<?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 Théodore Aouad on Medium]]></title>
        <description><![CDATA[Stories by Théodore Aouad on Medium]]></description>
        <link>https://medium.com/@theodorea?source=rss-67ee515d9e80------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*AlFq6CwT2XEZ8TVP.jpg</url>
            <title>Stories by Théodore Aouad on Medium</title>
            <link>https://medium.com/@theodorea?source=rss-67ee515d9e80------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 05:28:37 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@theodorea/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 of AI: New Articles Published This Month (July 2019)]]></title>
            <link>https://medium.com/sicara/07-2019-best-ai-new-articles-this-month-3e1fa3f6c321?source=rss-67ee515d9e80------2</link>
            <guid isPermaLink="false">https://medium.com/p/3e1fa3f6c321</guid>
            <category><![CDATA[best-of]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[articles]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[big-data]]></category>
            <dc:creator><![CDATA[Théodore Aouad]]></dc:creator>
            <pubDate>Mon, 12 Aug 2019 13:31:01 GMT</pubDate>
            <atom:updated>2019-12-06T13:37:48.222Z</atom:updated>
            <content:encoded><![CDATA[<h4>10 data articles handpicked by the Sicara team, just for you</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fLmaa4J-_6qr_IiUfc_p6g.png" /></figure><p>Welcome to the July edition of <strong>our best and favorite articles in AI that were published this month</strong>. We are a Paris-based company that does Agile data development. This month, we spotted <strong>articles about autonomous driving, artificial general intelligence, deepfakes and much more</strong>. Let’s kick off with <strong>the comic of the month</strong>:</p><figure><a href="https://xkcd.com/1132/"><img alt="" src="https://cdn-images-1.medium.com/max/468/1*jcTVH-5TiBjeiymMjxepEQ.png" /></a><figcaption>‘Detector! What would the Bayesian statistician say if I asked him whether the — ‘ [roll] ‘I AM A NEUTRINO DETECTOR, NOT A LABYRINTH GUARD. SERIOUSLY, DID YOUR BRAIN FALL OUT?’ [roll] ‘… yes.’</figcaption></figure><h3>1 —Facebook, Carnegie Mellon build first AI that beats pros in 6-player poker</h3><figure><a href="https://ai.facebook.com/blog/pluribus-first-ai-to-beat-pros-in-6-player-poker/"><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QqiOmJHn_AW2oY2sw4IJEw.png" /></a></figure><p>After conquering <strong>board games</strong> such as Chess and Go, <strong>video game</strong>s such as Dota, computers now aim to beat <strong>card games</strong>. If your ethics allow it, it is now possible to become rich by letting a <strong>program play Poker</strong> for you.</p><p>Programs that could win against professionals in a 1 vs 1 setting have been there for a while. This article presents <strong>Pluribus</strong>, the <strong>first AI able to beat human experts</strong> in six-player no-limit Hold’em, the most popular Poker format in the world. It was developed by <strong>Carnegie Mellon University (CMU) </strong>in collaboration with <strong>Facebook</strong>. If the bets were real, it would have won around $<strong>1000/hour</strong>. Does it motivate you to learn more about it?</p><p>Read <a href="https://ai.facebook.com/blog/pluribus-first-ai-to-beat-pros-in-6-player-poker/">Facebook, Carnegie Mellon build first AI that beats pros in 6-player poker</a> — from <a href="https://ai.facebook.com/">Facebook’s AI blog</a>.</p><h3>2 — Top 9 Ways Artificial Intelligence Prevents Fraud</h3><figure><a href="https://www.forbes.com/sites/louiscolumbus/2019/07/09/top-9-ways-artificial-intelligence-prevents-fraud/#5d5696b14b4f"><img alt="" src="https://cdn-images-1.medium.com/max/960/1*8Ri8fSb2xrmIP2PXAWx-8g.jpeg" /></a></figure><p>We have just seen a way of cheating in Poker, with a built-in program that can even beat professionals. Poker players, do not worry, as AI can also be used to <strong>fight fraud</strong>.