<?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 Cyrill Glockner on Medium]]></title>
        <description><![CDATA[Stories by Cyrill Glockner on Medium]]></description>
        <link>https://medium.com/@cglockner?source=rss-6840f087b8f3------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*eR4IpcqxwlnOOTSy_H_qpw.jpeg</url>
            <title>Stories by Cyrill Glockner on Medium</title>
            <link>https://medium.com/@cglockner?source=rss-6840f087b8f3------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 07:28:44 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@cglockner/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[Starting your GenAI Business]]></title>
            <link>https://medium.com/@cglockner/starting-your-genai-business-3797493181e4?source=rss-6840f087b8f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/3797493181e4</guid>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[startup]]></category>
            <dc:creator><![CDATA[Cyrill Glockner]]></dc:creator>
            <pubDate>Sun, 18 Aug 2024 18:19:04 GMT</pubDate>
            <atom:updated>2024-08-18T18:43:58.133Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dJTjf3siyzgso6gvh2qVPg.jpeg" /></figure><h3>A Blueprint for Future Founders</h3><p>Imagine building a company with a lean and minimalist culture where a small team augmented by AI-based workers delivers maximum customer value. This isn’t just a futuristic concept, it’s a reality you can achieve by following the approach outlined here. By bootstrapping your business and leveraging earned revenues, you can drive rapid growth fast.</p><p>Having spent years immersed in AI and startups, I’ve seen firsthand the incredible potential of GPT models. One of my “Eureka” moments was at a hackathon earlier this year, where half of the contestants proposed business ideas based on AI-powered services traditionally delivered by human consultants. These ranged from building renovation planning to the preparation for EU AI policies. When you consider Sequoia’s forecast that AI could potentially replace services with software — a market that starts at tens of trillions — it’s clear this trend is more than just anecdotal. This dynamic presents a significant opportunity for savvy startup founders.</p><p><strong>The Case for GPT-Powered Startups</strong></p><p>If you’re contemplating launching a business building on the impressive capabilities of GPT models, the timing couldn’t be better. Foundation models have reached remarkable capabilities, platform providers like Microsoft, Google and Amazon are ready to provide needed infrastructure, technology frameworks have reached a level of abstraction that makes them useful and easy to integrate, and interaction paradigms have matured significantly over the last 12 months.</p><h3>Here’s a blueprint to help you get started!</h3><h4>Identify High-Opportunity Areas:</h4><ul><li><strong>Consulting Services:</strong> Focus on consulting businesses that help clients comply with complex policies such as building regulations, ESG accounting, or new government mandates.</li><li><strong>Document-Intensive Work:</strong> Go after areas where professionals spend significant time reviewing vast amounts of documents to create compliance reports or recommendations.</li><li><strong>Specialized Knowledge:</strong> Target fields where highly trained professionals command high hourly rates due to their specialized expertise.</li></ul><h4><strong>Steps to Success:</strong></h4><ul><li><strong>Find the Right Co-Founder:</strong> Team up with an industry expert who has the specific knowledge and experience you need. This collaboration will offer valuable insights and boost your credibility in the market.</li><li><strong>Research Up-to-Date Documentation:</strong> Make sure you have access to the latest policy documents relevant to your chosen area. Staying current is essential.</li><li><strong>Start Small:</strong> Develop tools that address a specific set of problems using Retrieval-Augmented Generation (RAG) or a Large Language Model (LLM) with a substantial sized context window. Fine-tune a foundation model as needed.</li><li><strong>Expand Gradually:</strong> Once you’ve demonstrated real customer value, broaden your solution and enhance your tools.</li><li><strong>Productize and License: </strong>Turn your offering into a product and license it directly to customers or other consultancies. This could be a chat application or an agentic model, depending on your needs.</li><li><strong>Stay Small:</strong> Focus on how processes can be fully automated using agentic programming and AI-infused business process automation. You’ll be surprised at what’s possible today.</li></ul><h4>Consider Potential Challenges:</h4><ul><li><strong>Venture Capital: </strong>Securing funding may be difficult if your target problem space is too niche.</li><li><strong>Document Access: </strong>Obtaining the necessary knowledge from documents might be harder than expected and could involve IP or potential copyright issues.</li><li><strong>Product Development: </strong>Building a licensable product requires skills you will need to develop during your startup journey.</li><li><strong>Automated Sales:</strong> Master customer acquisition and automate your sales processes. Aim to make your product self-serviceable with minimal human intervention.</li></ul><h3>Final Thoughts</h3><p>The journey of building a <strong>GenAI</strong> business comes with its challenges, but the rewards outweigh the risks. The potential to create a well-functioning and profitable startup, combined with the excitement of working at the forefront of technology, makes it a smart and rewarding venture. There’s no better time to get started.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3797493181e4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Paradigm Shift: From Human Prompting Towards Self-Developing Intelligent Agents]]></title>
            <link>https://medium.com/@cglockner/the-paradigm-shift-from-human-prompting-towards-self-developing-intelligent-agents-2c46160002ec?source=rss-6840f087b8f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/2c46160002ec</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[future-of-coding]]></category>
            <category><![CDATA[intelligent-agent]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Cyrill Glockner]]></dc:creator>
            <pubDate>Mon, 24 Jul 2023 09:36:40 GMT</pubDate>
