<?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 Sparkd.ai on Medium]]></title>
        <description><![CDATA[Stories by Sparkd.ai on Medium]]></description>
        <link>https://medium.com/@sparkdai-hq?source=rss-92061bf597c4------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*i2sBuHObyAjKg77lyMPVig.png</url>
            <title>Stories by Sparkd.ai on Medium</title>
            <link>https://medium.com/@sparkdai-hq?source=rss-92061bf597c4------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 07:46:25 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sparkdai-hq/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 best sport tech companies taking advantage of Artificial Intelligence]]></title>
            <link>https://medium.com/@sparkdai-hq/the-best-sport-tech-companies-taking-advantage-of-artificial-intelligence-c092c6c45bcd?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/c092c6c45bcd</guid>
            <category><![CDATA[football]]></category>
            <category><![CDATA[sport-tech]]></category>
            <category><![CDATA[computer-vision]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Mon, 16 Aug 2021 15:55:15 GMT</pubDate>
            <atom:updated>2021-08-16T15:55:15.426Z</atom:updated>
            <content:encoded><![CDATA[<p>The innovative infographic tool, — called Market Map — once again sheds light and clarity on a specific market: Sports Tech.</p><p>The infographic visualises the market of AI-based startups from all over the world operating in the Sports Tech sector. But let’s start by defining this relatively new term, SportTech, which has recently risen to prominence. It is a new area that arises from the encounter between sport and technology. In other words, all the technology aimed at finding solutions for sport is part of Sports Tech. It ranges from innovation in the field of Athlete Performance to new ways of Fan Engagement and the digitisation of clubs and federations business processes, as we will discuss deeper in the following paragraphs.</p><p>Our focus is certainly on a particular technology that Sports Tech relies on, Artificial Intelligence. So, let’s see how AI is used in Sports!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*GY6F954zXndcouHP.png" /></figure><p>Preview of the Sports Analytics Lumascape Tool</p><h3>Sports Analytics</h3><p>Everyone should have seen at least once in their life, one of the most beloved cinematic cult classics of all time: <em>Moneyball</em>. The film is about how <a href="https://en.wikipedia.org/wiki/Oakland_Athletics">Oakland Athletics</a> General Manager, <a href="https://en.wikipedia.org/wiki/Billy_Beane">Billy Beane</a> (played by Brad Pitt), strapped with a minimal budget, built a competitive team relying on <a href="https://en.wikipedia.org/wiki/Sabermetrics">sabermetrics</a>, a form of sports analytics, for player evaluation and making decisions.</p><p>The success of analytics-based decision-making in baseball was soon noted by other professional sports leagues and today it is widely adopted in other sports as well. According to estimates, 1,012 is the number of startups globally in Digital Sport, which offer digital solutions to monitor athletic performance, optimize event management, analyze and improve the involvement and experience of fans, and innovate club business processes.</p><p>But let’s take a step back and then clarify what is meant by Sports Analytics.</p><p>Sports Analytics is a collection of relevant, historical, statistics that can inform players, coaches, and other staff in order to facilitate decision-making both during and prior to sporting events.</p><p>There are two main areas of sports analytics — on-field and off-field analytics.</p><ul><li><strong>on-field analytics</strong> deals with improving the on-field performance of teams and players (game tactics and scoring, athletes performance, players training, and fitness)</li><li><strong>Off-field analytics</strong>, instead, is applied to the business side of sports to increase ticket and merchandise sales, improve fan engagement, offer highlights of matches and digital content through both social media and operating platforms.</li><li>alongside these ones, another distinct area of sports analytics use is <strong>sports gambling</strong>, but we do not treat it.</li></ul><p>Now that we have mapped out a general framework about Sports Analytics, let’s move to the two main sectors that result from the pair of opposites on-field / off-field and in the different subcategories that Lumascape presents.</p><p>Download all Sport Tech Map data Get all the data (including website, mission statements, etc.) related to this Market Map <a href="https://sparkd.ai/the-best-sport-tech-companies-taking-advantage-of-artificial-intelligence/">here</a>.</p><h3>Activity &amp; Performance</h3><p>As mentioned, this category is closely related to the on-field activity, so that means two moments of analysis: during the matches and in training. It is also the most popular category, in fact, almost one in two digital sports startups (49%) operates in Athletic Performance, offering solutions to measure performance, prevent injuries and monitor the training and rehabilitation of the athletes.