com/@hypergiant/what-is-so-special-about-computer-vision-c4b3f5780106?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/c4b3f5780106</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Thu, 15 Nov 2018 21:44:14 GMT</pubDate>
            <atom:updated>2018-12-13T16:29:50.489Z</atom:updated>
            <content:encoded><![CDATA[<h3>What Is So Special About Computer Vision?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fp9OPVaSwl9fWd1kvLITQQ.png" /></figure><p><em>Written By Applied Sciences Team at Hypergiant</em></p><p>As intelligence becomes an increasingly important technology and machines are better equipped to comprehend their surroundings, the ability of a machine to see what humans cannot becomes paramount to the success of artificial-human integration. This translates directly into core performance indicators such as customer satisfaction, revenue growth, cost and waste reduction, and overall optimization.</p><p>In addition, <a href="https://transcripts.hypergiant.com/ai-computer">investing in computer vision is critical to the future success of all sectors</a> as some contemporary technologies — even those that came with high hopes — reveal their inability to truly communicate not only with each other, but with established organizational processes and needs. The good news is that, with artificial intelligence flowing through its veins, computer vision can now gather the data needed to both automate the mundane and empower the innovative.</p><h3>There is no great prior technology that compares to computer vision.</h3><p>The likes of light sensors and ultrasonic sound are certainly capabilities that we depend on, but they lack context. They are specific pieces of hardware or software built to analyze particular measurements of reality just as the body has separate apparatuses for hearing, touching, smelling, and tasting. In other words, the data being gathered is siloed with no system for seeing and deciphering what it means in concert.</p><p>The Internet of Things was, for many years, touted as the solution to this disconnection. It was meant to take the multitude of gadgets and technologies inside our homes, for example, and transform them into a cohesive structure for seamless management. Rather quickly, however, the realities of the intricacy of our networks and systems for storing information complicated the promise. When coupled with massive fragmentation and a lack of interoperability, these realities became painful barriers to overcome when it came to adoption. Many of these pain points are, to this day, due to the lack of certain functionalities, a collection of gaps that the introduction of AI serves to fill.</p><p>Instead of introducing novel technologies, computer vision bestows the context of sight onto technologies that people are already accustomed to. It is no coincidence that a small camera (the personification of computer vision) is sitting in almost everyone’s pocket or purse at this very moment. It is even sitting in many people’s cars; cameras that were originally placed for visual aid are now coupled with intelligent sensors that alert drivers of possible impediments or dangers.</p><p>This permeation enables computer vision to be everywhere and everything given the right application build. In fact, previous dependencies on researchers and members of academia to implement a custom system are all but gone. Today’s sensor options and the ubiquity of camera functionality allow for unique pieces of hardware and software to be put in place and practice faster and cheaper than ever before.</p><p>And that is not all. Computer vision amplifies visuals at a resolution incomprehensible to the human eye, thereby allowing for the identification of subtle elements or patterns that people are unable to see. The object types and speed at which they are detected have also both drastically increased. Such a large amount of data would be too cumbersome and time-consuming for humans to gather, but the same cannot be said for intelligent software.</p><h3>How does AI further push this potential?</h3><p><a href="http://transcripts.hypergiant.com/machine-intelligence-hypergiant-edition-0">Machine intelligence has progressed to the point where data can be processed into information</a> that the software can learn from. This machine learning layer, when paired with the accelerated hardware that allows users to interact with it, becomes the sight that machines need to share their “experiences” with their human counterparts.</p><p>The human brain has insights, emotions, references, and ideas that come to life just by looking at something. Today’s computers, on the other hand, are semantically blind — they cannot <em>think</em> about what they are seeing. This blindness is cured by AI and deep learning. For a long time, state-of-the-art computer vision meant pointing to a car and identifying it as such. Nowadays, the ability to actively learn is being built into machines, thereby removing the need to tell it what things are. The AI is trained experientially; much like a child, it is shown thousands of pictures and videos of cars until it understands on its own what a car actually is.</p><p>With cameras and computer vision, that same machine can now be fed a great deal more data, but in a format analogous to the way people absorb their environments. The AI then evaluates problems and opportunities in a manner that is relative to our experiences as humans. It can label them, train itself, and produce outputs within the familiar boundaries of how users extract relevant information. The symbiosis between the two systems — digital and biological — becomes intuitive in the greatest sense of the word.</p><p>The evident difference between humans and machines, on the other hand, is in accuracy. Task automation such as detecting an object and placing it somewhere is achieved through an algorithm whose precision and speed increases with each occurrence. Without the disadvantage of human inconsistency, the algorithm’s return on investment (ROI) is at a constant incline as the robot reaches a level of accuracy that was unimaginable mere decades ago. All it needs to do, first, is overcome the hurdle of acquiring good data.</p><p>The adoption and introduction of computer vision is not a matter of intelligent technology or hardware — the latter improves exponentially every day, while the former is sustained by continued research and case studies. The difficulty lies in the system utilized to create the data; in other words, the tool that formulates and initially augments data must understand the specific problem that each organization is trying to solve. Monolithic services from giants such as Microsoft, IBM, and Amazon may correlate large quantities of data, but (unless the user’s ask is quite generic) the result is not as qualitatively satisfying when compared to a custom solution.</p><p>For example, if access to the necessary data does not exist, then the intelligent system is required to synthesize it. This is specifically important for robotics (a substantial opportunity area for most industries) that need test environments involving a variation of stimuli such as lighting and textures synthesized for the machine’s learning experience. Data like this can, in fact, be fabricated by an intelligent machine that understands the unique demands and goals of the organization at hand. This results in superior model testing prior to any kind of real-world release.</p><p><em>Everything changes with computer vision and there is no industry that will not benefit. See examples and read more by downloading our PDF entitled </em><a href="https://transcripts.hypergiant.com/ai-computer">AI-Enabled Computer Vision Paves the Future of Food, Marketing &amp; Health</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c4b3f5780106" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is Neural Network Hype Killing Machine Learning?]]></title>