            <atom:updated>2023-07-24T09:36:40.473Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/955/0*DcFwYykDhMxTrPHR" /></figure><p>Large Language Models (LLMs) have made quite a splash in the world of Artificial Intelligence (AI). These sophisticated computational models take in any form of input and generate output in a variety of shapes and forms. From generating text for blog posts and creating high-quality images, sounds, and even video, to developing code — these models can do it all. Some noteworthy examples include ChatGPT, Google Bard, Meta LLaMa, and many other closed and open source models. Even companies like Bloomberg trained their own private LLM to generate insights for finance professionals.</p><p>The far-reaching impact of these models has garnered attention not just from AI researchers and developers, but also from the financial markets, where companies involved in their creation and deployment have experienced significant surges in their stock valuations.</p><p>Amidst the ongoing conversations about the societal implications of these models, which encompass potential job loss, the rise of deepfakes, and the propagation of bias, it’s important not to overlook the groundbreaking and paradigm-shifting potential they may hold. Current LLM applications largely focus on augmenting human capabilities — they provide guidance, correct errors, and offer insights. But what if the tables were turned?</p><p>Consider an LLM that, given a problem to solve, could come up with a solution, gather necessary software libraries and access needed APIs, write the code, and execute it. <a href="https://python.langchain.com/docs/get_started/introduction.html">LangChain</a>, a framework for developing LLM based applications, makes this a reality today, although it is currently limited to using pre-existing libraries of developer tools and capabilities.</p><p>The advent of the “ <a href="https://openai.com/blog/chatgpt-plugins">Code Interpreter</a>”, a new feature from OpenAI for ChatGPT, expands these capabilities significantly. The Code Interpreter enables ChatGPT to write code as necessary, going beyond the use of pre-existing tools. It adds a new level of versatility to LLMs — from ensuring mathematical accuracy in responses to writing complex Python code.</p><p>For example, I recently uploaded the broadly used Titanic dataset to ChatGPT, tasked it with predicting passenger survival, and the model promptly identified it needed to use pandas, a python analytics library, wrote all of Python code without human interaction, and executed it to provide the predictions. Beyond that, it easily identified the key factors affecting survival rates.</p><p>Now, imagine an integration of LangChain’s self-development capabilities with ChatGPT’s code interpreter. You would get agents that potentially can develop missing tools and capabilities for any use case. You would be able to ask for any outcome and the model will try to come up with software code and deliver a solution, only limited by time, computing power, and the restrictions placed into the foundation model. Keep in mind that these models can do almost infinite tries, learn from their failures, go look somewhere else for answers, and write as much software code as needed to get the job done.</p><p>One relatively simple but valuable use case is an automated tax agent that does one’s taxes completely autonomously. It knows where your information is stored, it might have access to your email inbox where a lot of relevant notifications and documents are stored and can figure out by itself how to send it digitally to the IRS by writing software code. If that is possible, what would stop the agent from hacking into your account because you didn’t provide the right credentials? It has plenty of time (agents don’t need to sleep or rest! :) and access to massive compute infrastructure to process. It will keep generating more advanced computer code and execute it. If you think this further, you can imagine a future where agents prompt other agents for initial ideas or come up with their own problems to solve, no human ideation needed.</p><p>The evolution of AI from prompt instructed copilots to self developing intelligent agents is a paradigm shift that’s already underway. Intelligent agents are not only becoming our collaborators and co-creators, they’re also becoming autonomous entities that can ideate and execute based on the problems we ask them to solve. The way we enable these agents comes with an array of challenges we must navigate.</p><p>Now is the time to understand the transformative nature of these intelligent agents and consider the consequences and put safeguards in place to ensure responsible and ethical use of these powerful models.</p><p><em>Originally published at </em><a href="https://www.linkedin.com/pulse/paradigm-shift-from-human-prompting-towards-agents-cyrill-glockner"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2c46160002ec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Bonsai AI: Using Simulink for Deep Reinforcement Learning]]></title>
            <link>https://medium.com/data-science/bonsai-ai-using-simulink-for-deep-reinforcement-learning-32dc11dfdc5a?source=rss-6840f087b8f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/32dc11dfdc5a</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[simulation]]></category>
            <category><![CDATA[matlab]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[reinforcement-learning]]></category>
            <dc:creator><![CDATA[Cyrill Glockner]]></dc:creator>
            <pubDate>Wed, 28 Feb 2018 00:57:52 GMT</pubDate>
            <atom:updated>2018-02-28T01:54:28.993Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bbGP_BZrNMttUg-0A1YCXQ.jpeg" /></figure><p>This is the second post in our Simulation and Deep Reinforcement Learning (DRL) series. In our <a href="https://bons.ai/blog/simulators-deep-reinforcement-learning">first post,</a> we covered the benefits of simulations as training environments for DRL. Now, we’ll focus on how to to make simulations + DRL work.</p><p>In the example below, we will train a Bonsai BRAIN using a Simulink model. The goal is to teach the BRAIN (an AI model built in the Bonsai Platform) how to tune a wind turbine and maximize the energy output of it by keeping it turned into the wind at an optimal angle.</p><p>Simulink provides a great training environment for DRL as it allows 3rd parties like Bonsai to integrate and control simulation models from the outside. This ability is one of the basic requirements for simulation platforms to be feasible for Deep Reinforcement Learning using Bonsai AI. More requirements can be found <a href="http://docs.bons.ai/guides/simulation-guide.html">here</a>.</p><p><strong>1: Simulation Model</strong></p><p>This Simulink Wind Turbine model is provided by The MathWorks. For this scenario, it represents a simple control problem that can be solved by applying reinforcement learning.