</p><h3>Wearable &amp; Equipment Tech</h3><p>Through smart wearable systems equipped with sophisticated sensors, it is possible to monitor the player’s health and collect data about the performance in real-time. These are the technological solutions that some startups have successfully implemented. Innovative and comfortable designs characterize these devices, worn as clothing or an accessory, record health and fitness data during sports activities, and usually send them to a mobile app or a display, where you can see all the information in the blink of an eye. They typically track multiple metrics like blood pressure, speed, heart rate, calories burned, distance, etc. Wearable devices, such as smart shin guards, “intelligent” Sweat Band, and Smart Yoga Pants to name a few, are starting to gain popularity as consumers can now quantify their results on a daily basis. The benefits of wearable are both for those who train alone as a cyclist, or an amateur swimmer, and for the coach of an entire volleyball team for example.</p><h3>Athletic Performance Tracking &amp; Coaching</h3><p>If it is not possible to make an entire football team wear tracking systems with GPS and other sensors, there is another way to mathematically analyze performance, it is Video Analysis. Cameras track the ball &amp; player movements, generating huge volumes of data stocked in data-prep platforms able to deliver analysis and insights that truly support the coach’s and club’s decision-making process, or helps player agencies to scout new talents. Also, it can confirm a player’s objective value when buying players between clubs. Some startups have their own customized cameras among their offerings, which is certainly an advantage in terms of system compatibility and functionality for the use case, or rather the sport in question. Lastly, a much less used source of data for these sports analysis platforms is that of tabular data.</p><h3>Fans &amp; Content</h3><p>It is probably the most known and exposed part of the Sports Tech ecosystem to the general public: this section is in fact about Fan Engagement, so how to entertain fans and build a long-term relationship with the aim of converting them into fans.</p><h3>Fan Experiences &amp; Social/Media Platforms</h3><p>This category contains all those usefulI instruments to connect fans with teams and athletes, but also with other fans. We can find chatbots, ticketing, merchandise solutions, which are based on sentiment analysis and historical data analysis.<br>Another widely used technology is video analysis and motion tracking systems. which in addition to offering data on the performance of athletes, can enhance the fan experience. It is the case of the platform subscription formula, that though watchers micropayments, automatically provide highlights of matches, replays, and commentary. This happens thanks to Machine Learning technology being able to identify the best in a match, capturing the main events, and offering them to fans, without the need to see all the game. Also, Augmented reality is used slightly less to allow stadium visitors to see live match information and player statistics.</p><h3>E-sport — Gaming</h3><p>The last category which includes end-to-end content solutions for sports leagues that leverages computer vision and deep learning is about monetization platforms that feature gamification elements. Some fan platforms enable viewers to play along during live broadcasts, archive, and on-demand content (quizzes, predictions, challenges…). Alternatively, these platforms and mobile apps offer personal performance analytics for competitive gamers and e-sports.</p><h3>Conclusion</h3><p>In conclusion, the trend that emerges from the analysis of the sector is that AI and Machine Learning are increasingly requested by fans for a satisfying sports experience and by clubs and federations for a more objective and detailed analysis of their players. This shows that there is a profitable market for organizations willing to invest in “connected” stadiums and events.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c092c6c45bcd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI for Business Growth (Part 2)]]></title>
            <link>https://medium.com/@sparkdai-hq/ai-for-business-growth-part-2-8bec8b72aeba?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/8bec8b72aeba</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[machine-intelligence]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 20 Nov 2019 09:31:01 GMT</pubDate>
            <atom:updated>2019-11-20T09:31:01.189Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FCguNiBQQDGz91r8g51l8w.jpeg" /></figure><p>This is Episode 2 of the “Embedding AI in Business” review: a key conversation between Paul Daugherty, Accenture’s Chief Technology and Innovation officer and Azeem, founder of the exponential view.</p><p>This exchange was so dense of lessons learned that we simply could not fit into a single post!</p><p>Here’s the lessons learned of this episode:</p><p><strong>1. AI has arrived as a torrent.</strong></p><p>AI has grown so fast in the enterprise world, for three main reasons: firstly, through AI we can solve customer problems faster. Secondly, data and infrastructure to implement AI is already there. Finally, access to AI is not democratised: all the discoveries in the areas of Machine Learning and Computer Vision</p><p><strong>2. Start with AI now and beat competition with the data network effect</strong></p><p>Machine Learning needs data to learn models from. If your models are performing well in your products, customers will use them more. This will result in more data created by your customers that you can use to further improve your AI. This is what we call AI locking loop: A reinforcing effect, a winner-take-all approach that privileges the companies that start early with AI.