<?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 Alexander-Derek Rein on Medium]]></title>
        <description><![CDATA[Stories by Alexander-Derek Rein on Medium]]></description>
        <link>https://medium.com/@alexanderderekrein?source=rss-d611dd14ec20------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*6gm-uJ2gyr0OMbYU.</url>
            <title>Stories by Alexander-Derek Rein on Medium</title>
            <link>https://medium.com/@alexanderderekrein?source=rss-d611dd14ec20------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 06:48:47 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@alexanderderekrein/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[Thoughts on Parenting]]></title>
            <link>https://medium.com/@alexanderderekrein/thoughts-on-parenting-bf22631ac692?source=rss-d611dd14ec20------2</link>
            <guid isPermaLink="false">https://medium.com/p/bf22631ac692</guid>
            <category><![CDATA[parenting]]></category>
            <dc:creator><![CDATA[Alexander-Derek Rein]]></dc:creator>
            <pubDate>Sat, 27 Jan 2018 04:27:49 GMT</pubDate>
            <atom:updated>2018-01-27T04:27:49.443Z</atom:updated>
            <content:encoded><![CDATA[<p>One of my best friends since childhood is visiting me right now. He will be a father soon and we’ve been discussing some of his thoughts on becoming a dad and comparing these with how we think our parents have experienced raising us.</p><p>I believe that my parents got a lot right raising my sister and I by setting some core values straight such as “you may never depend on a man in your life” for my sister and letting most other things very loose for us to figure out on our own. However, I believe that it’s been a great challenge or impossible for them to figure out what’s actually going on with us. Especially for my dad: he was 55 when I was born which means we’re two generations apart.</p><p>For my dad’s generation the social currency is money and things you can buy with money. For my generation the social currency is Instagram and the cool things you showcase there and the amount of <em>Insta</em> followers and likes you can accumulate.</p><p>My dad is so far away from technology that he has a hard time using a TV remote and is absolutely unable to use a smartphone. How is he supposed to understand what motivates my <em>Insta</em>-crazy generation?</p><p>When talking to my friend he expressed that he <em>just</em> wants his daughter to be cool and happy. But what will being cool mean for his daughter?</p><p>What’s even more striking to me is whether we will we be able to understand what’s cool for her generation. Teens today <a href="http://time.com/4946997/teens-sex-dating-drinking/">date much less and drink much less than we did</a>. That already doesn’t make sense to me. They are so connected and it must be so much easier to meet other teens. For my friends and I that was much harder since we couldn’t just meet online. So why isn’t this happening? (feel free to use the comments section in case you know)</p><p>Hence, I already have a hard time to understand what’s going on. And my buddy’s kid and ultimately also my kids will be generations apart.</p><p>I believe I will have a strong desire to connect with my kid(s) and to be able to understand what’s going on. But will I be able to?</p><p>Actually, maybe it’s more beautiful if I cannot fully understand. Who knows?</p><p>Thanks for taking the time to read this,</p><p>Derek</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bf22631ac692" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Gen Y vs. iGen]]></title>
            <link>https://medium.com/@alexanderderekrein/gen-y-vs-igen-be75f81cb20?source=rss-d611dd14ec20------2</link>
            <guid isPermaLink="false">https://medium.com/p/be75f81cb20</guid>
            <category><![CDATA[geni]]></category>
            <category><![CDATA[millenials]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[davinci]]></category>
            <category><![CDATA[gen-y]]></category>
            <dc:creator><![CDATA[Alexander-Derek Rein]]></dc:creator>
            <pubDate>Mon, 22 Jan 2018 04:42:44 GMT</pubDate>
            <atom:updated>2018-01-22T04:45:27.993Z</atom:updated>