            <link>https://medium.com/@hypergiant/is-neural-network-hype-killing-machine-learning-120041406f1?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/120041406f1</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[neural-networks]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Fri, 07 Sep 2018 13:43:30 GMT</pubDate>
            <atom:updated>2018-11-09T21:01:02.051Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wC1aGQQjAHDrPinlrEoCwg.png" /></figure><p><em>Written by Drew J. Lipman, PhD, Lead Data Scientist at Hypergiant</em></p><p>In the fall semester of 2016, I was attending a seminar series taught by <a href="http://ssather.people.clemson.edu/">Dr. Sather-Wagstaff</a> about the applications of algebraic topology, in particular <em>homologies</em>. Homology is a way to connect two sequences of objects, normally from diverse branches of mathematics, to each other in a meaningful way. Ideally, with these types of processes, you gain information about each sequence by studying the other. Throughout that semester, we learned several abstract concepts and ideas that have proven useful when approaching problems. Consider, for example, the difficult and relevant issue of determining the number of separate, geometric regions defined by a problem (and if there are any holes), and turning it into a computational issue — namely, computing <a href="https://en.wikipedia.org/wiki/Betti_number">Betti numbers</a>.</p><p>What does this have to do with <a href="https://www.hypergiant.com/what-is-machine-learning/">machine learning</a>, exactly? Ask yourself: Is taking an abstract problem and converting it into a numeric computation that can be solved not one of the fundamental problems of machine intelligence? The goal is not, has not, and should not be to build larger or more complex neural networks. The goal of machine learning has always been to make abstract problems understandable to machines. That is, to compute answers to problems. Data scientists and industrial <a href="https://www.hypergiant.com/ai-development/">AI developers</a> seem to get hung up on using neural networks, or <a href="https://medium.com/@karpathy/software-2-0-a64152b37c35">Software 2.0 as some call it</a>, as the only acceptable technique to a problem. Consequently, the news — and, therefore, public opinion — lacks information about more fundamental techniques for converting the abstract into the computational and is filled with information about how neural networks have achieved this result, instead.</p><p>That being said, neural networks do offer a wide range of applications; when trained on a large, properly-curated dataset with enough computational power and time, neural networks have produced results that match or exceed both human benchmarks and the performance of any other machine learning algorithm. This is why they represent such a rich area of research and development.</p><h3>The Faults at Hand</h3><p>What, then, are some fundamental problems with neural networks? While there are several, the ones we will discuss, in particular, are transparency of the resulting function, failure analysis for when things go wrong, and the lack of understanding in the community.</p><p>What do we mean by transparency, you may ask? Consider the following toy example: An insurance company hires a data scientist to help them streamline their process of identifying people eligible for coverage. After an appropriate amount of time analyzing the data, environmental factors, health metrics, and employment trends — not to mention, a lot of Natural Language Processing of news articles — the data scientist builds a neural network that inputs applicant data and outputs a list of various details such as the expected return value and insurance duration.</p><p>That may seem reasonable, but under a transparency law <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2972855">similar to what the EU proposed</a>, rejected applicants have the right to ask how their data was processed and used in <a href="https://arxiv.org/pdf/1606.08813.pdf">insurance risk assessments</a>. A neural network does not really offer the kind of transparency required, especially in the insurance business where it is immoral, and illegal, to deny an individual coverage based on factors such as ethnicity. So, for example, how does a company defend itself against the allegation that its protocol denies people insurance based on ethnicity? How do you prove that the neural network is not, in fact, doing that? Neural networks are very good at correlating data, and oftentimes make correlations that are not there or (worse) are illegal. Sure, you can simply remove that data from the training set, but information such as hometown is important and does provide a large amount of statistical evidence regarding sensitive characteristics such as ethnicity and religion. In short, even if you do not provide the training set with the kind of data that might produce illegal correlations, the neural network might make those connections anyway.</p><p>Now, consider autonomous vehicles. This is the great, long-hoped-for promise of computer vision. If nothing else, easing the tribulations of long distance driving makes these worthwhile. Inevitably, however, any accident brings about the question of whether the driver, the vehicle’s mechanics, or the AI is responsible for the accident. Who’s to be sued for damages?</p><p>Knowing <em>why </em>an artificial intelligence failed at a task is often just as important as knowing <em>when </em>it failed. There is an example, often used, in which <a href="https://www.usatoday.com/story/tech/2015/07/01/google-apologizes-after-photos-identify-black-people-as-gorillas/29567465/">a neural network image classifier labeled African Americans as gorillas</a>. We know that the neural network failed, but how did it fail and what can Google do to fix the problem? Is it truly a cultural diversity problem as suggested in the article? Or is it an artifact of the training set reflecting population dynamics and the fact that African Americans only appear in 16% of human photos? Could it have been due to lighting conditions confusing the classifier or some other reason? Given the <a href="https://arxiv.org/pdf/1712.09665.pdf">susceptibility of neural networks to malicious real world attacks</a> and <a href="http://ieeexplore.ieee.org/document/8203770/">fault injection</a>, determining the cause of any failure is very important.</p><p>The issue with neural networks is that, many times, we just do not know what caused an incorrect classification.</p><h3>Some Much-Needed Clarification</h3><p>Ask yourself, what are neural networks? If you answered “machine learning models based on the behavior of neurons,” then congratulations, you have bought into the hype. Try this: Ask your friendly, neighborhood neural biologist what artificial neural networks have to do with biological neural networks. One out of three times, you will be laughed out of the office, argued with, or given a look of disgust. People believe that biological neural nets are similar to artificial ones — that they actually inspired them — when they are not.