</p><p><strong>Classic techniques</strong> to fight fraud consist of <strong>rule-based engines</strong> and simple predictive models. However, security breaches nowadays are getting more and more <strong>complex </strong>and the previous systems cannot compete. The countermeasure to the adaptation of scammers is to <strong>adapt the defense systems to them</strong>, using AI. But what does it mean exactly?</p><p>With AI, fraud prevention can use <strong>both</strong> <strong>past experiences</strong> and <strong>current trends</strong>. Moreover, it can do so in <strong>real-time</strong>, compared to previous methods which could take <strong>6 to 8 weeks</strong> to get the same results. Do you want to read more about the benefits of AI against fraud?</p><p>Read <a href="https://www.forbes.com/sites/louiscolumbus/2019/07/09/top-9-ways-artificial-intelligence-prevents-fraud/#5d5696b14b4f">Top 9 Ways Artificial Intelligence Prevents Fraud</a> — by <a href="https://www.forbes.com/sites/louiscolumbus/#718f310439e0">Louis Colombus</a>.</p><h3>3 — Microsoft Invests In and Partners with OpenAI to Support Building Beneficial AGI</h3><figure><a href="https://news.microsoft.com/2019/07/22/openai-forms-exclusive-computing-partnership-with-microsoft-to-build-new-azure-ai-supercomputing-technologies/"><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9M8GccHt9z3M0aGW.jpg" /></a><figcaption>Sam Altman, CEO of OpenAI (left), and Microsoft CEO Satya Nadella</figcaption></figure><p>The pinnacle of AI, the one we see in every futuristic movie with intelligent robots, is the concept of <strong>Artificial General Intelligence (AGI)</strong>. It represents a machine that has the <strong>same learning abilities as humans do</strong>. It could even <strong>surpass us</strong> by making <strong>connections across disciplines</strong> humans never saw.</p><p>AGI is a controversial term for many researchers. How can we define “general intelligence”? Is human intelligence even that “general”? To answer these questions and advance on this domain, <strong>OpenAI has received a $1 billion investment from Microsoft</strong>. They aim to tackle the <strong>technical difficulties</strong> of building an AGI and ensure a <strong>safe and secure</strong> deployment in society.</p><p>This partnership also implies that a lot of current OpenAI techs will become <strong>licensed</strong>. However, I believe that <strong>combining resources and intellect</strong> with some of the best researchers in AI can lead to many <strong>exciting discoveries</strong>.</p><p>Read <a href="https://news.microsoft.com/2019/07/22/openai-forms-exclusive-computing-partnership-with-microsoft-to-build-new-azure-ai-supercomputing-technologies/">OpenAI forms exclusive computing partnership with Microsoft to build new Azure AI supercomputing technologies</a> — from <a href="https://news.microsoft.com/">Microsoft News Center</a>.</p><h3>4 —Introducing the Plato Research Dialogue System: A Flexible Conversational AI Platform</h3><figure><a href="https://eng.uber.com/plato-research-dialogue-system/"><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*2SI7vr_odRNI7kVj.png" /></a></figure><p>Over the past decade, we have gone from <strong>looking</strong> at computer screens to <strong>interacting</strong> with them thanks to <strong>touch-screen technology</strong>. Children often believe that they can interact with every screen they see, because they grew up with it. Well, for tomorrow’s children, they may grow used to<strong> talking to the electronic devices around them</strong>.</p><p><strong>Uber </strong>just<strong> </strong>developed <strong>Plato Research Dialogue System </strong>to build, train and deploy conversational AI agents. It helps deal with the <strong>different steps</strong> required for these agents to interact with humans. It is designed to be <strong>modular</strong> and <strong>flexible</strong>. Now, everyone, <strong>experienced or not</strong> with this domain, can play around and create their <strong>own conversational agents</strong>.