            <content:encoded><![CDATA[<p>iGen is the generation succeeding Gen Y. Definitions about birth years vary but let’s define the birth years as 1997 to present (the iPhone was released in 2007 and become mainstream in the years after).</p><p>It’s controversial whether iGen will use the power of the internet to their advantage or whether they will just remain caught in their Instagram infused vanity.</p><p>My stance on this is that iGen will be the smartest generation ever. Not necessarily in terms of median or average smartness but the p99 smartness will outperform Gen X.</p><p>I believe this not because of smartphone access but because of the internet progress that they were born into. Facebook was founded in 2004, Youtube in 2005, Quora in 2009, and Coursera in 2012. Hence, the amount of knowledge that is available for every (pre-)teen to indulge is extremely vast. And websites like Quora help connect the dots. Let’s take learning a programming language as an example. Gen X had to learn programming through books and had to find a way to connect the dots e.g. which language do I choose and why. iGen can just ask Google or online communities to connect the dots on their behalf. So making sense out of things is much easier than it used to be for Gen Y.</p><p>I recently read Walter Isaacson’s recent <a href="http://www.simonandschuster.com/books/Leonardo-da-Vinci/Walter-Isaacson/9781501139154">Da Vinci biography</a>. Isaacson proposes a striking answer to the question of why DaVinci had become such an outstanding scientist and painter: because he had not attended a regular school but was self-taught and driven by his vast curiosity.</p><blockquote>“Describe the tongue of the woodpecker” — item on Da Vinci’s todo list</blockquote><p>Similarly, iGen kids can pursue their curiosity very easily online after school with barely any limits to what they can learn and create. The programming skills Mark Zuckerberg needed to create Facebook are just one click to an online programming bootcamp or a Coursera course away.</p><p>Hard to imagine what Da Vinci would have created with that knowledge at hand. Maybe just some Instagram pictures.</p><p>We won’t find out anymore about Da Vinci but the future will show what iGen will create.</p><p>Thanks for reading,</p><p>Derek</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=be75f81cb20" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The History of Big Data]]></title>
            <link>https://medium.com/@alexanderderekrein/the-history-of-big-data-5d4aa3899afd?source=rss-d611dd14ec20------2</link>
            <guid isPermaLink="false">https://medium.com/p/5d4aa3899afd</guid>
            <category><![CDATA[mapreduce]]></category>
            <category><![CDATA[database]]></category>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[presto]]></category>
            <dc:creator><![CDATA[Alexander-Derek Rein]]></dc:creator>
            <pubDate>Tue, 07 Nov 2017 15:37:19 GMT</pubDate>
            <atom:updated>2017-11-07T15:38:28.249Z</atom:updated>
            <content:encoded><![CDATA[<p>Big data has been one of today’s most popular buzzwords. Let’s debunk it:</p><p><strong>When is data “big”</strong></p><p>Usually the threshold is about 5TB of data that needs to be searchable/queryable/analyzed/mined fast. The video data in 4k from your last ski trip which may easily exceed 5TB does not fall into that category as it just needs to be stored but not queried in an efficient way.</p><p>5TB is the threshold as traditional search/query/analysis/mining solutions usually start being hard or expensive to operate/maintain/scale at that point.</p><p>But if you’re below the magic 5TB threshold you are usually better off using a traditional database such as PostgreSQL or MySQL.</p><p>Real world big data appliances of course are often way larger than 5TB. As an example, the FB data warehouse was 300 petabytes in 2013 [1].</p><p><strong>I/O vs. Moore’s Law</strong></p><p>One fundamental problem underlying big data is that Moore’s law — which describes that processing power doubles every two years and thus grows exponentially — I/O speeds however increase only in a linear fashion.</p><p>This means that I/O would be the bottleneck for databases that need to analyze larger amounts of data.</p><p><strong>The Birth of Map Reduce</strong></p><p>This was one of the motivations for Google to introduce MapReduce [2]. The pattern is simple: because I know that I am I/O bound on a single machine I distribute my task across many machines to get parallel disk I/O.