</p><p>A neuron (artificial or organic) receives inputs, often of varying strength, from other neurons and does something as a result. That is the extent of the similarities between the two. Basically, artificial neural nets and biological neural nets have about as much in common as I have to Charlize Theron. In short, one is an example of beauty and efficiency, while the other is an amusing facsimile created late one night in a basement. This is not to say that there is no value in using biology to inspire programs, but (as is the case with most biologically-inspired algorithms) the similarities are topical at best, and more about having a cool-sounding name to get interest — and, subsequently, funding and citations — for your research.</p><h3>Where Does That Leave Us?</h3><p>We do not understand how neural networks make decisions, and cannot explain it to others if needed; fixing neural networks when they go wrong is <a href="https://ai.googleblog.com/2015/06/inceptionism-going-deeper-into-neural.html">such a monumental task</a> that Google often just begins again; and the public generally does not understand what they are or how they are designed. All of that being said, when a problem comes up and a so-called “intelligent” solution is called for, what do most want? A neural network! When do they want it? Right now! There seems to be no real interest in pursuing those classical methods that often provide solutions of equal or superior quality for the datasets provided.</p><p>Which begs the question: Is this hype killing the development of other <a href="https://www.hypergiant.com/machine-learning-techniques/">machine learning techniques</a> that provide equivalent (or greater) value?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=120041406f1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Oil & Gas Workers Are Evolving In Relation To Technology]]></title>
            <link>https://medium.com/@hypergiant/how-oil-gas-workers-are-evolving-in-relation-to-technology-e455709e9857?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/e455709e9857</guid>
            <category><![CDATA[oil-and-gas]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[organizational-culture]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[knowledge-management]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Thu, 30 Aug 2018 14:14:36 GMT</pubDate>
            <atom:updated>2018-11-09T21:04:23.331Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0Q5_9_bjwZqqCCcRgLtedw.png" /></figure><p><em>Written by Marc Boudria, VP of Artificial Intelligence at Hypergiant</em></p><p>Artificial Intelligence (AI) in the oil and gas industry stands to <a href="https://www.marketsandmarkets.com/Market-Reports/artificial-intelligence-oil-gas-market-87246288.html">reach US$2.85 billion by 2022</a>. Why? Because data is never special. Oil rigs may generate somewhere around 50 terabytes a year, but that kind of big data needs to be applicable to be useful and, unfortunately, humans do a terrible job of classifying things into datasets. Indeed, a good scenario will see 10% of the resulting datasets actually be beneficial.</p><p>Most competing firms are also known to have access to the <em>same </em>datasets. This means that how they each choose to clean up, interpret, transform into information and disseminate that data throughout the organization is where the opportunity lies. More and more organizations understand that, even though their focus is often on mitigating risk (think health and safety), risk is now necessary in order to move the needle. They are ready to change the way they treat their people and their screens, and must use AI to do it.</p><h3>Current Employees</h3><p>The older generations now retiring from the workforce possess a great deal of knowledge that is, in fact, intuitive and difficult to convey to younger entrants. But <a href="http://transcripts.hypergiant.com/evolution-oil-gas-worker-download-whitepaper">when paired with a strong, AI-powered database structure</a>, the learning process can be made infinitely easier by the adoption of user experience rules and trends that are more resonant and relevant to the next generations of workers — Millennials and Z’s — coming in.</p><p>Current employees, given their proficiency in the work and processes, are best suited to be the teachers and stewards of any new, AI-powered system. Beyond their knowledge of traditional systems, these pioneers must also accept that every generational worker after them will have different situations to deal with. The work focus of future employees will involve interacting with machines as they would with coworker tools more than anything else. And it is the job of current employees (together with <a href="https://www.hypergiant.com/ai-programming/">AI specialists</a>) to answer how this all happens from a knowledge perspective.</p><p>It stands to reason then that creating a document of information and sticking it in Sharepoint is not knowledge management. So, what exactly is at stake here? The answer is democratizing veterans’ knowledge in a quantified way that is useful for both the machines and humans. (This does not exist today.) Current employees’ responsibilities will shift towards that of AI instructors; they will continue to do their jobs alongside an intelligent machine, effectively defining what the future of that system will be. At the end of the day, these pioneers — those that have the traditional skills and knowledge — need to be the ones to validate that the machine is correct.</p><p>The evolution of that machine intelligence will, ultimately, result in the creation of new jobs and opportunities. <a href="https://www.gartner.com/doc/3827163/predicts--artificial-intelligence">Predictions suggest</a> that AI will create more jobs than it eliminates by 2020 and will also be seen recovering 6.2 billion hours of worker productivity across sectors a year earlier. Promising, indeed.</p><h3>Future Employees</h3><p>Machine intelligence does not spell the end of human interaction in the oil and gas industry, but <a href="https://www.gartner.com/doc/3869463/ai-future-work-oil-gas">it does propose some modifications</a>. Jobs such as oversight assessment and quality assurance or training and maintenance of the AI system are some of the more obvious ones. Not only does this mean field work, but it also means personnel in control centers monitoring both the machines and employees involved. Other roles will skew more towards performance management and leading bigger teams of digital specialists. All in all, future employees in this industry stand to be remote digital knowledge workers more than anything else.</p><p>Entering the industry will not look the same either — a picture painted clear by the potential brought on by natural language processing (NLP). NLP will eliminate the need for future employees to have deep, industry knowledge simply by mitigating the challenge of learning (and retaining) the current glossary of acronyms needed to speak intelligently with other industry members. Not to mention that once an employee is hired, he or she will enter a much more homogenized and automated organization. Just think about it… Satellite radar, in combination with geological drones and historical findings in the region, can be used to detect where wanted minerals are. The hot spots can then be identified by a robotic unit and only after a threshold is met would the human supply chain kick in. This level of automation is all possible once <a href="https://www.hypergiant.com/capabilities-of-artificial-intelligence/">machine intelligence</a> is finding and looking into goal-driven patterns.