</p><p>Read <a href="https://eng.uber.com/plato-research-dialogue-system/">Introducing the Plato Research Dialogue System: A Flexible Conversational AI Platform</a> — from <a href="https://eng.uber.com/plato-research-dialogue-system/">Uber’s engineering blog</a>.</p><h3>5 — Conference on Uncertainty in AI</h3><figure><a href="http://auai.org/uai2019/"><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*apL33k0AuWnQs8h1.png" /></a></figure><p><strong>Conferences</strong> are wonderful to discover the <strong>latest development</strong> in AI. You might know some famous ones such as <strong>CVPR</strong> or <strong>ICLR</strong>. This month, the <strong>Conference on Uncertainty in Artificial Intelligence (UAI)</strong> took place in Tel Aviv.</p><p>Studies on <strong>uncertainty</strong> have increased a lot in the past years. Indeed, we want our models to be able to deal with <strong>intrinsic ambiguity</strong>. For example, let’s say I trained a <strong>neural network</strong> to classify different <strong>breeds of dogs</strong>. If I input an image of a <strong>cat</strong>, the model will output a dog’s breed, while I‘d rather have the network <strong>refusing</strong> to give a prediction.</p><p>This was only an example of how we could use uncertainty and probability in AI. Researchers at the <strong>University of Helsinki</strong> gave a method to sample from <strong>Directed Acyclic Graphs (DAG)</strong>. They won the <strong>best student paper award</strong>. If you are interested in their method and familiar with DAGs, you can read their paper: <a href="http://auai.org/uai2019/proceedings/papers/345.pdf">Exact Sampling of Directed Acyclic Graphs from Modular Distributions</a>. You can also read the winner of the <strong>best paper award</strong>: <a href="http://auai.org/uai2019/proceedings/papers/144.pdf">General Identifiability with Arbitrary Surrogate Experiments</a>.</p><p>If you are interested in probability theory and AI, then you can dive deeper into the different papers and talks given at this conference. See their website: <a href="http://auai.org/uai2019/">The Conference on Uncertainty in AI</a>.</p><h3>6 —Open-sourcing CraftAssist, a platform for studying collaborative AI bots in Minecraft</h3><figure><a href="https://ai.facebook.com/blog/craftassist-platform-for-collaborative-minecraft-bots/"><img alt="" src="https://cdn-images-1.medium.com/max/561/1*ZYX0aC79jCKbt3wJKLWDKQ.png" /></a></figure><p>Almost all AI programs are <strong>specialized</strong>. They can <strong>see</strong> with <strong>Computer Vision</strong>, they can <strong>hear</strong> with <strong>Speech Recognition</strong>, understand <strong>language</strong> with <strong>Natural Language Processing</strong>. Technologies as vocal assistants have appeared to <strong>combine</strong> all these specialties into one <strong>unique</strong> agent.</p><p>Like every milestone, we can climb step by step and first solve an easier problem. To approach this goal of <strong>global assistant AI</strong>, <strong>Facebook</strong> researchers first conquered the famous video game <strong>Minecraft</strong>. They released <strong>CraftAssist</strong>, a platform to create <strong>bots</strong> able to <strong>assist</strong> the users by <strong>performing various difficult tasks</strong>. We can <strong>interact</strong> with the bot via an in-game <strong>chat</strong>. They receive orders in <strong>natural language</strong>, understand them and execute them.</p><p>This has <strong>tremendous applications</strong> in a future where the bot would be a <strong>moving robot</strong>, the chat would be our <strong>voice</strong>, and the video game would be the <strong>real-world</strong>. CraftAssist is an <strong>open-source platform</strong>. You can even play with the framework to build <strong>your</strong> <strong>own assistants</strong>.</p><p>Read <a href="https://ai.facebook.com/blog/craftassist-platform-for-collaborative-minecraft-bots/">Open-sourcing CraftAssist, a platform for studying collaborative AI bots in Minecraft</a> — from <a href="https://ai.facebook.com/blog/">Facebook AI’s blog</a>.