</p><p>If there’s no data at the core of the DNA of your company, you will struggle in the medium and longer term to compete with the players leveraging the data network effect.</p><p><strong>3. Uses Machine Learning in every part of your business.</strong></p><p>Amazon adopted this strategy early after the Deep Learning revolution in 2012. Jeff Bezos in one famous memo stated: “ we need to embed AI in everything we do”.</p><p>They started with a program that would train all their managers and senior execs on the capability of Machine Learning and AI. Moreover, in business plans of every amazon program manager they introduced a mandatory item: how can I improve my part of the business through AI.</p><p><strong>4. Next wave of AI business value is created by combining Symbolic AI and Machine Learning</strong></p><p>For a lot of business applications, especially in verticals such as Manufactoring and Talent Management &amp; HR, the name of the game is to take into previous knowledge and rules devised by human experts. For this reason, Adam and Azeem argue that traditional AI methods as symbolic graph reasoning are regaining a “raison d’etre” and complement more modern learning techniques.</p><p>Listen to the <a href="https://overcast.fm/+HiEbbzd1o">full podcast</a>.</p><p>Deep dive with more readings :</p><ul><li><a href="https://hbr.org/2019/01/the-future-of-ai-will-be-about-less-data-not-more">“The Future of AI Will Be About Less Data, Not More”</a> (HBR, Jan. 14, 2019)</li><li><a href="https://newsroom.accenture.com/news/accenture-interactive-launches-groundbreaking-artificial-intelligence-solution-to-tackle-elderly-loneliness.htm">“Accenture Interactive Launches Groundbreaking Artificial Intelligence Solution to Tackle Elderly Loneliness”</a> (Accenture, April 30, 2019)</li><li><a href="https://hbr.org/2019/04/using-ai-to-make-knowledge-workers-more-effective">“Using AI to Make Knowledge Workers More Effective”</a> (HBR, April 19, 2019)</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8bec8b72aeba" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI for business growth (Part 1)]]></title>
            <link>https://medium.com/@sparkdai-hq/ai-for-business-growth-part-1-a14edc9b4ed8?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/a14edc9b4ed8</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 13 Nov 2019 09:31:01 GMT</pubDate>
            <atom:updated>2019-11-13T09:31:01.549Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*Pqrh2eIDWBSHdoio09GZ3w.jpeg" /></figure><p>AI is the fastest-growing industry that I have seen in the last thirty years of my career”. This is one of the opening statements of Paul Daugherty, Accenture’s Chief Technology and Innovation officer.</p><p>We re-listened to this fantastic conversation with Azeem from the Exponential view and gained a lot of insights on how AI will force us to redesign the world we live in.</p><p>This is what we retained:</p><p><strong>1. AI is just an approximation of human abilities.</strong></p><p>While many are talking about Generic Artificial Intelligence making humans irrelevant, Paul argues that AI is a set of capabilities covering three broad categories: sensing, comprehension and learning. Seen in this way, AI extends humans and allows them to do old things in new ways.</p><p><strong>2. AI releases us from the precise and complicated logic of Computers</strong></p><p>How many of us are still struggling on a daily bases with complicated Excel sheets or with out-of-date software difficult to use? Not anymore.. Ai is liberating us from all this complication. By being “smarter” than traditional software, it is more flexible and more “human” in the interaction. More specififcally, it allows us to draw usable insights from squiggly, ambiguous data as images, speech and natural language.</p><p><strong>3. AI is the new Electricity..</strong></p><p>Embracing AI might mean re-writing the very nature of your business, Azeem argues that AI transformation will be similar to transition to <a href="https://en.wikipedia.org/wiki/Electrification">electrified factories</a> that took place around 1900.</p><p>Prior to electrification, factories were powered by big steam engines. They were releasing energy in a large lump and it was difficult to manage. As a consequence tools were big and distributed around the steam engine.</p><p>Electricity enabled to flow power into small amounts and distribute it over different parts of the factory. As a consequence, throught electricity, we were able to break down processes, modularise and speed-up production chains.</p><p>AI will have the same transformational power to unbundle complex, monolithic processes and favour automation through a variety of different “smart” tools across organizations.</p><p>Listen to the <a href="https://overcast.fm/+HiEbbzd1o">full podcast</a>.</p><p>Deep dive with more readings :</p><ul><li><a href="https://hbr.org/2019/01/the-future-of-ai-will-be-about-less-data-not-more">“The Future of AI Will Be About Less Data, Not More”</a> (HBR, Jan. 14, 2019)</li><li><a href="https://newsroom.accenture.com/news/accenture-interactive-launches-groundbreaking-artificial-intelligence-solution-to-tackle-elderly-loneliness.htm">“Accenture Interactive Launches Groundbreaking Artificial Intelligence Solution to Tackle Elderly Loneliness”</a> (Accenture, April 30, 2019)</li><li><a href="https://hbr.org/2019/04/using-ai-to-make-knowledge-workers-more-effective">“Using AI to Make Knowledge Workers More Effective”</a> (HBR, April 19, 2019)</li><li><a href="https://www.technologyreview.com/s/613281/google-cancels-ateac-ai-ethics-council-what-next/">“Hey Google, sorry you lost your ethics council, so we made one for you”</a> (MIT Technology Review, April 6, 2019)</li><li><a href="https://blog.usejournal.com/from-principles-to-action-how-do-we-implement-tech-ethics-d59aebb05ed8?sk=d01ca6f6d827b44414d34730dd920ab2">“From Principles to Action: How do we Implement Tech Ethics?”</a> (Industry Ethicists, April 17, 2019)</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a14edc9b4ed8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Cracking Creativity with Computers: How AI is Learning to Write, Paint and Think]]></title>