</p><p>Better yet, machine intelligence empowers the system to simultaneously learn the way each and every employee thinks and acts. It remembers this better than the user does and alters its approach when influencing each particular user’s decision-making process. This is not to say that the user does not also gut check the machine — after all, the AI (like humans) is also capable of developing bad habits. Indeed, processes are built and designed to enable this exact kind of verification.</p><p>The truth is that many traditional oil and gas workers have learned what they know on the job and, therefore, find themselves occasionally straying from established and documented processes to achieve success. This becomes a noteworthy obstacle when hiring new generations. Not only will they be trained less for the oil and gas industry specifically and more for engineering as a whole, but they will also be searching for the employer with the best tool set for new hires. All of a sudden, the AI system becomes the differentiating factor for top talent looking for its next employer.</p><p><em>This article originally appeared in </em><a href="https://www.epmag.com/how-oil-gas-workers-are-evolving-relation-technology-1712876"><em>E&amp;P Magazine</em></a><em>, but there’s a lot more where it came from! We recently published an entire paper on supporting the evolution of the oil and gas worker with design, science, and engineering. </em><a href="http://transcripts.hypergiant.com/evolution-oil-gas-worker-download-whitepaper"><em>Download the full PDF here.</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e455709e9857" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Technologies and Technical Considerations of Data Privatization]]></title>
            <link>https://medium.com/@hypergiant/the-technologies-and-technical-considerations-of-data-privatization-7bae047aa97e?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/7bae047aa97e</guid>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[bitcoin]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[big-data]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Mon, 13 Aug 2018 12:31:56 GMT</pubDate>
            <atom:updated>2018-11-12T20:34:01.846Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PPj8q-FYSAH0pflxEajMxQ.png" /></figure><p><em>Written by Drew J. Lipman, PhD, Lead Data Scientist at Hypergiant</em></p><p><a href="https://medium.com/@hypergiant/what-data-privatization-is-and-why-it-matters-to-you-32ccc4a7025e">In part one of this series</a>, the evolution of big data, what it means to privatize data, and the merits of privatizing personal data (especially for the consumer) were discussed. In part two, the merits of privatized data <a href="https://medium.com/@hypergiant/how-and-where-data-privatization-can-be-implemented-3adfa5dd09b6">were discussed in terms of corporations</a>, along with the exploration of the high-level details involved in the potential development of a widespread, privatized data system. Here, to bring the topic to a close, some of the technical considerations and technologies that may play a role in a privatized data system are examined — with a focus on ease of understanding.</p><h3>Cloud Security</h3><p>Since the year 2000, cloud computing has been a growing piece of corporate life, with cloud storage security growing apace with it. As with many security protocols involving humans and computers, implementation and user interface are at least as important as the theory behind the security. Bluntly speaking, no matter how theoretically sound the security is, if your password is weak or password-structure is known, then there will be issues.</p><p>Given this, how can cloud computing security be implemented for privatized data? Many corporations maintain an Identity Management System that employees are required to go through as a prerequisite for accessing confidential or classified information; to shift a verification protocol as closed as this to the cloud (even when utilizing multi-factor identification) would be fundamentally insecure.</p><p>As for the data itself, cloud storage is susceptible not only to traditional hacking approaches, but also to a growing number of side channel attacks (those outside the mathematical model), such as the <a href="https://www.cs.unc.edu/~reiter/papers/2014/CCS1.pdf">sharing of a CPU cache</a>. Maintaining the segregation of privatized data storage servers from other servers may be necessary because of this — just one of a long list of concerns regarding data security in cloud computing.</p><p>We must note that these are not new problems, but they are ones that the <a href="https://www.wired.com/story/equifax-breach-response/">current system is failing to solve</a>. To counter them, however, increasingly <a href="https://arxiv.org/pdf/1610.00043.pdf">sophisticated coding theory applications</a> are making it so that data stored in the cloud is among the least likely to be lost due to an accident.</p><h3>Blockchain</h3><p><a href="https://en.wikipedia.org/wiki/Blockchain"><em>Blockchain</em></a> is yet another magic term floating around. Many of us have heard it, a few of us understand it, and even fewer understand it well enough to <a href="https://cryptonote.org/cns/cns003.txt">implement it successfully</a>. At a very high level, a blockchain is a decentralized and distributed ledger of transactions secured by cryptography in such a way that it resists modification. At this point in time, your eyes may be closing due to math-lecture-fatigue (MLF), but — trust us — it gets clearer.</p><p>Suppose we have an asset — in this case, a piece of personal data. A blockchain would then be a part of a ledger recording who has permission to access that data. This is why blockchains have become an integral part of modern cryptocurrency (the context within which most people first hear the term). The key takeaway is that a blockchain records the history of transactions that it has been involved in in a way that is all-but-impossible to reproduce (it would be easier to reproduce paper currency). Its transactions can, therefore, be enforced legally.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jRniYwYhWT3ZfNqsRUG94g.png" /></figure><h3>Cryptocurrency</h3><p>While not directly part of data privatization, cryptocurrency is often the currency of choice for many software companies and also where most blockchain research has been performed. Due to the convoluted nature of modern cryptography (What do <a href="https://en.wikipedia.org/wiki/Elliptic_curve_cryptography">elliptic curves over fields with finite characteristic</a> have to do with anything?) and its interconnectivity with cryptography, explanations of the material can be both annoyingly complex and stunningly simplistic.</p><p>Our interest in it, however, lies in its ability to produce <a href="http://firstmonday.org/ojs/index.php/fm/article/view/548">smart contracts</a> that control access to privatized data. This is related to public key cryptography — a core part of the Internet. Look at the address bar of your web client. The “s” in “https” signifies a secure connection. To form such a connection, information must be shared between your computer and the server without anyone else knowing what it is. This is the fundamental problem with public key cryptography and it is where elliptic curves get involved.</p><p><em>Want to learn more about how machine intelligence factors into your data privacy? We cover these technologies and more, not to mention </em><a href="https://www.hypergiant.com/blockchain-strategy/"><em>blockchain strategy</em></a><em> and the applications of data privatization across consumer and corporate needs, in our recent paper. Download the </em><a href="http://transcripts.hypergiant.com/data-privatization-download-whitepaper"><em>full PDF on data privatization here</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7bae047aa97e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How and Where Data Privatization Can Be Implemented]]></title>