</p><h3>7 — Facebook AI Self-Supervision Challenge</h3><figure><a href="https://sites.google.com/view/fb-ssl-challenge-iccv19/home?fbclid=IwAR05mm_1Cf3Ny3pTOJPgf6QVgkIFSNtDzNv8HFcBhLDfLf4gfyzMr8KL0rs"><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*oe4HnDstZW3B6BkB.png" /></a></figure><p>Another way of pushing innovation in AI across the world, besides the release of <strong>public datasets</strong>, is to motivate people through <strong>challenges</strong>. On the other hand, <strong>unsupervised learning</strong> has been less studied than <strong>supervised learning</strong>, even though it is considered by many as the future of AI.</p><p>To balance out the equation, Facebook just announced the first session of the <strong>Facebook AI Self-Supervision Challenge</strong>, with a prize pool going up to $<strong>8000</strong>. Participants will compete in <strong>four different task</strong>s: three tasks of <strong>Image Segmentation</strong> and one of <strong>Object Detection</strong>. A bonus of $<strong>500</strong> is offered to winners who<strong> open-source their code</strong>. I believe it is another step in the right direction for research in AI. Sharing people’s work and research is always beneficial for science and knowledge.</p><p>Do you want to participate for the thrills of the competition, the price, or just follow the event? The competition started on July 25, 2019, and will close on October 15, 2019. Check out the challenge’s homepage: <a href="https://sites.google.com/view/fb-ssl-challenge-iccv19/home?fbclid=IwAR05mm_1Cf3Ny3pTOJPgf6QVgkIFSNtDzNv8HFcBhLDfLf4gfyzMr8KL0rs">Facebook AI Self-Supervision Challenge</a>.</p><h3>8 — The Best (And Scariest) Examples Of AI-Enabled Deepfakes</h3><figure><a href="https://www.forbes.com/sites/bernardmarr/2019/07/22/the-best-and-scariest-examples-of-ai-enabled-deepfakes/#4595c17e2eaf"><img alt="" src="https://cdn-images-1.medium.com/max/960/0*0ZWrKmNBqANlHLoy" /></a></figure><p>As we have seen at the beginning of this best of with Poker, AI can also be used for <strong>malicious intents</strong>. As it grows more <strong>effective</strong> and more <strong>dangerous</strong>, this raises the recurrent question, nonetheless <strong>important</strong>, of <strong>ethics in AI</strong>.</p><p>This article talks about different uses of <strong>Deepfake</strong>. This is the name we give to the different techniques of <strong>faking content using AI</strong>, like videos or recordings. A famous example is the<strong> video of Mark Zuckerberg </strong>released in June where he talks about controlling the future with data. This realistic video is indeed a <strong>total fake</strong> and raises awareness of the <strong>efficiency of the technology.</strong> It is becoming harder and harder to spot <strong>the real from the fake</strong>.</p><p>It is an important debate to have: <strong>should we still be pushing AI researches while such malicious applications exist?</strong> What do you think?</p><p>Read <a href="https://www.forbes.com/sites/bernardmarr/2019/07/22/the-best-and-scariest-examples-of-ai-enabled-deepfakes/#51219f2d2eaf">The Best (And Scariest) Examples Of AI-Enabled Deepfakes</a> — by <a href="https://www.forbes.com/sites/bernardmarr/#4e58fcb630c8">Bernard Marr</a>.</p><h3>9 — Lyft dataset for autonomous driving</h3><figure><a href="https://level5.lyft.com/dataset/"><img alt="" src="https://cdn-images-1.medium.com/max/600/1*-tWlCxHZ412MzymrbWCMbA.gif" /></a></figure><p><strong>Autonomous driving</strong> is still a milestone to overcome. The <strong>real world impact</strong> of driverless vehicles is <strong>limitless</strong>: from <strong>safety</strong> to <strong>energy consumption optimization</strong>, passing by <strong>comfort and time-saving</strong>. However, <strong>data</strong> has been <strong>lacking</strong> for self-driving cars, <strong>until now.</strong></p><p><strong>Lyft</strong> just released its <strong>brand-new dataset on autonomous driving</strong>. This is a major game-changer as it will help people from <strong>all around the world</strong> to move forward in this challenging task. Every time a big company shares its technology, it is a step in the right direction in the improvement of <strong>global human knowledge</strong>.