            <link>https://medium.com/@sparkdai-hq/cracking-creativity-with-computers-how-ai-is-learning-to-write-paint-and-think-8292c3209151?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/8292c3209151</guid>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 06 Nov 2019 09:31:03 GMT</pubDate>
            <atom:updated>2019-11-06T09:31:03.271Z</atom:updated>
            <content:encoded><![CDATA[<p>We will let you down. Despite all the buzz and all the sensational claims, a machine learning algorithm cannot beat Picasso!</p><p>In this great chat, Marcus du Sautoy, the author of <a href="https://www.amazon.co.uk/Creativity-Code-Marcus-du-Sautoy/dp/0008288151">The Creativity Code</a> and Oxford professor, discusses with John Tornhill (Financial Times — Tech columnist) about cracking creativity with computers.</p><p>Here’s what we retained:</p><ol><li><strong>AI can help humans to find new ways to unleash creativity</strong>.</li></ol><p>We believe that we are at the pinnacle of creativity but we aren’t. There are unlimited ways in which human creativity can express itself. The good news? AI can help to dissipate the clouds over the next creativity peak standing in front of us!</p><ol><li><strong>AI has already exhibited an extraordinary level of creativity in a number of different famous examples</strong>.</li></ol><p>While there’s consensus among scientists about the inability of an AI algorithm to replace human creativity, in very specific problems, Machine Learning models have “invented” novel ways to solve them.</p><p>The most notorious is the “Move 37” in the <a href="https://www.wired.com/2016/03/two-moves-alphago-lee-sedol-redefined-future/">Alpha go game against World champion Lee Sedol</a>. As a consequence of that revolutionary move, we play go on a completely different way.</p><p><strong>Less Artificial intelligence and more Augmented Intelligence.</strong></p><p>AI is moving towards more cooperation between humans and machines, it is not mere replacement.</p><p>This is especially interesting for complex tasks where the machine can’t operate on its own yet and the human is struggling on its own. Fascinating opportunities in this area of human-machine interaction.</p><ol><li><strong>How many types of creativity?</strong></li></ol><p>Marcus provides these three categories of creativity, that it is helpful to keep in the back of your mind for your next creative project!</p><ol><li><em>Exploratory</em> — take something and push it to the extreme (e.g. Bach).</li><li><em>Combinatory creativity</em> — apply the style of one composer to another, or mix elements of different styles (e.g. fusion cuisine).</li><li><em>Transformational creativity</em> — break the rules, have a moment of genius!</li><li><strong>Computers and AI struggle being creative with text.</strong></li></ol><p>Despite what you might think, for a computer, it is easier to create from scratch a realistic image than a piece of text.</p><p>In fact, with text, you have to deal with a lot of nuances, while music and images are easier to create because they are full of patterns that can be replicated.</p><p>For instance, think about humor: it mixes presentation, performance skills with linguistic ambiguity and deep understanding of human psychology. Not so easy for a computer!</p><p>Listen to the episode on <a href="https://overcast.fm/+He0sOZz0U">Overcast</a></p><h3>OUR SUGGESTIONS FOR FURTHER READING</h3><ul><li>buy <a href="https://www.amazon.co.uk/Creativity-Code-Marcus-du-Sautoy/dp/0008288151">The Creativity Code</a> book, seems really good ( 4.27 / 5 Goodread score, v high but only 42 votes)</li><li>check out some examples of “Artificial Creativity” in t<a href="https://deepdreamgenerator.com/#tools">he deep dream tools</a></li><li>know more about the key technology behind Artificial Creativity, watch <a href="https://www.youtube.com/watch?v=C1YUYWP-6rE&amp;t=156s">this short YouTube Video</a> about GANs (Generative Adversarial Neural Networks)</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8292c3209151" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Tensor flow 2.0]]></title>
            <link>https://medium.com/@sparkdai-hq/tensor-flow-2-0-6e3e038bd72a?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/6e3e038bd72a</guid>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[tensorflow]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 30 Oct 2019 09:31:02 GMT</pubDate>