            <link>https://medium.com/@hypergiant/how-and-where-data-privatization-can-be-implemented-3adfa5dd09b6?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/3adfa5dd09b6</guid>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[virtual-assistant]]></category>
            <category><![CDATA[machine-intelligence]]></category>
            <category><![CDATA[big-data]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Fri, 03 Aug 2018 10:46:32 GMT</pubDate>
            <atom:updated>2018-08-03T10:46:32.699Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tPW-fEvvTL_WT2M4XcT1HA.png" /></figure><p><em>Written by Drew J. Lipman, PhD, Lead Data Scientist at Hypergiant</em></p><p>Having discussed <a href="https://medium.com/@hypergiant/what-data-privatization-is-and-why-it-matters-to-you-32ccc4a7025e">why data privatization is good for the consumer</a>, how does it benefit corporations? And how can the public leverage enough control to convince major corporations to not keep users’ data? The obstacles here are twofold: (1) getting enough people to go in the same direction when most do not realize that there is a problem and (2) using that mass movement of public opinion to convince corporations that it is in <em>their </em>interest to do something.</p><h3>Corporate Interests</h3><p>Out of these two issues, the second is (oddly enough) easier to approach. What are corporations interested in and how can that be used as leverage? At their core, major corporations become major corporations by always having one eye on the bottom line. And, since their interests are easily understood, they are easier to discuss. In order to make data privatization a reality for such enterprises, the collection, storage, and sale of consumer data needs to become less profitable.</p><p>Easier said than done. Big Data is a billion-dollar industry and consumer data usage has been a part of corporate Big Data for almost two decades. Google, for example, was founded in 1998 based on its ability to understand what people are interested in and looking for — a service offering that requires access to people’s data.</p><p>How do we make an approach like Google’s unprofitable? Well, unless you are in the habit of <a href="https://www.quora.com/Whats-the-difference-between-VPN-and-TOR">browsing via VPN or TOR</a>, your Internet service provider (along with Google, Microsoft, or any search engine of your choice) knows what you search for and click on. And, given how irate people get over even the smallest slow down of their regular Internet browsing, VPN and TOR are not ideal tools for affecting a major, cultural change.</p><p>Alternatively, we can make it cheaper to <em>not </em>collect, store, and disseminate data. First of all, storing individualized data in the Cloud would not only grant (limited) permission to major corporations, but it would also remove their data storage costs. Second, if the access permissions for a single corporation were broad enough, then the number of calls upon the data would not be a problem either. Lastly, with the advent of <a href="https://dl.acm.org/citation.cfm?id=1536414.1536440">Homomorphic Encryption</a>, it may soon be feasible to do large computations directly in the Cloud, without having to decrypt any data sets — another development that would lighten a corporation’s hardware or vendor costs.</p><p>Method options certainly exist, but garnering enough interest in privatized data to make them feasible is another challenge entirely. The only way to convince a large group of people to do something, in fact, is to convince them that they <em>want </em>to do it. To paraphrase Locke: a technology only lasts as long as people consent to its usage. Perhaps the best (and only) approach to convincing people to implement privatization in their day-to-day lives is education — an exchange that you are participating in at this very moment. The more people understand data usage, and their options surrounding it, the better we are as a whole.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IB4V_JAYCjnLSyk2iJnYug.png" /></figure><h3>Mutually Beneficial Privatization</h3><p>Several applications of privatization stand to better user experience as much as they stand to optimize business processes. Online identity verification is a good example. Rather than your email or cell phone, a properly implemented system of privatized data could be used to verify your identity. As it should be given email’s failings in this department.</p><p>Another privatization application relates to education. If data privatization is applied to include educational data, then each individual’s qualifications and certifications become easily accessible in real time. For example, a Computer Science student with privatized data showing his qualifications in JavaScript, C++, and Knowledge Graphs can validate his competencies even without a degree. In other words, his education to date has value — even if he did not complete some line-in-the-sand collection of requirements set by people outside of his chosen employer or industry.</p><p><em>Want to learn more about how machine intelligence factors into your data privacy and that of your users or consumers? Download our </em><a href="http://transcripts.hypergiant.com/data-privatization-download-whitepaper"><em>full PDF on data privatization here</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3adfa5dd09b6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Data Privatization Is and Why It Matters to You]]></title>
            <link>https://medium.com/@hypergiant/what-data-privatization-is-and-why-it-matters-to-you-32ccc4a7025e?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/32ccc4a7025e</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[virtual-assistant]]></category>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[machine-intelligence]]></category>
            <category><![CDATA[privacy]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Fri, 20 Jul 2018 17:53:58 GMT</pubDate>