</p><p>This white paper was the foundation for the inception of multiple big data companies that have since been created based on the map reduce promise such as Cloudera and Hortonworks.</p><h3></h3><p>1969:-what&#39;re you doing with that 2KB of RAM?-sending people to the moon2017:-what&#39;re you doing with that 1.5GB of RAM?-running Slack</p><p><strong>The World after Map Reduce</strong></p><p>The need for parallel disk I/O was one of the motivations for Google to build Dremel/BigQuery [3]. BigQuery is an interactive big data database.</p><p>Facebook implemented the very same principle in it’s Presto database which has been significantly gaining popularity. Presto also powers Amazon Athena which democratizes presto since it is serverless and has a very low maintenance penalty.</p><p>Parallel disk I/O also motivated many other big data projects such as Elasticsearch.</p><p><strong>Status Quo &amp; Next Steps</strong></p><p>Currently all of these technologies are being improved and cloud computing providers are significantly contributing to the democratization of big data tools. Thus, making it much easier for developers to start using big data technologies. For many use cases big data solutions will be the default choice going forward.</p><p>Big data analytics and data retrieval will become very important for AI as algorithms are trained on this data and making it accessible is a prerequisite to training.</p><p>Thanks for reading,</p><p>Derek</p><p>[1] <a href="https://www.facebook.com/notes/facebook-engineering/presto-interacting-with-petabytes-of-data-at-facebook/10151786197628920/">https://www.facebook.com/notes/facebook-engineering/presto-interacting-with-petabytes-of-data-at-facebook/10151786197628920/</a></p><p>[2] <a href="https://static.googleusercontent.com/media/research.google.com/en//archive/mapreduce-osdi04.pdf">https://static.googleusercontent.com/media/research.google.com/en//archive/mapreduce-osdi04.pdf</a></p><p>[3] <a href="https://cloud.google.com/files/BigQueryTechnicalWP.pdf">https://cloud.google.com/files/BigQueryTechnicalWP.pdf</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5d4aa3899afd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI ASICs in phones: what it means for the world and you]]></title>
            <link>https://medium.com/@alexanderderekrein/ai-asics-in-phones-what-it-means-for-the-world-and-you-9df18bb5c9e6?source=rss-d611dd14ec20------2</link>
            <guid isPermaLink="false">https://medium.com/p/9df18bb5c9e6</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[apple]]></category>
            <category><![CDATA[android]]></category>
            <category><![CDATA[asics]]></category>
            <dc:creator><![CDATA[Alexander-Derek Rein]]></dc:creator>
            <pubDate>Tue, 31 Oct 2017 14:54:38 GMT</pubDate>
            <atom:updated>2017-10-31T14:55:08.185Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>As always — let’s begin with some history ;)</strong></p><p>In 2012 an algorithm called <em>AlexNet</em> won the prestigious ImageNet Challenge [1]. This marked a point of inflection as <em>AlexNet</em> is a neural net. Prior to that more classic machine learning (ML) algorithms had been best performing in the challenge.</p><p>Ever since this moment deep learning has been conquering many other areas in which classic machine learning had prevailed prior.</p><p><strong>The Compute Caveat</strong></p><p>However, there has been one big blocker in deep learning’s crusade: compute. Deep learning algorithms are very compute intense (and thus power and battery intense). Therefore, they have traditionally not been able to run on on-device e.g. phones in real-time but only in the cloud.</p><p>Evaluating the algorithm in the cloud has several drawbacks:</p><ol><li>Non-realtime: roundtripping to the cloud introduces latency that makes real-time impossible</li><li>Network reliability: you cannot run the algorithm offline</li><li>Privacy: the data has to transfer to the internet. Not good for some privacy sensitive tasks</li></ol><p><strong>The ASIC sweetness</strong></p><p>One way to lift the problems above are “application-specific integrated circuits” (ASICs). There exist ASICs for Bitcoin mining and also AI ASICs have been used by Google’s Compute Cloud for a while [2]. The goal is to carry out mining or AI applications more efficient while consuming less power.