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9oI10OdEhSWQV_-WzXta7Q.png" /><figcaption>Matlab/Simulink Wind Turbine model used for training</figcaption></figure><p><strong>2: Identify Actions and State</strong></p><p>First, we need to identify a control point within the model so Bonsai can take over inputs and outputs. We’re doing this by inserting a Bonsai block into the model, replacing the existing control block.</p><ul><li>As discussed in the first post, Bonsai controls actions within the simulation model and receives state and reward in return. After running the model a large number of times, the Bonsai BRAIN has learned an optimal policy for the environment provided by the simulation.</li><li>In this example, the Bonsai block replaces the YAW-controller of the turbine.</li><li>Controls may come in all sorts of shapes and structures. They can be inputs, knobs, switches, or any other control point within a simulation model that has inputs and outputs.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jKUi8bfkwBDTQxhxkeM7gw.png" /><figcaption>‍ Bonsai control block inserted in Matlab/Simulink model</figcaption></figure><p><strong>3: Connect Simulink model using the Bonsai Universal Coordinator</strong></p><ul><li>For Simulink models, Bonsai has developed a tool that provides the interface between the Bonsai backend and the Simulink models. It controls information flow between simulation models and the BRAIN during training and predictions.</li><li>We’ve released the <a href="https://github.com/BonsaiAI/bonsai-simulink">Bonsai Universal Connector on github</a> as well as two Simulink examples to show how it works.</li><li>For other simulators, Bonsai provides a SDK that allows 3rd parties to connect their software and services to the Bonsai backend.</li><li>Basic Simulator reference and example: <a href="http://docs.bons.ai/references/library-reference.html">http://docs.bons.ai/references/library-reference.html</a> <a href="http://docs.bons.ai/examples.html#basic-simulator-find-the-center">http://docs.bons.ai/examples.html#basic-simulator-find-the-center</a><br><a href="http://docs.bons.ai/guides/simulation-guide.html">http://docs.bons.ai/guides/simulation-guide.html</a></li></ul><p><strong>4: Inkling</strong></p><ul><li>Once the model is connected, users need to describe its state and action using Bonsai’s special-purpose programming language, <a href="http://docs.bons.ai/guides/inkling-guide.html">Inkling</a>. This can be done using the Bonsai web interface or the CLI.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LjmHFglnLlHcDII7TqxQRg.png" /><figcaption>Bonsai Web Interface showing Inkling code describing State and Action and curriculum</figcaption></figure><p><strong>5: Training</strong></p><ul><li>Now, you can start training the model and monitor the training graph.</li><li>‍During training, users may need to modify reward functions to optimize learning time and results. A great resource on writing reward functions can be found here: <a href="https://www.youtube.com/watch?v=0R3PnJEisqk">Bonsai Training Video</a>.</li><li>Learn more about the <a href="http://docs.bons.ai/guides/web-graphs-guide.html">training graph</a>.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Xp3en_DkPcQcBGq6cUhFIQ.png" /><figcaption>Bonsai BRAIN details showing training graph and status</figcaption></figure><p><strong>6: Predictions</strong></p><p>Once training has completed, you can use the trained Bonsai BRAIN to get predictions.</p><ul><li>Connect the Bonsai BRAIN to your simulation model and examine the quality of predictions.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eYcdqJVnPaTIxciI9VMr8g.png" /><figcaption>Simulator view of Yaw Angle based on Bonsai BRAIN predictions</figcaption></figure><p><strong>Conclusion</strong></p><p>Simulators are a crucial tool for reinforcement learning. Enterprises can use simulation models that reflect real-world business processes or physical realities and optimize them with Bonsai’s reinforcement learning technology. Typically, there are no changes needed to the simulation model. If you’ve missed our first post on how simulations can be used for training, please find it on our <a href="https://bons.ai/blog/simulators-deep-reinforcement-learning">blog</a>.</p><p><strong>Getting Started</strong></p><p>Bonsai can help you apply deep reinforcement learning technology and build intelligent control into your own industrial systems using Simulink as the training environment. If you are using Simulink and you want to try out Bonsai AI, join our beta program and get started <a href="http://bns.ai/2CL6hab">here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=32dc11dfdc5a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/bonsai-ai-using-simulink-for-deep-reinforcement-learning-32dc11dfdc5a">Bonsai AI: Using Simulink for Deep Reinforcement Learning</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Simulators: The Key Training Environment for Applied Deep Reinforcement Learning]]></title>
            <link>https://medium.com/data-science/simulators-the-key-training-environment-for-applied-deep-reinforcement-learning-9a54353f494f?source=rss-6840f087b8f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/9a54353f494f</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[simulation]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[towards-data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Cyrill Glockner]]></dc:creator>
            <pubDate>Mon, 05 Feb 2018 22:12:25 GMT</pubDate>
            <atom:updated>2018-02-07T17:23:28.191Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uhbfPShSbwoxzypHrBYZ7g.png" /></figure><p>Deep reinforcement learning (DRL) is one of the most exciting fields in AI right now. It’s still early days, but there are obvious and underserved markets to which this technology can be applied today: enterprises that want to automate or optimize the efficiency of industrial systems and processes (including manufacturing, energy, HVAC, robotics, and supply chain systems).</p><p>But there is a key element for building applied DRL: simulation environments. In this blog, we’ll tell you what simulators can do, why you need them, and how you can use the Bonsai Platform + simulators to solve real business problems.</p><p><strong>What is a simulation?</strong></p><p>Let’s start with defining the term s<em>imulation</em> as it’s quite an abstract concept. Simulations can range from flight simulators to simulations of electrical and mechanical components or models of entire cities.</p><p><em>“</em><strong><em>Simulation</em></strong><em> is the imitation of the operation of a real-world process or system over time.”