            <atom:updated>2018-07-20T17:54:46.709Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0OhKBZnRblbsRnj63-TIbw.png" /></figure><p><em>Written by Drew J. Lipman, PhD, Lead Data Scientist at Hypergiant</em></p><p>Unless you have been living in a cave, you are probably familiar with (or have at least heard of) Big Data. What Big Data means, how it impacts your life, and what its existence means for the world around you, however, are less understood. And, while the issue of how Big Data’s very existence impacts your life is rarely discussed, it might be of special interest to know exactly how <em>your </em>personal information is being used as data points within the larger web and scope.</p><p>One response to this is an approach to personal data control called <em>data privatization</em>. While <a href="http://julkaisut.valtioneuvosto.fi/bitstream/handle/10024/78439/MyData-nordic-model.pdf?sequence=1&amp;isAllowed=y">this concept is not new</a>, there is <a href="https://docs.google.com/file/d/0B0Mrcy3jPHMYOC1BcWQ5dlItb0E/edit">huge potential growth in the market</a> for methods and programs that privatize data, and savvy consumers (along with those who market to them) should make themselves familiar with these concepts.</p><h3>What Privatization Means</h3><p>We are in the simultaneously wonderful and horrible situation where our data is being both actively collected and processed. What does this mean for consumers? Major corporations are collecting, storing, using, and, in some cases, <a href="http://www.npr.org/2017/03/28/521831393/congress-overturns-internet-privacy-regulation">selling data about you</a> without your permission or with permission buried so deep in a user agreement that you don’t even know you’ve agreed. Moreover, unless you want to live without the services they provide, there is not a whole lot that can be done to stop these organizations — given how the system is currently designed. Even if the data sold is sanitized (with names and addresses removed), it is often not enough to protect your privacy.</p><p>For example, when Netflix issued a challenge to help improve their movie suggestion algorithm, they provided a sanitized database. However, with a little side channel knowledge from outside the dataset (terminology stolen from cryptography), <a href="https://www.cs.utexas.edu/~shmat/shmat_oak08netflix.pdf">your account can be identified</a>. If you are interested in making your data private, meaning you have it and no one else does, <a href="https://us.norton.com/internetsecurity-how-to-ten-ways-to-keep-your-data-private.html">there</a> <a href="https://www.forbes.com/sites/learnvest/2017/01/27/7-ways-to-protect-your-private-cellphone-data-from-hackers/#6e296f6213ff">are</a> <a href="https://www.wired.com/video/2017/06/how-to-make-your-browsing-data-more-private-than-a-thousand-incognito-windows/">options</a>. Most, however, are more about protecting yourself from hackers than from legitimate, “trusted” vendors collecting information about how, when, and where you interact with them.</p><p>Since many of us are unwilling, or unable, to do without our Netflix, Amazon, Ebay, Facebook, LinkedIn, or Google accounts, there is very little that we can do to make our data truly private. Instead, we ask the obvious question: why do corporations collect data about their users? The immemorial answer is money. Big Data about customers is worth billions of dollars. So, if companies are making money using your data, and it is not really feasible to stop them from collecting it, then what can we do?</p><p>One option is to <em>privatize</em> your data. If someone is to make money from your data, then why not you? While the concept of privatized data has been around for a while, it is only recently that Machine Intelligence techniques have evolved enough to make it a reasonable approach. Anyone nowadays with a smartphone or Windows has Machine Intelligence (MI) installed — a Machine Intelligence that would make true privatization feasible if it were working for the consumer instead of the corporation.</p><h3>Consumer Benefits</h3><p>A simple search for your name on Google, and possibly a few other common websites, reveals more information about you than anyone (with the possible exception of your mother) cares to know about you. Most people do not even know what information about them is out there — after all, how often do people do a simple search for themselves?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CLsi7GI6dJaJj3FPUGoJtQ.png" /></figure><p><em>Want to learn more about how machine intelligence factors into your data privacy? Download the </em><a href="http://transcripts.hypergiant.com/data-privatization-download-whitepaper"><em>full PDF on data privatization here</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=32ccc4a7025e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Blockchain Regency: Consent and Consensus]]></title>
            <link>https://medium.com/@hypergiant/the-blockchain-regency-consent-and-consensus-d1f1708dde78?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/d1f1708dde78</guid>
            <category><![CDATA[policy]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-intelligence]]></category>
            <category><![CDATA[bitcoin]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Fri, 06 Jul 2018 21:19:17 GMT</pubDate>
            <atom:updated>2018-11-12T20:47:10.559Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-MHcK9DttaCG6AXJ7cEYKQ.jpeg" /></figure><p><em>Written by Ross A. McIntyre, VP of Strategy at Hypergiant</em></p><p>To wax <a href="https://www.youtube.com/watch?v=v1cVl7KHsGA">Seinfeldian</a> for a moment, I have to ask: What’s the deal with purebred dogs? And how about the British royal family? Or exotic derivatives, have you heard of those? Oh, and <a href="https://www.hypergiant.com/blockchain-strategy/">blockchain strategy</a>, amirite?</p><p>Until the advent of the Declaration of Independence, most people around the world lived under some form of autocratic rule — either a dictatorship or a hereditary monarchy. These rulers were often the individuals able to seize power over a group or nation (e.g., Genghis Khan). And, many times, such leaders would found a hereditary monarchy based on some form of “<a href="https://www.britannica.com/topic/divine-right-of-kings">divine right</a>” (as in Europe or South America) or “<a href="https://courses.lumenlearning.com/boundless-worldhistory/chapter/the-zhou-dynasty/">Mandate of Heaven</a>” (as in China). Under such rule, people were considered subjects rather than citizens, while enforced obedience and fealty became law.</p><p>Nowhere was there “<a href="http://democracyweb.org/consent-of-the-governed-principles">consent of the governed</a>” as the limitations frequently placed on such rulers by the wealthy, the bourgeois, and the landed gentry only led to a system of consent by aristocracy. Even when the regent’s powers were checked through agreement (e.g., the <a href="https://en.wikipedia.org/wiki/Magna_Carta">Magna Carta</a>), he/she still maintained unrepresentative rule, the conceit of citizen sovereignty awaiting its next opportunity for exposure: when the thirteen American colonies began their revolution. (Caveat here being that the world’s oldest democracy doesn’t actually offer direct election, employing instead the mechanism of the Electoral College, which, from the standpoint of the majority, has fresh failures.) On some level — pain of death and jackbooted enforcers notwithstanding — monarchs only have the power that is ascribed to them by their subjects.</p><blockquote>“<em>In free governments, the rulers are the servants and the people their superiors and sovereigns.</em>” — Benjamin Franklin</blockquote><p>Now, take purebred canines: dogs of modern breeds whose pedigrees are documented in a stud book that can then be registered with a breed club that may or may not be part of a national kennel club. This distinction says nothing about the quality of the dog — in terms of health, temperament, or sagacity — but rather denotes only that parentage is documented and traceable. Nowadays, some kennel clubs guarantee parentage through DNA testing, but most breed clubs still rely on the breeder’s word. Dog breeding was initially the (almost) exclusive avocation of the wealthy, whose reputations were considered to be on the line.</p><blockquote>“<em>If you steal my money, you’re just stealing trash. It’s something, it’s nothing: it’s yours, it’s mine, and it’ll belong to thousands more. But if you steal my reputation, you’re robbing me of something that doesn’t make you richer, but makes me much poorer.