</p><p>Three years ago Tim Cook announced that Apple is working on machine learning ASICs that run on the phone. Apple’s A11 Bionic chip that was introduced with the iPhone 8 and X has a “neural engine” that “enabled FaceID” [3]. Google released the Nexus 2 a few weeks ago and it also ships with the <em>Pixel Visual Core</em> an <em>Image Processing Unit</em> (IPU) that is designed to run ML algorithms [4]. In the initial release the IPU only runs ML algorithms that enable depth-of-field features on the Nexus 2 without requiring Google to add a second lens to the Pixel as Apple is doing it for the same purpose since the iPhone 7. Google also announced that they would allow others to use the IPU in the future through new software releases. Last week Google released Android Oreo 8.1 in beta which contains a new <em>Neural Networks API</em> which allows everybody to target the IPU.</p><p><strong>“Tremendous potential to do on-device processing”</strong></p><p>At the 30 year anniversary of Taiwanese chip manufacturer TMSC which manufactures all of Apples’ chips Apple COO Jeff Williams said in his remarks that he sees “tremendous potential to do on-device processing” and “we can’t think of anything more significant than this” [3].</p><p>Qualcomm which manufactures whips for many Android OEMs is also including AI ASICs in their chips [5].</p><p><strong>Developer Access</strong></p><p>Apple supports running models created through Google’s popular TensorFlow framework. Qualcomm is working on enabling Caffe2 [5]. As part of the <em>Android Neural Networks API</em> Google envisions developers to run TensorFlow, Caffe2, and other models [4].</p><p>Currently scientists already use these frameworks to train and evaluate their models but they deploy them in the Cloud. Now they can basically take the models they’ve already trained and deploy them on phones.</p><p>Hence, the entry barrier for on-device AI processing is low.</p><p><strong>What can we expect from this?</strong></p><p>Neural nets will continue to conquer many areas of machine learning as they provide best results. AI ASICs on phones will be commodities and most smartphones will be equipped with the capability over the next two years. Also the computer vision and voice algorithms will continue to improve.</p><p>Now that we can run many of these in realtime on-device we will see a lot of great apps and game changing companies because of this.</p><p>When Apple added GPS and internet to the iPhone companies like AirBnb became possible because of the ubiquitous internet and in the case of Uber GPS access.</p><p>I believe this will enable a new set of game changing companies as classification and voice algorithms can be run on-device and thus the entry barrier is lowering.</p><p>Let’s see what the future brings.</p><p>Thanks for reading,</p><p>Derek</p><p>[1] <a href="https://en.wikipedia.org/wiki/AlexNet">https://en.wikipedia.org/wiki/AlexNet</a></p><p>[2] <a href="https://cloud.google.com/tpu/">https://cloud.google.com/tpu/</a></p><p>[3] <a href="https://www.macrumors.com/2017/10/23/jeff-williams-attends-tsmc-30th-anniversary/">https://www.macrumors.com/2017/10/23/jeff-williams-attends-tsmc-30th-anniversary/</a></p><p>[4] <a href="https://www.blog.google/products/pixel/pixel-visual-core-image-processing-and-machine-learning-pixel-2/">https://www.blog.google/products/pixel/pixel-visual-core-image-processing-and-machine-learning-pixel-2/</a></p><p>[5] <a href="https://www.qualcomm.com/news/onq/2017/08/16/we-are-making-device-ai-ubiquitous">https://www.qualcomm.com/news/onq/2017/08/16/we-are-making-device-ai-ubiquitous</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9df18bb5c9e6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Neural Nets: nobody can reconstruct their decisions so it’s only a question of time until you will…]]></title>
            <link>https://medium.com/@alexanderderekrein/neural-nets-nobody-can-reconstruct-their-decisions-so-its-only-a-question-of-time-until-you-will-a53fc790e50?source=rss-d611dd14ec20------2</link>
            <guid isPermaLink="false">https://medium.com/p/a53fc790e50</guid>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Alexander-Derek Rein]]></dc:creator>