            <atom:updated>2019-10-30T09:31:02.532Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*GMdJ2h5dosPs8yqWFe_KjA.jpeg" /></figure><p><strong>Flexibility, flexibility and flexibility.</strong></p><p>While the most common case is by far, transfer learning (e.g. adapt a pre-trained model to your problem) TF offers flexible solutions for users with different Deep Learning skills, if you are a researcher you can design your own layer, while if you are a developer with limited time, you can use a pre-trained model (e.g. Inception or Resnet ).</p><ol><li><strong>[NEW] — Keras is the easiest way to interact with tensor flow</strong>.</li></ol><p>The strategy of google is clear: more of the same.. enabling non-expert users to get their transfer learning use cases done will become even easier through Keras.</p><ol><li><strong>[NEW] — Monolithic architecture is unbundled without breaking compatibility</strong>.</li></ol><p>TF 2.0 has better integration among elements, and more harmonization between the different tools and projects so that you will be able to seamlessly run models on cluster or mobile.</p><ol><li><strong>[NEW] — Improved Portability</strong>.</li></ol><p>While you already have a lot of options at the moment, it is still very challenging to move models around. With TF 2.0 there is more cohesiveness in the ecosystem with the ability to port models from TS to JS, to TF extended and TF mobile.</p><ul><li>Listen to the entire <a href="https://overcast.fm/+OcVdvvtHA">podcast on overcast</a></li></ul><h3>OUR SUGGESTIONS</h3><ul><li>read some of the TensorFlow case studies published in the <a href="https://www.tensorflow.org/about/case-studies/">Google Blog</a></li><li>watch the <a href="https://www.youtube.com/watch?v=2FmcHiLCwTU">introduction to Tensor Flow</a> by Siraj Raval. Fun and very interesting!</li><li>experiment with Tensor Flow in a few clicks by running <a href="https://colab.research.google.com/notebooks/mlcc/first_steps_with_tensor_flow.ipynb?utm_source=mlcc&amp;utm_campaign=colab-external&amp;utm_medium=referral&amp;utm_content=firststeps-colab&amp;hl=en">this colab notebook</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6e3e038bd72a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The three years test: Our prediction check on Fei-fei Li discussion with A16z on A.I.]]></title>
            <link>https://medium.com/@sparkdai-hq/the-three-years-test-our-prediction-check-on-fei-fei-li-discussion-with-a16z-on-a-i-7a61f72fa565?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/7a61f72fa565</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 23 Oct 2019 09:31:02 GMT</pubDate>
            <atom:updated>2019-10-23T09:31:02.082Z</atom:updated>
            <content:encoded><![CDATA[<h3>The three years test: Our prediction check on Fei-fei Li discussion with A16z on A.I. | Autonomous Vehicles and Startups</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*o10nvWQRwrajWaAkyCzbvg.jpeg" /></figure><p>We re-listened to “<a href="https://overcast.fm/+BlzGXgYK0">When Humanity Meets A.I.</a>”, a podcast featuring Prof. Fei-Fei Li, from Stanford University (@drfeifei) and Frank Chen from leading VC firm a16z, (@withfries2).</p><p>This podcast was recorded almost three years ago, in June 2016! That’s why we noted the main predictions that were made at the time and then we fact checked them.. here’s our results:</p><p><strong>Prediction n.1: A new generation of Deep Leaning chips will blossom.</strong></p><p>Dedicated they will exploit redundancy in operations to optimise computation in a paralysed way.</p><p>⇒ [<strong>VERIFIED</strong>] — Movidius launched a <a href="https://techcrunch.com/2017/07/20/movidius-launches-a-79-deep-learning-usb-stick/">79$ Deep Learning USB stick</a> in 2017, and today Deep learning chipsets are gaining market traction.</p><p><strong>Prediction n.2: Dualism between Unsupervised and Supervised Learning</strong> <strong>will be resolved</strong>.</p><p>By automating the generation of labels, one of the biggest pains of Supervised Learning will be removed. As a consequence the gap between Supervised and Unsupervised Learning will be reduced.</p><p>⇒ [<strong>VERIFIED</strong>] — Many works were published in the last three years covering techniques such as <a href="https://www.youtube.com/watch?v=7I0Qt7GALVk">self-supervised learning</a> and synthetic <a href="https://arxiv.org/pdf/1801.06397.pdf">data generation</a>. Prof. Fei-Fei was spot on!</p><p><strong>Prediction n.3: Machines will learn to learn.</strong></p><p>Machines will learn like humans: incrementally and iteratively. They will learn to explore knowledge, to ask humans the right questions in the right order: No more annotations will be needed, machines will learn by example.</p><p>⇒ [<strong>UNVERIFIED</strong>] — While a lot of papers were recently published in conferences such as <a href="http://cvpr2019.thecvf.com/">CVPR</a>, <a href="http://iccv2019.thecvf.com/">ICCV</a> and <a href="https://nips.cc/">NIPS</a> on the subject we are still waiting for this complex technology to mature.</p><p>Listen to the full podcast on <a href="https://www.notion.so/lucamarchesotti/9b183b74e9df4a1ebd07da48edb7122b?v=5803ba5ea96d486388508cccc40abbf4&amp;p=5fa07e7da34749bbb4eaafc04f6fb6a0">Itunes</a>, <a href="https://overcast.fm/+BlzGXgYK0">Overcast</a>, <a href="https://www.notion.so/lucamarchesotti/9b183b74e9df4a1ebd07da48edb7122b?v=5803ba5ea96d486388508cccc40abbf4&amp;p=5fa07e7da34749bbb4eaafc04f6fb6a0">Spotify</a></p><p><strong>Discover more</strong></p><ul><li>Check out the research of <a href="http://vision.stanford.edu/">Stanford Vision and Learning lab</a></li><li>Get inspired by the <a href="https://www.womeninai.co/">Women in Ai initiative</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7a61f72fa565" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The future of Autonomous Vehicles — bold predictions from field experts and top-notch VCs]]></title>