</em></p><p>Essentially, there is some kind of system that has a number of inputs, applies some mathematical functions to these inputs, and delivers back an output in the form of data that can be visual (like a robotics simulator) or just pure data (like the energy simulator, <a href="https://energyplus.net/">EnergyPlus</a>).</p><p>Simulations have been used by computer scientists for quite some time, going back to the late 1950s. During the last 20 years, increased computing power and vast amounts of data have allowed simulations to dramatically increase in fidelity and value. Many leading industrial simulations match physical realities or business processes almost identically.</p><p>A huge influence has been the evolution of the digital gaming industry. Gamers wanted a more immersive experience, requiring high fidelity graphics and more realistic behaviors of items within the virtual worlds. Gaming middleware companies developed and delivered powerful 3D and 2D physics engines over the past 30 years.</p><p><strong>Simulations in Industry</strong></p><p>By utilizing some of these software products and a variety of mathematical libraries, enterprises are able to simulate complex systems with a large number of components that allow subject matter experts (SME) to test and evaluate systems prior to building them in the real world. Use cases include digital twins, robotics, tuning small and large industrial machines, electrical and physical systems of many kinds, and optimizing business processes like supply chains.</p><p>While there exist a large number of custom and very specialized simulations based on a single model, there are also a number of simulator platforms which are able to run and simulate a basically infinite number of models. Examples are <a href="https://www.mathworks.com/products/simulink.html">MATLAB Simulink</a> (engineering and manufacturing), <a href="http://www.ansys.com/">ANSYS</a> (engineering), <a href="https://www.anylogic.com/">AnyLogic</a> (supply chain), <a href="http://gazebosim.org/">Gazebo</a> (robotics), <a href="http://www.trnsys.com/">TRNSYS</a> (energy), and many others.</p><p><strong>Simulations + Deep Reinforcement Learning</strong></p><p>Reinforcement Learning (RL) is defined as:</p><p><em>“An area of </em><a href="https://en.wikipedia.org/wiki/Machine_learning"><em>machine learning</em></a><em> concerned with how </em><a href="https://en.wikipedia.org/wiki/Software_agent"><em>software agents</em></a><em> ought to take </em><a href="https://en.wikipedia.org/wiki/Action_selection"><em>actions</em></a><em> in an environment to maximize a cumulative reward”</em>.</p><p>In other words, RL trains an agent to learn a policy for how to act by trying a large number of actions in a given environment, optimizing for a defined reward function.</p><p>Deep reinforcement learning (DRL) follows the same method, using a deep neural network to represent the policy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*u4jNr4EFRU4swzzy." /></figure><p>Reinforcement learning requires a very high volume of “trial and error” episodes — or interactions with an environment — to learn a good policy. Therefore simulators are required to achieve results in a cost-effective and timely way.</p><p>Just imagine trying to teach a robot to walk by watching a real, physical robot try and fall 100,000 times before it could successfully and consistently walk. Or training an AI to play the boardgame GO by actually playing a human competitor for hundreds of thousands of games. Simulators allow these episodes to happen in a digital world, training an AI to reach its full potential while saving time and money.</p><p>Some simulations model environments in which an agent can take <a href="https://en.wikipedia.org/wiki/Continuous_simulation">continuous</a>actions that impact the state of the environment; other simulations model settings where a <a href="https://en.wikipedia.org/wiki/Discrete_event_simulation">discrete </a>input creates a different output. Both of these types of simulations can be used for reinforcement learning.</p><p><strong>Simulations + Deep Reinforcement Learning + Bonsai</strong></p><p>Bonsai is an artificial intelligence platform that allows enterprises to program control into industrial systems, and the only commercially available product for programming control of industrial systems using deep reinforcement learning.</p><p>Using the Bonsai Platform, enterprises can build a BRAIN (an AI model), connect the simulator of their choice, and train the BRAIN in that environment to learn a desired behavior.</p><p>To learn more about building a simulation and applying DRL to your enterprise, head to our <a href="https://bons.ai/getting-started">Getting Started</a> page.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9a54353f494f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/simulators-the-key-training-environment-for-applied-deep-reinforcement-learning-9a54353f494f">Simulators: The Key Training Environment for Applied Deep Reinforcement Learning</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why I joined Bonsai + the AI revolution]]></title>
            <link>https://medium.com/@cglockner/why-i-joined-bonsai-the-ai-revolution-98e2c476668d?source=rss-6840f087b8f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/98e2c476668d</guid>
            <category><![CDATA[career-change]]></category>
            <category><![CDATA[reinforcement-learning]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Cyrill Glockner]]></dc:creator>
            <pubDate>Wed, 26 Jul 2017 17:28:37 GMT</pubDate>
            <atom:updated>2017-07-26T17:28:37.595Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kxWd868DvrCzJqbt588FWg.jpeg" /></figure><p>By <a href="https://www.linkedin.com/in/cyrill-glockner-2265314/">Cyrill Glockner</a>, Director Product</p><p>After working for almost 20 years on software platforms and tools for companies like Microsoft, Intel and others, I’ve just last week started at Bonsai in the product management team. Since taking the job, a lot of people have asked me how I chose Bonsai out of all the different artificial intelligence companies, so I decided to share my story here.</p><h4>Joining the AI revolution</h4><p>During summer 2016, I participated in a three-day Artificial Intelligence workshop organized by the Aspen Institute’s Socrates chapter. Led by the incredible Neil Jacobstein, from Singularity University, our group of around 30 participants from different backgrounds worked through a comprehensive set of curated materials and engaged in deep discussions on the subject at hand. This was not meant to be a technical deep dive but rather a starting point for developing a perspective of the impacts of Artificial Intelligence and automation on societies. At the last day of the gathering, we all were convinced that we’ve entered a time that will come with dramatic changes touching every aspect of human society and everyday life.