</em>” — Shakespeare, Othello, Act III, Sc. iii</blockquote><p>Finally (stick with me), let’s look at derivatives. Categorically, a derivative is a <a href="https://www.investopedia.com/terms/d/derivative.asp">financial security whose value is derived from the value of underlying assets</a> (e.g., stocks, bonds, commodities, etc.). And it is a form of contract that generally requires little or no money to be invested up front. Derivatives can be used for <a href="https://www.thebalance.com/what-are-derivatives-3305833">commodities such as oil or gold, but are also flexible enough to include currencies, stocks, bonds, and interest rates</a>. Interestingly enough, the seller doesn’t have to own the underlying asset. He/she can simply fulfill a contract by giving the buyer enough money to acquire the asset at the agreed-upon market price or by offering another derivative contract to offset the initial one’s value. Derivatives, therefore, are easier to sell than the assets themselves.</p><p>Exotic derivatives — those whose values are difficult to ascertain due to their complex features or formulas — can feel less “real” than other financial securities. But, to put it simply, the fact that payments may be complex formulas is neither here nor there — the derivative remains a contract by which two parties exchange payments. When you sign up with a mobile service provider, for example, you agree to pay a monthly fee in exchange for voice and data, both of which are very real even if they have no tangible manifestation. This constitutes an abstract concept that has a mutually agreed upon value determined by consent and consensus in the marketplace.</p><p>Which brings us to blockchain. <a href="https://blockgeeks.com/guides/what-is-blockchain-technology/">Blockchain takes myriad forms</a>, but (for the sake of this piece) let’s describe it as a decentralized public database that keeps public records in an additive way alone: data can be appended, but not removed (a feature called <a href="https://medium.com/@chrshmmmr/consensus-in-blockchain-systems-in-short-691fc7d1fefe"><em>persistence</em></a>). Records are also visible to the public, making blockchains auditable insofar as they allow authenticity verification for each transaction. The advantage of a decentralized database such as this over a traditional, centralized database is as follows: traditional systems require user-controlled access by known (and trusted) persons, while a blockchain is operated by entirely unknown parties (individuals, groups, computer agents, and so forth). Lack of faith is something of a feature rather than a bug. That’s because, while any entity with unknown trustworthiness can submit info to a blockchain, the distributed operators (investors, owners, or what have you) must evaluate and agree upon addenda before they are permanently integrated into the blockchain. In other words, consensus must be gained.</p><p>Given the vast inconsistencies between disparate, subjective views of reality — and our inability to agree upon the ontology of existence — there is a societal drive towards obtaining a consensus view. Theocracies have very different viewpoints from their secular counterparts, religion and faith being effortful categorizations and classifications of the fundamental whys and wherefores. Despite seemingly empirical evidence that runs contrary to teachings, people often stick to their beliefs. How else do we explain the persistence of <a href="https://en.wikipedia.org/wiki/Flat_Earth">flat-Earthers</a>?</p><blockquote>“<em>Reality is merely an illusion, albeit a very persistent one</em>.” — Albert Einstein</blockquote><p>(A quick sidebar: I felt very betrayed by reality when I learned, many pounds later, that vodka has calories. I believed society to be in universal agreement that clear beverages have no calories. Sadly, I was either wrong about the consensus or the collective impact of said consensus on objective reality.)</p><p>Now, <a href="https://en.wikipedia.org/wiki/Consensus_reality">consensus reality</a> is an agreed-upon reality based on a consensus view. But what is to prevent a group of blockchain members (assuming anonymity has been compromised) from putting forth a subjective view of reality that conflicts with, you know, actual reality? What is there to prevent a kennel club from falsifying pedigree or a large enough group of investors from dictating a skewed valuation of a derivative?</p><p>Let’s look at paper currency — one of the most ubiquitous examples of consensus reality — for a second. The market (made up of individuals) puts forth the idea that this form of paper is worth <em>x</em>. Back in the day, that indicated that the government backed each banknote with a corresponding amount of gold. But, after the <a href="https://en.wikipedia.org/wiki/Smithsonian_Agreement">Smithsonian Agreement</a> was reached (and the global market had time to respond), the US government changed the definition of the dollar by removing gold references from statutes. And, since October of 1976, the international monetary standard has been composed entirely of <a href="https://en.wikipedia.org/wiki/Fiat_money">fiat money</a>. (And, yes, the whole notion that gold has some intrinsic value is, in actuality, also based on consensus reality. While the simple fact that it is physical lends it some value, the consensus is in the magnitude of the value relative to other fungible commodities.) Fundamentally, currency is backed by the reputation or good name of the government that produces it; absent that public reputation or societal mores associated with mendacity, why should we trust blockchain? There is no implicit mechanism in the technology that prevents consensus hijacking.</p><p>Largely, it’s because the <a href="https://www.britannica.com/biography/Marie-Jean-Antoine-Nicolas-de-Caritat-marquis-de-Condorcet">Marquis de Condorcet</a> tells us to relax. Marie Jean Antoine Nicolas de Caritat (the aforementioned marquis) was an 18th-century philosopher and mathematician whose ideas and writings are considered reflective of the ideals of the <a href="https://en.wikipedia.org/wiki/Age_of_Enlightenment">Age of Enlightenment</a> and philosophy of <a href="https://en.wikipedia.org/wiki/Rationalism">rationalism</a>. His <a href="http://mathworld.wolfram.com/CondorcetsJuryTheorem.html">jury theorem</a> promotes the following idea: If individual votes are independent of one another (i.e., people formulate their own opinions without being influenced by other jurors), and each voter makes the correct decision at least half of the time, then the probability of the group’s majority coming to a correct decision improves as jury size increases. Wikipedia is a fantastic example of consensus reality: <a href="https://www.livescience.com/32950-how-accurate-is-wikipedia.html">veracity is self-reinforcing</a> given time and the appropriate number of contributors.</p><p>As a society, we accept certain things — objective truth being one of them — as reality based on consensus. But the social contract that underpins concepts like reputation and trustworthiness is not the only force pushing us towards a majority, consensus view. <a href="https://www.stat.berkeley.edu/~mossel/teach/SocialChoiceNetworks10/ScribeAug31.pdf">Math also plays its part</a> and, even with the increased pervasiveness of anti-scientific beliefs and “<a href="https://en.wikipedia.org/wiki/Alternative_facts">alternative facts</a>,” few argue against math’s objective truth.</p><p>At the proper scale, it would seem that honesty is not only the best policy, it is also the majority policy.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d1f1708dde78" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Product Strategy: Bridging the Chasm of Frustration]]></title>