</p><p>The dataset includes all types of information, from <strong>LiDAR raw inputs</strong> to high quality <strong>labeled bounding boxes of traffic agents</strong>. If you want to be the first to design a level 5 autonomous vehicle, or if, like me, you are just interested in the field, check out the <a href="https://level5.lyft.com/dataset/">lyft level 5 dataset</a>.</p><p><a href="https://level5.lyft.com/dataset/">Dataset | Lyft Level 5</a></p><h3>10 —ArchiGAN: a Generative Stack for Apartment Building Design</h3><figure><a href="https://devblogs.nvidia.com/archigan-generative-stack-apartment-building-design/?linkId=70968833"><img alt="" src="https://cdn-images-1.medium.com/max/625/1*FxRcRUqjXGd1cyA0iiGIpg.gif" /></a></figure><p><strong>GANs</strong> (Generative Adversarial Networks) have proven to be an effective <strong>generation paradigm</strong>. <strong>Researchers</strong> have delivered consistently on new results and experimentation with GAN, with<strong> landscape and face generation</strong> or even <strong>making</strong> <strong>photographs move</strong>.</p><p><strong>Nvidia</strong> began thinking about more <strong>applications to GAN</strong>. Stanislas Chaillou believes we can use AI to <strong>help architects</strong>. In this article, he presents an AI he built upon <strong>GAN</strong> and <strong>Pix2Pix</strong>, a model to perform <strong>image segmentation</strong>. His program can convert a <strong>simple sketch</strong> to plans that contain <strong>building footprints </strong>(basically where we can walk), the <strong>emplacement of doors, windows and walls</strong>, and can even give the <strong>furniture layout</strong> of the floor.</p><p>This model can help architects solve <strong>highly constrained problems</strong>. It is also a <strong>proof of concepts</strong> of what could be done in the future to help architects using AI.</p><p>Read <a href="https://devblogs.nvidia.com/archigan-generative-stack-apartment-building-design/?linkId=70968833">ArchiGAN: a Generative Stack for Apartment Building Design</a> — by <a href="https://devblogs.nvidia.com/author/sjchaillou/">Stanislas Chaillou</a>.</p><p><a href="https://devblogs.nvidia.com/archigan-generative-stack-apartment-building-design/?linkId=70968833">ArchiGAN: a Generative Stack for Apartment Building Design | NVIDIA Developer Blog</a></p><p>We hope you’ve enjoyed our list of the best new articles in AI this month. Feel free to suggest additional articles or give us feedback in the comments; we’d love to hear from you! See you next month.</p><p><a href="https://blog.sicara.com/03-2019-best-ai-new-articles-this-month-8580e0eb5f9b">Read the March edition</a><br><a href="https://blog.sicara.com/04-2019-best-ai-new-articles-this-month-d5bacff26ded">Read the April edition</a><br><a href="https://blog.sicara.com/05-2019-best-ai-new-articles-this-month-4a202e173e5a">Read the May edition</a><br><a href="https://blog.sicara.com/06-2019-best-ai-new-articles-this-month-d61c763b03db">Read the June edition</a></p><p>Read the original article on Sicara’s blog <a href="https://www.sicara.ai/blog/2019-08-12-best-ai-july-2019">here</a>.</p><h3>By the way, we published these articles on our blog in July:</h3><ul><li><a href="https://blog.sicara.com/image-registration-sift-deep-learning-3c794d794b7a">Image Registration: From SIFT to Deep Learning</a></li><li><a href="https://blog.sicara.com/tf-explain-interpretability-tensorflow-2-9438b5846e35">Introducing tf-explain, Interpretability for TensorFlow 2.0</a></li><li><a href="https://blog.sicara.com/meta-learning-for-few-shot-computer-vision-1a2655ac0f3a">Few-Shot Image Classification using Meta-Learning</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3e1fa3f6c321" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sicara/07-2019-best-ai-new-articles-this-month-3e1fa3f6c321">The Best of AI: New Articles Published This Month (July 2019)</a> was originally published in <a href="https://medium.com/sicara">Sicara&#39;s blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>