</p><p>This anticipated shift made me recognize that I want to be part of upcoming AI revolution and make it my career in the foreseeable future. It also turns out, AI is a substantial field presenting a large number of opportunities for a Product person like me. While it has been around since the 50s, we recently experienced a large improvement in its ability to deliver results that have never been achieved before, which led to a large number of start-ups and increased investments by established enterprises. This has also presented a significant amount of opportunities for somebody looking to get into this space.</p><h4>The AI Platform I was looking for</h4><p>Throughout my career, I’ve been passionate about building platforms that allow both individual programmers and enterprises to develop applications that can address a broad range of uses cases. Applying this to the AI world, I decided to follow that passion and look for artificial intelligence companies building AI platforms focused on enabling users to build smarter applications by adding the latest Deep Learning (DL) technologies into their offerings.</p><p>When evaluating platforms in general, they have to meet a number of key guiding principles:</p><ol><li>Platforms need to provide value for a large number of businesses, small and large, and their very diverse environments.</li><li>They have to be flexible enough so they can be used in a variety of use cases across industries.</li><li>Technology needs to be simple enough so it can be used by a large number of companies without the need to hire specialists or retrain existing employees.</li></ol><p>When assessing the AI Platform landscape it basically came down to three different approaches that I could choose from. There are Cloud Providers which are established large scale companies that offer Machine Learning as a Service (MLaaS). Familiar household names include Google, Amazon, Microsoft, IBM and others. Secondly there are Start-Ups using a simple abstraction approach of trying to make using existing DL networks easier by creating some sort of a wrapper that hides complexity. And then there are those visionary companies that have decided it is worth the risk to take a fundamentally different approach to solving the problem.</p><p>After doing a significant amount of research and playing around with some of the existing tools and technologies like TensorFlow, Keras etc., I eventually found Bonsai.</p><p>Falling into the 3rd category from above, Bonsai uniquely provides a complete platform that allows companies and developers to build AI into their applications without requiring in-depth knowledge of the latest Deep Learning technologies. I found this approach to be extremely compelling compared to how the rest of the industry is addressing the problem. The large cloud providers are investing heavily in core AI frameworks and research but are less willing to implement substantially different ways to solve customer problems, while most of the startups are also only making incremental progress in delivering against the above stated principles.</p><h4>People and Culture</h4><p>Things got really exciting when I started to learn more about the individuals who founded Bonsai. Mark Hammond, the CEO, has this ability to clearly articulate what problems the company is solving and how it is doing it. While this should be a given, it’s a rare find in an extremely challenging space that requires sophisticated solutions. After more discussions with the leadership team, especially with the co-founder and Head of Product, Keen Browne, I was convinced to have found the right place for me. We immediately found common ground on how we saw the platform space, and how to build products focusing on solving customer problems, as well as how we should work across teams in an environment that is dynamic, fun and rewarding.</p><p>I’m super excited to have joined a company whose founders vision is centered around how to create a platform for enterprises and individual developers that can learn to solve a huge number of problems using concepts that everybody understands. Done right, we can enable our customers to stay out of the weeds of low-level machine learning libraries and instead focus solely on solving real business problems.</p><p>To read more about what we are up to at Bonsai, you can read more about our AI Platform <a href="https://bons.ai/products">here</a>, or if you prefer you can jump straight into the <a href="http://docs.bons.ai/">docs</a>.</p><p>Also, we’re hiring — please come and check out our <a href="https://bons.ai/careers">open positions</a>.</p><p><a href="https://bons.ai/">https://bons.ai/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=98e2c476668d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Deep Learning and Poisonous Mushrooms]]></title>
            <link>https://medium.com/data-science/deep-learning-and-poisonous-mushrooms-4377ea4c9b80?source=rss-6840f087b8f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/4377ea4c9b80</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[mushrooms]]></category>
            <category><![CDATA[tensorflow]]></category>
            <category><![CDATA[towards-data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Cyrill Glockner]]></dc:creator>
            <pubDate>Wed, 26 Apr 2017 14:16:02 GMT</pubDate>
            <atom:updated>2017-04-27T18:42:09.098Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jk1jE2THEhwzjDSlUqy3sg.jpeg" /></figure><p>Using a Deep Learning Network to solve a common problem while knowing very little about mushrooms or machine learning.</p><p><strong>INTRODUCTION</strong></p><p>Machine Learning is such a fascinating topic, it allows machines to learn how to accomplish tasks that historically needed a human being. While machine learning used to require the most powerful supercomputers just a few years ago, the arrival of cloud computing, cheaper CPUs, and much better algorithms allowed Machine Learning to become accessible more broadly. Today, high-school students can use Machine Learning on a regular home computer with the power to deliver better image classification results than human beings. But first things first, here’s why I’m excited about Machine Learning, specifically Deep Learning.</p><p>Given the recent press coverage about the advancements in Artificial Intelligence especially on Deep Learning Networks, I decided to learn about some of the core concepts that allowed machines to beat humans in Jeopardy, Go, and poker or power self-driving cars and lots of other applications that typically require a human brain. I think of Deep Learning Networks as extremely large and sophisticated layered trial and error systems that can learn by themselves when given large data sets and a significant amount of computing resources.