            <link>https://medium.com/@sparkdai-hq/the-future-of-autonomous-vehicles-bold-predictions-from-field-experts-and-top-notch-vcs-e9e88da0a51c?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/e9e88da0a51c</guid>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 16 Oct 2019 09:31:01 GMT</pubDate>
            <atom:updated>2019-10-16T09:31:01.862Z</atom:updated>
            <content:encoded><![CDATA[<h3>The future of Autonomous Vehicles — bold predictions from field experts and top-notch VCs</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*e7RQgUv_1g9BPgzVy2ALlw.jpeg" /></figure><p>As Autonomous Vehicles (AVs) are a major driver of Artificial Intelligence Use Cases we dug into this fascinating discussion about the state of AV and the challenges the near future will bring to the automotive industry.</p><p>Detroit, Silicon Valley, Ford, Tesla… Sonal Chokshi from a16z one of the leading VC in US interviews two expe</p><p>Here’s our main lessons learned from this great episode:</p><ol><li><strong>The current AV industry is the exact replica of the old Automotive industry.</strong> If look back at titans of the car industry in the US you think about geniuses such as Alfred P. Sloan, Henry Ford. Today we have Elon Musk and Jeff Bezos, but we have the same patterns of innovation.</li><li><strong>To predict the trends of the autonomous vehicles industry look at how mobile phone innovation has worked in the last ten years</strong>. Don’t expect a defining autonomous vehicle movement. There was no such moment for mobile phones aside from in 2012/2013 with the flourishing global success of Snapchat and Instagram.</li><li><strong>We are laying down the train tracks while at the same time inventing the trains</strong>. Old trivial problems like fitting air conditioning into a car will need to be solved again when electric vehicles take over. The very nature of a car will change, and almost every single component of a vehicle will be impacted.</li><li><strong>Expect a renaissance of the automotive market</strong>: The AV industry will see emerging a universe of new companies broadly distributed into two categories: Software companies and infrastructure companies, like Bosch for instance, that provides the infrastructure for AVs to operate in the physical environment.</li><li><strong>AV innovation will have different pace, depending on use case by use case basis</strong>. Different use cases will develop in parallel, at different speeds, in different market segments (E.g. bus shuttles, robo taxis, autonomous trucks, or agriculture trucks) depending on customers’ needs and complexity of the tasks. Many say that Autonomous trucks will be the first to hit the market.</li><li><strong>Protectionism, protectionism, protectionism !</strong> Governments will be much more protectionist towards autonomous vehicles than with other internet tech. The AV industry is of national and also regional interest: governments will want the AV technology ecosystem growing up locally, all around the automotive industry, instead of being delocalised as other internet services.</li></ol><p>Listen to the <a href="https://a16z.com/2019/02/23/autonomous-vehicles-state-of-autonomy-2019-mobile-commodity-capitalism/">full podcast here</a></p><p><strong>Read more</strong></p><ul><li>discover what the <a href="https://www.wired.com/story/trolley-problem-teach-self-driving-car-engineers/">“Trolley problem”</a> taught us about Self driving cars. (SPOILER ALERT…Information flows in human brains (and bodies) very slowly while computers have the time to plan the best action in advance in case a collision happens, or a big problem concerning the ethics of AI: e.g. who should the machine kill?</li><li>guess how much it would cost to make a sandwich on your own… and then read this article about the (<a href="https://www.theguardian.com/lifeandstyle/2015/sep/23/1500-sandwich-from-scratch-andy-george-youtube-food-supply-chain">Sandwich experiment</a>. (SPOILER ALERT: The development of an ecosystem of companies around the AV industry is a natural way to spread risk and to supercede inefficiencies. While full verticalization is expensive (the sandwich experiment link), it’s about the convenience of unbundling costs. Stuff that you don’t have on the market. It’s about reducing inefficiencies, such as in the example of the sandwich, doing it yourself costs much less. )</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e9e88da0a51c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Ethics of Artificial Intelligence]]></title>
            <link>https://medium.com/@sparkdai-hq/the-ethics-of-artificial-intelligence-1ba4e195fd6f?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/1ba4e195fd6f</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 09 Oct 2019 09:31:01 GMT</pubDate>