            <pubDate>Tue, 17 Oct 2017 04:34:22 GMT</pubDate>
            <atom:updated>2017-10-17T04:34:22.647Z</atom:updated>
            <content:encoded><![CDATA[<h3>Neural Nets: nobody can reconstruct their decisions so it’s only a question of time until you will be a false negative</h3><p>Over the last few years neural networks have managed to outperform classic machine learning in many domains. As an example, the best image classification and voice algorithms such as Alexa and Siri are powered by neural networks.</p><p>Neural networks have also been further democratized with the release of frameworks such as Google’s Tensorflow, Facebook’s Caffe, MXnet, and others. This means that everybody who prior considered classic machine learning algorithms to train models is now able to easily access neural networks as an option as well.</p><p><strong>Great, but why should you care?</strong><br>One major difference from neural networks to classic machine learning algorithms such as Support Vector Machines or Decision Trees (just to name a few very popular ones) is that it’s almost impossible to understand WHY a certain decision was made. This is especially true in some of the best performing incarnations of neural networks such as deep learning or convolutional neural networks (CNN).</p><p>This means that if a neural network takes action on a specific classification or forecast nobody will be able to reconstruct how the model came to the conclusion and what the meaning of the neurons in the network is.</p><p>Let’s take a look at popular examples such as banking and self-driving cars:<br>1) Lending: your bank might deny you a credit based on a forecast a deep neural net made. You might ask the bank why and even the engineers who trained the model will not be able to make a statement on that<br>2) Self-driving cars: the car might suddenly stop because it detected a stop sign. On closer analysis engineers detect that there was no stop sign. But there is no way for them to determine WHY the algorithm detected a stop sign</p><p><strong>Summary</strong><br>Neural networks perform best in many areas of machine learning. That and the improving ease of access through general framework will continue to drive their adoption.<br>This means algorithms whose decisions nobody can reconstruct will continue to shape our world.</p><p>It’s only a question of time until you will be a false negative and nobody will be able to confirm or deny it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a53fc790e50" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Google vs. Uber lawsuit: why is the LiDAR technology which the lawsuit is about so important?]]></title>
            <link>https://medium.com/@alexanderderekrein/google-vs-uber-lawsuit-why-is-the-lidar-technology-which-the-lawsuit-is-about-so-important-b03de848f1fd?source=rss-d611dd14ec20------2</link>
            <guid isPermaLink="false">https://medium.com/p/b03de848f1fd</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <category><![CDATA[uber]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[google]]></category>
            <dc:creator><![CDATA[Alexander-Derek Rein]]></dc:creator>
            <pubDate>Fri, 13 Oct 2017 00:18:06 GMT</pubDate>
            <atom:updated>2017-10-15T22:28:55.913Z</atom:updated>
            <content:encoded><![CDATA[<p>Uber acquired Otto — a self-driving truck startup founded by former Google engineer Anthony Lewandowsky. Google accuses Lewandowsky of stealing Google’s LiDAR blueprints [1].</p><p><strong>Who cares about LiDAR blueprints?</strong></p><p>LiDARs are 3d scanners. 3d data yields better results in many CV classification problems e.g. cyclist detection. Google’s blueprints aim to produce cheaper LiDARs. This matters as currently the price for a off-the-shelf LiDAR is around ~$75k (they are also heavy). That is much more than the price of an off-the-shelf car which is why it’s being considered a burden on the way to mass producing self-driving cars.</p><p><strong>Do you actually need a LiDAR?