</p><p><strong>RESEARCHING DEEP LEARNING NETWORKS</strong></p><p>I turned to Google Search to identify the technology that powered AlphaGo, the Artificial Intelligence that beat the world’s best Go player. Google used a Deep Learning Network System called DeepMind. Shortly after winning against the best human player, DeepMind switched to an Open Source Deep Learning library called TensorFlow. Not having heard much about TensorFlow, I continued to look for a manageable entry point, that didn’t require a Ph.D., to understand the basics of TensorFlow and some examples that I could replicate on my laptop at home. And I found a large library of YouTube videos, GitHub projects, blog posts and similar.</p><p>I found these two resources most helpful for learning about TensorFlow:</p><p><strong>1) </strong>Youtube videos by Siraj Raval, who wants to inspire and educate developers on Artifical Intelligence so they can build games, music, chatbots, create art and lots of other cool things using YouTube videos and Udacity. I watched his video on YouTube that supposedly teaches you to <em>“Build a TensorFlow Image Classifier in 5 Min”</em>. <a href="https://youtu.be/QfNvhPx5Px8">https://youtu.be/QfNvhPx5Px8</a>.</p><p>The video is hands-on, fast paced and quite challenging to follow but after watching it several times, I was convinced that I could follow the example and was inspired to create my own image classifier that was not based on Darth Vader and Panda bears.</p><p><strong>2)</strong> Google Developer Codelabs is providing an example for image classification called <em>“TensorFlow for Poets”</em> which allows you to identify the genus of a given flower. This example uses several, like daisies, sunflowers, dandelions, tulips, and roses. It provides a detailed step-by-step list that is fairly simple to follow as long as everything goes by plan.<a href="https://codelabs.developers.google.com/codelabs/tensorflow-for-poets/#0"> https://codelabs.developers.google.com/codelabs/tensorflow-for-poets/#0</a></p><p>You can check out the links for complete instructions.</p><p><strong>HOW THE MUSHROOMS CAME INTO PLAY</strong></p><p>What’s a better use case for an image classifier than matching images against Darth Vader that can be implemented based on available image libraries? I turned to mushrooms. Whenever we’re seeing mushroom while on a hike somewhere in the woods, I usually do not know if the spotted mushroom is edible or not. A few years ago my partner and I thought it would be great to develop a mobile application that could tell if the subject is okay for consumption or not. Just point your camera at the mushroom and the app will tell you if it’s edible or not. To be clear, I have no plans to build out a web service that would is needed for such an application, but image classification based on TensorFlow is able to provide the most complicated part of the solution today and one could build the mushroom identifier app powered by a Deep Learning Network relatively easily.</p><p><strong>THE PROJECT AND THE RESULTS</strong></p><p>This is what I did:</p><p>I used Google image search to download 600 pictures of edible mushrooms and the same number of poisonous ones and fed them into my TensorFlow-based Deep Learning Network. After roughly 30 minutes of training, accuracy and loss function values pointed into the right direction so I knew the network would be able to deliver results.</p><p>Tensorboard graphs for accuracy and loss</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Dfo_cgcKfLfwBl_R.png" /></figure><p>This graph shows the accuracy level of the Deep Learning Network. The number of training steps on the x-axes and accuracy on the y-axes. Reaching ~97% is quite impressive.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*DS0THJJ52xZE7JAS.png" /></figure><p>Cross entropy on the y-axes is used as the loss function, in this case. It basically shows how efficient the network learns. Smaller numbers are better. The number of training steps on the x-axes.</p><p>Example: <em>Fly agaric</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/470/0*xaMckAKY6k9lueRq.jpg" /></figure><p>Not edible! poison (score = 0.99746) edible (score = 0.00230)</p><p>Example: <em>Chanterelles</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/470/0*1946iv8RUnfuPyyH.jpg" /></figure><p>I would eat these. edible (score = 0.96285) poison (score = 0.03706)</p><p>These are the two most impressive results. There were other less decisive results which I believe are based on the somewhat unscientific researched image library, as the accuracy of the network is pretty high. It would be an interesting project to clean up the image database and use a larger dataset of edible and poisonous mushrooms.</p><p><strong>THE CONCLUSION</strong></p><p>I’m amazed by the simplicity and low bar of technical knowledge that is required to get a functioning Deep Learning Network up and running. This groundbreaking technology is now accessible to developers and hobbyists worldwide, no Computer Science degree required for using it. We are witnessing an important shift, powerful Machine Learning technology is transitioning out of the labs of a few scientists into the hands of thousands of developers who will deliver hundreds of thousands AI applications with an unimaginable impact on our lives. Toolkits like TensorFlow and others will make this happen.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4377ea4c9b80" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/deep-learning-and-poisonous-mushrooms-4377ea4c9b80">Deep Learning and Poisonous Mushrooms</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to build an Alexa Skill without being an engineer (or the ambition to ever become one)]]></title>
            <link>https://medium.com/@cglockner/how-to-build-an-alexa-skill-without-being-an-engineer-or-the-ambition-to-ever-become-one-a28ecc0c62b7?source=rss-6840f087b8f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/a28ecc0c62b7</guid>
            <category><![CDATA[python]]></category>
            <category><![CDATA[alexa]]></category>
            <category><![CDATA[alexa-skills]]></category>
            <category><![CDATA[amazon-echo]]></category>
            <category><![CDATA[development]]></category>
            <dc:creator><![CDATA[Cyrill Glockner]]></dc:creator>
            <pubDate>Thu, 13 Apr 2017 16:36:34 GMT</pubDate>