            <atom:updated>2019-10-09T09:31:01.330Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*Cz-OmFL6Ksi7tbrM6ThJNA.jpeg" /></figure><p>The way we develop the next generation of learning software will have a tremendous impact on our life. Prof. Joanna Bryson (<a href="https://twitter.com/j2bryson?lang=en">@j2bryson</a>) from University of Bath dissected this fascinating topic with Azeem Azhar (<a href="https://twitter.com/azeem?lang=en">@azeem</a>), founder of The Expontial view in <a href="https://hbr.org/podcast/2019/04/designing-responsible-ai">this incredible podcast</a>.</p><p>Intelligence, ethics, and digital transformation they covered them all. Here’s our main lessons learned:</p><ol><li><strong>AI is the pen and paper of the 21st Century.</strong> Artificial Intelligence and Machine learning are the “<em>Cognitive prosthetics that enable humans to offload information to the cloud</em>”. Through voice recognition, photos and smart devices we are able to save a big chunk of our life on the internet. AI and Machine learning give us the power to record but also to access all that data!</li><li><strong>Bias is one of the most burning problems with AI from an ethical perspective</strong>. This issue is being heavily researched at the moment. If the results are successful, a weakness can be turned into a strength: AI and Machine Learning models can be used to detect, measure, and ultimately remove bias from human-centered processes!</li><li><strong>Explainability is a very serious and largely unresolved issue with Deep Learning systems</strong>. Understanding why a certain action was taken or suggested by a machine learning algorithm is not only important for understanding if a moral or ethical guideline was respected, but doing so has also been imposed by recent regulations such as GDPR.</li><li><strong>Artificial Intelligence can be put at good use if applied as a tool that enhances human intelligence</strong>.</li><li><strong>We need to maintain AI and Deep Learning models over time.</strong> As we do with other machines (e.g. cars), we need to check the status of an AI model to avoid bad data is indadvertedly introduced! Since these systems are continuing to learn, we need to ensure biases don’t affect our AI models.</li></ol><h3>OUR SUGGESTIONS FOR FURTHER READING</h3><ul><li>understand what the ethical issues we are facing are and the emergence of super-intelligence <a href="https://nickbostrom.com/ethics/ai.html">https://nickbostrom.com/ethics/ai.html</a></li><li>read one of Prof. Bryson’s seminal articles: <a href="https://science.sciencemag.org/content/356/6334/183">“Semantics derived automatically from language corpora contain human-like biases”</a>(April 14, 2017)</li><li>get tense by reading the <a href="https://wiki.lesswrong.com/wiki/Paperclip_maximizer">paperclip example</a>: which is about how AI, that is apparently innocuous, can pose a very serious existential threat</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1ba4e195fd6f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The state of Artificial Intelligence]]></title>
            <link>https://medium.com/@sparkdai-hq/the-state-of-artificial-intelligence-19efc6aed8c1?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/19efc6aed8c1</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 02 Oct 2019 09:31:03 GMT</pubDate>
            <atom:updated>2019-10-02T09:31:03.068Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*CTFXBQ3Po5-KtEUwDOyyNg.jpeg" /></figure><p>You might have heard about Open AI: Elon Musk initiative to address big AI challenges.</p><p>At the end of last year, following the publication of the AI Index Report (a dense outlook on the state of AI) Azeem Azhar, technologist, founder and host in <a href="http://www.exponentialview.co/">www.exponentialview.co</a> interviewed Jack Clark policy director at OpenAI.</p><p>This is what we retained from an incredibly rich discussion covering both the societal and geopolitical aspects of the AI revolution.</p><ol><li><strong>Artificial Intelligence Monks vs. AI peasants</strong>. The staggering speed of the AI transition is causing tension between AI Monks (AI researchers) and AI Peasants (AI practitioners). AI researchers have kept the flame of AI alive for more than two decades following the AI winter in the ’80s. However, it is hard for them to redefine their role now that interest in AI has surged. The upside is that AI practitioners make the AI and Machine Learning landscapes more diverse, heterogeneous, and vibrant.</li><li><strong>Speed is the most defining feature of the wave of AI innovation:</strong> while mobile phones and the internet took a decade to thrive, AI is ramping up much faster. This was confirmed by a recent statement by the CTO of Accenture who declared, “AI is the fastest growing IT trend I have seen in thirty years of [my] career.”</li></ol><p><strong>The AI lords and the data farmers.</strong> We live in an era of Neo-Feudalism where big data and AI companies (i.e. the lords) govern the rest of the people by providing some level of service and security (i.e. services like Google, Twitter, etc.) and by getting resources (i.e. data) in return. The big difference from Feudalism is that the resources in the past were food and hence they were consumed, whereas now it is data which can be infinitely re-used.</p><ol><li><strong>NEW TREND — 🚀Every Data center can be turned into a configurable, gigantic learning machine.</strong> The factories where AI is made (i.e. Data centers) are now gigantic, multi-purpose computational machines that can be reconfigured depending on the problem you need to solve. You can re-train the same deep neural network stack to understand language, to recognize images, or move a robotic arm.</li><li><strong>Deep learning researchers are GPUs greedy.</strong> The human brain consumes 25 Watts; a GPUs consumes 10 times as much. At Google and Alibaba, DeepMind researchers are training deep neural networks with thousands of GPUs! At the moment, the most common approach towards Deep Learning is brute force because we are still at an exploration phase, and computational resources are still relatively cheap.