</strong></p><p>Tesla is commonly perceived as one of the leading companies in the self-driving car space but they don’t feature LiDARs in their cars [2]. Critics such as the former head of Google’s self-driving car project Chris Urmson therefore accuse Tesla of never being able to provide full (aka level 5) autonomous driving using their current hardware despite Elon Musk having promised exactly that (which might pave the way for a class action lawsuit against Tesla if it doesn’t happen) [3].</p><p>It’s interesting because the LiDAR question divides the self-driving car space a little:</p><p>Mobileye (which partners with BMW and Fiat Chrisler) and which was recently acquired by Intel, stands with Tesla in its belief that level 5 autonomous driving can be achieved without LiDARs [4][5].</p><p>Mobileye categorizes the path to level 5 into two core components:</p><p>+ Sensing: understanding the world around the car including free space, moving objects, traffic lights, etc</p><p>+ Driving policy: how does the car interact with other cars on the road</p><p>Sensing is where the LiDARs contention comes into place. Functional safety standards require autonomous cars to have a backup for all sensors [6]. Many players use LiDAR as a backup or primary sensor (Waymo uses it as the primary sensor) [2]. Mobileeye (and Tesla) claim they can provide backup sensors using other hardware sensors such as radar or using hybrid hardware/software solutions such as high-definition maps with an ultra-high time-to-reality refresh rate [6]. Mobileeye will soon have such maps as their technology is deployed in many consumer BMWs starting in 2018 [7]. Tesla of course also has such maps but Tesla builds way less cars than BMW per year(2.5mio vs 100k)[8][9]. However, Tesla can easily update their software which Mobileeye cannot because they deploy through BMW.</p><p><strong>What’s next?</strong></p><p>Time will tell whether level 5 mass consumer self-driving cars will be deployed using LiDARs or not. Current mass consumer vehicles such as BMWs and Teslas certainly cannot increase retail prices by ~$75k just to add a LiDAR that may or may not provide value in the future so they will continue to invest into alternatives.</p><p>The LiDAR space stays hot with Cruise (GM) recently acquiring a LiDAR startup and others also trying to get the LiDAR cost and weight down [10][2].</p><p>Thanks for reading</p><p>Derek</p><p><strong>References:</strong></p><p>[1] <a href="https://techcrunch.com/2017/04/21/waymo-accuses-uber-of-hiding-a-secret-lidar-device/">https://techcrunch.com/2017/04/21/waymo-accuses-uber-of-hiding-a-secret-lidar-device/</a></p><p>[2] <a href="https://cleantechnica.com/2016/07/29/tesla-google-disagree-lidar-right/">https://cleantechnica.com/2016/07/29/tesla-google-disagree-lidar-right/</a></p><p>[3] <a href="https://www.recode.net/2017/9/4/16250252/chris-urmson-aurora-self-driving-car-autonomous-vehicle-human-safety-recode-decode-podcast-swisher">https://www.recode.net/2017/9/4/16250252/chris-urmson-aurora-self-driving-car-autonomous-vehicle-human-safety-recode-decode-podcast-swisher</a></p><p>[4] <a href="https://www.forbes.com/sites/samabuelsamid/2017/08/16/fiat-chrysler-signs-on-to-bmwintelmobileye-automated-driving-partnership/#307114db63f4">https://www.forbes.com/sites/samabuelsamid/2017/08/16/fiat-chrysler-signs-on-to-bmwintelmobileye-automated-driving-partnership/#307114db63f4</a></p><p>[5] <a href="https://www.recode.net/2017/3/13/14907184/why-intel-15-billion-mobileye-deal">https://www.recode.net/2017/3/13/14907184/why-intel-15-billion-mobileye-deal</a></p><p>[6] <a href="http://www.mobileye.com/our-technology/">http://www.mobileye.com/our-technology/</a></p><p>[7] <a href="https://www.youtube.com/watch?v=b_lBL2yhU5A">https://www.youtube.com/watch?v=b_lBL2yhU5A</a></p><p>[8] <a href="https://www.reuters.com/article/us-tesla-recall/tesla-to-recall-11000-model-x-suvs-due-to-seat-issue-idUSKBN1CH32R">https://www.reuters.com/article/us-tesla-recall/tesla-to-recall-11000-model-x-suvs-due-to-seat-issue-idUSKBN1CH32R</a></p><p>[9] <a href="https://www.recode.net/2017/4/3/15160462/tesla-ford-deliveries-record-sales">https://www.recode.net/2017/4/3/15160462/tesla-ford-deliveries-record-sales</a></p><p>[10] <a href="https://techcrunch.com/2017/10/09/cruise-acquires-strobe-to-help-dramatically-reduce-lidar-costs/">https://techcrunch.com/2017/10/09/cruise-acquires-strobe-to-help-dramatically-reduce-lidar-costs/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b03de848f1fd" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>