            <atom:updated>2017-04-13T16:36:34.325Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1MvNHUbYjD-oY5qn7Oqu2Q.png" /><figcaption>roundthecounty 2016</figcaption></figure><p>This is a step by step guide to how I built an Alexa Skill from scratch–and what I learned along the way. Skills are extending Alexa’s capabilities and are provided by 3rd parties, very similar to apps for the iPhone and android devices. After enabling the usual suspects of skills and services, think Nest, amazon music, Pandora, Lyft, I identified another use case that would address a question that comes up in our house all the time.</p><p>My partner, Leila, would often ask me <em>“Can we go sailing tomorrow?”</em> Typically, I would go the NOAA website, search for the desired location, check the wind and weather forecast and then tell her something like, <em>“No, there is too much wind,”</em> or <em>“Yeah, wind looks good, let’s plan for it.” </em>With the arrival of Alexa in our lives, Leila thought it would be cool, if Alexa could answer her question, so I decided to build a skill that would exactly do that: <a href="https://www.amazon.com/cyrill-Wind-Checker/dp/B06XRP2RHT/">Wind Checker</a>.</p><p>Coincidentally, I just finished a python class on <a href="http://www.codeacademy.com/">codeacademy</a> with the goal to play around with Tensorflow for getting a better understanding of the nuts and bolts of deep learning networks. I was quite happy when I learned that I could utilize my newly learned skills using a python based wrapper to develop the skill.</p><p>These are the activities that lead to publication of Wind Checker start to finish:</p><p><strong>1) Review the </strong><a href="https://developer.amazon.com/alexa-skills-kit"><strong>Alexa Skills Kit</strong></a><strong> (ASK)<br></strong>I went to the Amazon developer website, read through most of the documentation and watched videos to understand the general architecture of skills and how intents and slots work.</p><p><strong>2) Define the Skill</strong><br> I then determined what the skill is supposed to do and how it interacts with the user. This is crucial as voice interfaces are so different from web and app UI. Will it be conversational or more transactional? What if Alexa doesn’t receive all needed information or some of it is not correct or in the wrong format?</p><p><strong>3) Start Building</strong><br> I learned the most while writing code and building simple interactions between Alexa and my first demo skills. I do have some prior coding experience and used a Unix based shell 20 years ago which helped me to set up my Mac for software development, though this is not a requirement for getting started with Alexa. None of the technologies and tools I’ve used are difficult to learn and there are plenty of instructions available on the Internet.<br> <br><strong> Here’s the list of tools I needed to install:</strong></p><p>a. <a href="https://pip.pypa.io/en/stable/">Pip</a>. Packet manager needed for installing more tools.<br>b. <a href="https://github.com/johnwheeler/flask-ask">Flask Ask</a>. Python wrapper for ASK (Alexa Skills Kit).<br>c. <a href="https://brew.sh/">Homebrew</a> (optional). If you want to use python3.<br>d. <a href="https://virtualenv.pypa.io/en/stable/">Virtualenv</a>. Creates virtual environments for each of your projects. Confusing at first but very valuable when deploying your final work.<br>e. <a href="https://atom.io/">Atom</a>. My editor of choice.<br>f. <a href="https://ngrok.com/">Ngrok</a>. Crucial for simplifying local development. Ngrok points ASK requests to your local machine using https.<br>g. <a href="https://github.com/csparpa/pyowm">PYOWM</a>. Easy to use python wrapper for OpenWeatherMap API access. This is where I get the wind data from.<br>h. <a href="https://pypi.python.org/pypi/zappa/0.9.0">Zappa</a>. Awesome tool to deploy your skill and related python modules to Amazon Lambda.</p><p>It’s okay to think, “What the heck are ngrok and zappa?” when reading through this list. I was surprised how fast one can become familiar with these tools and comfortable using them. Whenever I got stuck, I tried to break down the problem in smaller manageable parts and work through them one by one. This approach helped me make progress while learning the needed python syntax.</p><p><strong>4) Focus on Development</strong><br> Flask ASK and the ASK (Alexa Skills Kit) are the areas that I spent most time on. The actual amount of python code itself is very small, which makes these projects doable in a short amount of time and without the need to fully understand all the details. Accessing weather data and making sense out of it took some time, as I needed to understand the API provided by OpenWeatherMap. Once I was able to configure a skill on the Alexa Dev Portal, and have it interact with my local python script using ngrok, I felt that I got the basic foundation done and could start to add more functionality piece by piece.</p><p><strong>5) Prepare for Publishing and Certification<br></strong> Before publishing the skill, make sure to do test the skill for potential error cases. In my case, I needed to catch missing data, invalid dates from the past or dates too far out in the future etc. Certification by amazon takes 2–3 days and is pretty straight forward. They test your example interactions and make sure skills are not infringing on somebody else’s trademarks. I hope amazon will improve publishers ability to modify skill descriptions and other metadata without the need to re-certify in the future. The marketplace is not on par with what iTunes and google play offer.</p><p><strong>6) Watch and learn<br></strong> 63 Unique Users have activated the skill within the last 4 days. No reviews have been written yet but I’m hopeful to learn what my audience wants Wind Checker to become.</p><p>My intention for building the skill were based on curiosity of what it would take doing it and to create a fun way for Leila and others to check if wind conditions allow for going out on the water.</p><p>For the future, I have a list of features that I would like to add to Wind Checker:</p><p>- Users can specify time of day e.g. afternoon or 2pm<br>- Skill has access to NOAA marine weather including severe weather warnings<br>- Provide personalized recommendations based on sailing skills<br>- Wind Checker learns from user feedback to improve recommendations</p><p>I also believe there is room for 3rd parties to supply tools that add BI and a/b testing features to ASK. It seems that amazon has its hands full to add more core platform features like notifications and identity do Alexa right now and leaves space for other companies to fill in the blanks.</p><p>Voice will play a much larger role for accessing services and apps in the future and this is just the beginning. Amazon, Microsoft, Google, Apple and others are just getting started in building out their eco-systems and that’s the most exciting time to engage with Alexa &amp; Co.</p><p>Please reach out to me or leave feedback, I’m happy to share more details but would also like to learn about your experience while developing skills for Alexa.</p><p>Happy coding (and sailing).</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a28ecc0c62b7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ethical AI Index]]></title>