</li></ol><h3>OUR SUGGESTIONS</h3><ul><li>Discover the numbers behind the staggering growth in interest about AI in the <a href="https://aiindex.org/">AI Index Report 2018</a>.</li><li>Read more about how <a href="https://ai.googleblog.com/2018/03/using-evolutionary-automl-to-discover.html">computers can discover the best neural network architectures</a> without human intervention (machines that #learn2learn).</li><li>Be surprised at the super steep rate of improvement in language understanding by checking out the <a href="https://gluebenchmark.com/">GLUE benchmarking</a> (from 66% in May ’18 to 80% in October ’18; the non-expert human baseline is 90%).</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=19efc6aed8c1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A powerful voice from the Intellectual Dark Web talks about A.I., Science and Math]]></title>
            <link>https://medium.com/@sparkdai-hq/a-powerful-voice-from-the-intellectual-dark-web-talks-about-a-i-science-and-math-152a2743c038?source=rss-92061bf597c4------2</link>
            <guid isPermaLink="false">https://medium.com/p/152a2743c038</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Sparkd.ai]]></dc:creator>
            <pubDate>Wed, 25 Sep 2019 09:31:02 GMT</pubDate>
            <atom:updated>2019-09-25T09:31:02.590Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*1mRQkViiE3h2eUX8QT59xQ.jpeg" /></figure><p>Eric Weinstein, known for his contributions to Economic Theory and Mathematical Physics, recently talked to Lex Friedman from MIT.</p><p>This multifaceted, complex conversation touched the philosophical and societal implications of AI. It also diverged into Trumpism, Social Media and String Theory. Here’s our key learnings.</p><h3>1. GENERAL ARTIFICIAL INTELLIGENCE IS ALREADY HERE.</h3><p>Recently, in several examples (e.g. alpha-zero generating incredibly powerful and novel chess moves), AI is already exhibiting superior levels of creativity comparable to human creativity.</p><h3>2. WE HAVE SMARTER TECH AND AI, BUT WE AIN’T WISER IN THE WAY WE USE IT..</h3><p>Since World War 2, we have been living in a remarkably stable period where no major global systemic failures have happened.</p><p>The two worrying aspects of this situation is that: a) the technological sophistication of tools such as AI has soared but collecting wisdom on how to use them hasn’t, b) younger generations think that they are operating in a video game where the catastrophe scenario (nuclear intervention) will never really happen.</p><h3>3. ARTIFICIAL INTELLIGENCE (AND SOFTWARE) IT IS ALREADY REPLICATING ITSELF</h3><p>AI and more broadly speaking Software, can already replicate itself in an evolutionary way. To produce selective Software systems that self-replicate you need three ingredients: 1/ Heritability, 2 / Variability of population, 3/ Differential success. All these ingredients have a digital counterpart. #scary</p><h3>4. SOCIAL MEDIA IS A NEW EMERGING FORM OF SYNTHETIC PARASITISM</h3><p>In the natural world, parasitism is the main vehicle through which low intelligence species can outwit higher intelligence species and propagate their line of succession. However, in many cases, parasitism is indistinguishable from symbiosis: species co-exist by providing marginal mutual benefits to each other.</p><p>In the digital world, social media is evolving as a new form of intelligence co-existing with humans. Social media provides mild benefits but it is also highly addictive. Moreover, it has its own goal: creating human behaviours that optimise revenue. Considering all this social media scarily fits into a parasite / symbiotic profile!</p><p>Listen to the episode on <a href="https://www.notion.so/lucamarchesotti/9b183b74e9df4a1ebd07da48edb7122b?v=5803ba5ea96d486388508cccc40abbf4&amp;p=5fa07e7da34749bbb4eaafc04f6fb6a0">Itunes</a>, <a href="https://www.notion.so/lucamarchesotti/9b183b74e9df4a1ebd07da48edb7122b?v=5803ba5ea96d486388508cccc40abbf4&amp;p=5fa07e7da34749bbb4eaafc04f6fb6a0">Spotify</a>,<a href="https://overcast.fm/+OcVcbCk4k">Overcast</a></p><p><strong>Read more</strong></p><ul><li>Read the story of <a href="https://eu.news-leader.com/story/sports/outdoors/2015/08/04/fascinating-story-mussels-reproduce/31133857/">Unionoid mussels</a>, an arguably low-intelligence species that is outwitting a superior species (swimming fish) as part of their reproductive process. By parasitising fish, they disperse their offspring over a much wider range than could be accomplished otherwise.</li><li>Discover why <a href="https://techcrunch.com/2019/02/17/openai-text-generator-dangerous/">OpenAI that built a text generator so good, it’s considered too dangerous to release</a> and reflect about the AI dangerous innovation dilemma: better containing such ideas or discussing them publicly?</li><li>Learn what makes a system exhibiting <a href="https://en.wikipedia.org/wiki/Artificial_general_intelligence">Artificial General Intelligence</a></li><li>Check out who is part of the <a href="https://bigthink.com/culture-religion/eric-weinstein-intellectual-dark-web">“intellectual dark web”</a> along with Eric Weinstein.</li><li>consider buying <a href="https://www.edge.org/conversation/john_brockman-possible-minds">Possible Minds, 25 Ways of Looking at AI by John Brockman (1.30.19)</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=152a2743c038" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>