            <link>https://medium.com/@hypergiant/product-strategy-bridging-the-chasm-of-frustration-8a6eaafcc391?source=rss-601d9e7f0ae------2</link>
            <guid isPermaLink="false">https://medium.com/p/8a6eaafcc391</guid>
            <category><![CDATA[product-strategy]]></category>
            <category><![CDATA[design-thinking]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-intelligence]]></category>
            <dc:creator><![CDATA[Hypergiant]]></dc:creator>
            <pubDate>Mon, 18 Jun 2018 14:55:39 GMT</pubDate>
            <atom:updated>2018-11-12T20:48:48.871Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V8EDsqhdjd03vma440srbw.jpeg" /></figure><p><em>Written by Ross A. McIntyre, VP of Strategy at Hypergiant</em></p><p>As clients move away from engagements that are strictly strategic, limited to user experience and visual design, or focused solely on digital development, there is an increased need for studios and agencies to provide services that holistically engage strategy, design, and <a href="https://www.hypergiant.com/ai-development/">AI development</a> players throughout each project’s lifecycle. Moreover, when the goal is creating a digital product, experience, or service, projects’ strategic aspects must be reimagined.</p><p>Enter product strategy, an arena for building multifaceted approaches based on several key principles:</p><ul><li>A focus on innovation over optimization and a belief that “parity is for chumps”;</li><li>A willingness to be provocative with a purpose and a penchant for speaking the truth;</li><li>Providing a bridge between business strategy and the product roadmap;</li><li>Rapid ideation, creation, and articulation at a high degree of fidelity;</li><li>An awareness of the various <a href="https://www.hypergiant.com/machine-intelligence-applications/">machine intelligence (MI) applications</a> and techniques available.</li></ul><h3>The Art of Making Progress</h3><p>Initially, attempts to bridge “Design Thinking” and (for lack of a better term) “Design Doing” left a gap between the traditional application of service-design-led strategy to business transformation and the somewhat-more-tactical methodologies underlining digital product creation. For clients, this marked a clear <em>chasm of frustration</em> — a yawning divide between the opportunities unearthed in a strategic or discovery phase (or service design blueprint) and the conviction required to craft a well-defined, service-driven product. Indeed, while colossal maps are brilliant for exposing key user journeys — thereby linking customer lifecycles with a panoply of touch points and familiarizing clients with multiple, differentiated possibilities — they are not always accompanied by direction and prioritization (i.e., “Now, you should do exactly <em>this</em>”).</p><p>Product strategists, on the other hand, burgeon decisiveness via an interplay between design, business, and technology. Their ultimate goal is to craft remarkable products that users love by determining what should be created and why, what well-articulated business problem it solves, and how to best innovate with technology. Traditionally, this involves the following actions and capabilities:</p><ul><li>Identifying the most important questions that a company’s products and services should address;</li><li>Translating insights into actionable solutions that can be executed and delivered;</li><li>Connecting design efforts to an organization’s business strategy;</li><li>Using technology to transform business needs into improved operations or new offerings;</li><li>Integrating design as a fundamental aspect of building meaningful products;</li><li>Prioritizing how products and services are launched;</li><li>Helping their clients manage change, so that new solutions stick;</li><li>Providing clients with an understanding of <a href="https://medium.com/@hypergiant/a-world-without-ai-is-scary-progress-isnt-4dfd77a1c2ba">how MI applies to their business needs</a>.</li></ul><p>By combining the above, product strategists ensure that (at a minimum) the product is purposeful, responsive, intuitive, contextual, and engaging — all while remaining compliant with more specific quality measurements, as well. They carefully weigh the needs of the business, user, and data (especially critical when delivering MI- or AI-powered products) to define the strategy and foster direction. And it is only by synthesizing all three of the aforementioned aspects and rationalizing them against one another that a suitable product is created.</p><p>Taking solely the user’s position when crafting an experience (aka, a user-centric approach), for example, will lead to a product that does not satisfy the needs of the business that is, most likely, footing the bill. Conversely, being responsible with business needs alone is what plagues so many minimum viable products (MVPs) in the larger marketplace. A simple tally of business goals rarely addresses users’ core emotional or logistical needs, and often results in products (or services) that lack the necessary degree of subtlety or nuance.</p><p>The final consideration — that of the data — hinges upon future, data mining expectations. What does the client want to see supported by analytics and what is the roadmap for implementing machine learning (or some other aspect of MI) to deliver on those expectations? How can explicitly-provided data return value to that same user by providing him/her with an elevated vantage point from which to effectively see the interrelationships between data? From a data science perspective, is the data hygienic, properly formatted, or granular enough without creating unnecessary entropy? Many companies gather data, but few companies gather it intelligently.</p><h3>Adapting for the Fourth Industrial Revolution</h3><p>There are considerable elements related to product strategy that are massively affected by MI and <a href="https://medium.com/@hypergiant/welcome-to-the-fourth-industrial-revolution-439bd6a75f82">the so-called Fourth Industrial Revolution</a>. Some of the new requirements are as follows:</p><ul><li><strong>Automate</strong>: consider this as one of the most basic forms of MI, and possible in most experiences;</li><li><strong>Elevate</strong>: use MI to provide clients with a 10,000-foot view of their own data and identify relationships that are difficult to see “in the weeds”;</li><li><strong>Eliminate</strong>: apply Progressive Disclosure to ensure that users are not exposed to superfluous information inapplicable to them (properly applied MI offers significant personalization improvements);</li><li><strong>Modularize</strong>: architect products in ways that facilitate MI technology integration in future states;</li><li><strong>Centralize</strong>: encourage clients to create a canonical source of truth when data is spread across multiple, disparate systems (thereby promoting a revisiting of data structure and hygiene, as well);</li><li><strong>Collect</strong>: gracefully introduce explicit data collection opportunities to accompany implicit collection (users’ demographic and psychographic details have tremendous value and considerable risk, <a href="https://www.nytimes.com/2018/03/19/technology/facebook-cambridge-analytica-explained.html">as seen with Cambridge Analytica</a>).</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8a6eaafcc391" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>