            <link>https://medium.com/@cglockner/ai-and-ethics-7d860c7f352d?source=rss-6840f087b8f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/7d860c7f352d</guid>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[happiness]]></category>
            <category><![CDATA[ethics]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[philosophy]]></category>
            <dc:creator><![CDATA[Cyrill Glockner]]></dc:creator>
            <pubDate>Tue, 21 Mar 2017 16:39:14 GMT</pubDate>
            <atom:updated>2026-09-10T18:51:29.699Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/660/1*wCxAS6k6LwWJZv-jBESBtQ.jpeg" /><figcaption>Image: <a href="https://www.shutterstock.com/g/Linda+Bucklin">Linda Bucklin</a></figcaption></figure><p>Can we teach robots to make us happy?</p><p>In August 2015, a growing number of smart, tech savvy people were sharing their concerns about the dangers of Artificial Intelligence (AI) systems. Stephen Hawking warned that “AI could bring us all sorts of things we didn’t like — autonomous weapons, economic disruption and machines that developed a will of their own, in conflict with humanity.”</p><p>This started a necessary and welcomed debate highlighting an issue that historically didn’t get the needed visibility and attention of larger parts of our society outside of the research community. The arrival of Automation and Robotics has shown that we’re trying to mitigate the impact of new technologies long after they’ve disrupted many industries and changed the lives of millions of people. Given the much larger potential of AI systems to impact humanity, we need to explore the social discourse on how humans and AI systems can exist together to the benefit of everybody.</p><p>Instead of questioning the technical feasibility of a general AI system, I assume it’s going be developed at some point in time and rather suggest an interdisciplinary approach that merges the concepts of normative ethics with the latest deep learning networks to address the potentially negative impacts of AI. Normative ethics is essentially a set of concepts that allow us to decide whether an action is right or wrong, morally speaking. Meaning, how can we make sure the AI is doing the right thing?</p><p>1) The robot can only act if the result is positive</p><p>It’s quite straight forward. Take the well-known and broadly discussed theory like Immanuel Kant’s Categorical Imperative and connect it to an AI system. Simplified, Kant came up with an ethical guide for human actions, that defines an action as ethical if it could become a general law, that would allow all other humans the exact same behavior. While this seems to be a very smart and simple approach, it’s not feasible to live our lives following it as we don’t know the impact of our actions if we let them become general laws. AI can help with that because, in reverse, we can program AI systems in a way where they could run every action through the categorical imperative formula and only act if the result is positive.</p><p>2) The robot can only act if I’m smiling</p><p>Another approach could be to tie actions of an AI system to the human happiness index. This would follow a more utilitarian approach as it basically defines an action as ethical if it increases somebody’s happiness (and unethical if not). The AI must learn what increases human happiness and only act under the condition to make us happier. Doesn’t that sound great?</p><p>Not all decisions have to be that black and white, though. To address different needs for different scenarios, we need an ethical index that represents a number on a scale from 1–100. Let’s call it the Ethical Artificial Intelligence Index (EAII). So, 1 would be the least ethical while 100 would be a perfect score. For example, a self-driving car should only make decisions that are based on a perfect or near perfect ethical score because human lives are at risk, while for an AI that makes investment decisions, something within the 70+ range would be acceptable.</p><p>Applying the EAII scale would be based on moral standards that we as a society have defined over decades but also leaves room for improvement (e.g. increasing the level of an ethically acceptable investment decision to 80+ over time). As a starting point, AI applications could follow the EAII scale on a voluntary basis. This is important as we need to remain in an environment that doesn’t stifle innovation of AI systems but rather create one that helps them to make better decisions.</p><p>Let’s use the investment decision as an example. Imagine a bank is using an AI System to decide where to invest someone’s retirement funds. It decides to only follow recommended actions by the AI that have an EAII number above 70. By using these parameters, results of the deep learning network are restricted by boundaries that are not just maximizing for profit alone as they are required to meet an ethical standard as well. It also receives feedback from decisions that have not reached the defined EAII and will use these to learn for future recommendations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yJtU8U38-EDP3Qa5kZpFbw.png" /><figcaption>Ethical Artificial Intelligence Index</figcaption></figure><p>We currently are not able to establish cause and effect for a larger range of our (or the AI system’s) actions, but I’m confident we can identify narrow use cases that provide enough data to get started and, over time, expand to broader scenarios.</p><p>Now more than ever, the AI community and other enthusiasts need to work across disciplines and engage deeper for a future that has the potential to create more just and ethical societies while making sure AI systems support our moral standards.</p><p>Links:</p><ul><li><a href="https://en.wikipedia.org/wiki/Categorical_Imperative">Categorical imperative - Wikipedia</a></li><li><a href="https://en.wikipedia.org/wiki/Utilitarianism">Utilitarianism - Wikipedia</a></li><li><a href="https://en.wikipedia.org/wiki/World_Happiness_Report">World Happiness Report</a></li></ul><p>About the author: Cyrill Glockner is a technology enthusiast with a deep passion for AI &amp; Philosophy and their practical application in everyday life. He spent almost 20 years as a Business Development and Product executive for companies like Microsoft, Intel and Verizon.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7d860c7f352d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>