<?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 Pauli Isoaho on Medium]]></title>
        <description><![CDATA[Stories by Pauli Isoaho on Medium]]></description>
        <link>https://medium.com/@pauli.isoaho?source=rss-265e035d125------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*ro1FH39X7Ybx4mOd.</url>
            <title>Stories by Pauli Isoaho on Medium</title>
            <link>https://medium.com/@pauli.isoaho?source=rss-265e035d125------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 10:00:46 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@pauli.isoaho/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[Uber robocar aftermath, and beyond]]></title>
            <link>https://medium.com/@pauli.isoaho/uber-robocar-aftermath-and-beyond-9800684ed5e3?source=rss-265e035d125------2</link>
            <guid isPermaLink="false">https://medium.com/p/9800684ed5e3</guid>
            <category><![CDATA[autonomous-vehicles]]></category>
            <category><![CDATA[uber]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <dc:creator><![CDATA[Pauli Isoaho]]></dc:creator>
            <pubDate>Fri, 23 Mar 2018 18:23:41 GMT</pubDate>
            <atom:updated>2018-03-23T18:23:41.083Z</atom:updated>
            <content:encoded><![CDATA[<p>18th of Mach 2018 it happened, autonomous car, with human driver, hit a person, who was killed. Immediately news was like terrorist news, “Uber killer robocar etc.” Then news started to be more reasonable, like “Even human driver could not have saved the over driven person”, “She jumped from nowhere”, “It was not robot car’s fault” and indeed if you watch that footage you can see that she came from now where. Here is the video <a href="https://www.youtube.com/watch?v=YRUUT3AKK50">https://www.youtube.com/watch?v=YRUUT3AKK50</a></p><p>The good thing here is that everything is logged, like airplanes famous black box, so it can be easily seen what robocar did and what it didn’t. This is much better situation than asking later from (human) eyewitnesses or passengers or even from driver, which motives are most likely biased. Soon we will know what actually happened, only facts not assumptions.</p><p>Robocars don’t see! It is common misunderstand to use word “see” for robots. Robots (cars) <em>perceive </em>world around them and use this information to make decision what to do next. Robot cars use multiple sensors for perceiving world around. Almost all robocars have following sensors, LIDAR, RADAR and Stereo camera. We don’t go here for details but here is short explanation what they are and what they do. Lidar sends laser beam and because it sweeping both horizontally and vertically, it creates point cloud of environment and can detect objects about round 100 meters away. RADAR is based on radio frequency waves and it can detect objects round 150 meters away but not so accurately than LIDAR. Stereo cameras are main used keep robot car in lane and when in enough light, also to detect objects up to 30 meters. Note that all these sensors are overlapping, so robocar is not relying only to one sensor but many. Below is picture of Uber autonomous car sensors</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_63h0FPuB_OyvLLx." /></figure><p>So, now these sensors are perceiving environment and feeding distance measurement data to next step, analyser, with thousands of measures in minute. Now the analyser which is now days deep learning neural network tries to identify objects from thousands of measures and also estimate their speed and direction aka trajectory. If objects trajectory is on autonomous car’s trajectory, so in collision course, analyser calculates the best action to do, like braking and/or avoiding.</p><p>Human is useless! The biggest problem with evolving robocars is not technology or even price but human. When human driver is focused, he/she can react under 500ms, but what happen when you sit to autonomous car? The more the car is automated the more driver is relaxed and starts doing something with phone, laptop or newspaper. Now, if an autonomous car fails for reason or another, they still do, driver’s reaction time is in seconds, and most cases much too late.</p><p>The machine should have done better! Without knowing exactly what sensors model (range, measurement frequency) Uber uses in their autonomous car, it can’t be said with 100% confidence what was cause for this fatal accident but is clear that technology failed. RADAR and LIDAR had most likely easily spotted, i.e. measured the object, bicyclist, as we humans know, because it was on their range. Now comes the trickier part, analyser tried to analyse <em>what</em> is that object and what is its <em>trajectory. </em>Some wrote that here analyser failed to classify this object to bicycle or walking human or something known but that can’t be the case, because the analyser still could calculate the unknown object’s trajectory. The promise of autonomous car is that they are <em>better </em>drivers as humans, not <em>worse</em>.</p><p>Finally, who will be prosecuted? The human driver, Uber, Volvo, no-one (Okay, not the victim anyway) In many countries the human is responsible driver always, whatever is the autonomous car’s autonomy level. But as stated earlier, it is almost <em>unhuman</em> for human to keep focused all the time when car is, kilometre after kilometre, driving smoothly, accurately and reliable. So, should we then make car manufacture guilty? The biggest car manufacturer in Europe has said that they will insurance all their autonomous car and this insurance will cover all accidents happened to robocars.</p><p>So, what will happen now? Courts and lawyers will do their work, but what will this mean to robocar development? This will not be the last robocar accident where human is killed. Depends little bit of who or which will be named as guilty, if it’s technology, then most likely current robocar driving permissions will be halted, for a while. Little bit like when a plane crash, then same model planes are grounded. Anyway, this was not a breaking news, not even any news, this was inevitable, it was foreseen, it was just matter of time, now it happened, honeymoon is over. Let’s learn from, take notes, and then move on, robocars are coming!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9800684ed5e3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI inflation]]></title>
            <link>https://medium.com/@pauli.isoaho/ai-inflation-557efb518009?source=rss-265e035d125------2</link>
            <guid isPermaLink="false">https://medium.com/p/557efb518009</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Pauli Isoaho]]></dc:creator>
            <pubDate>Wed, 21 Mar 2018 06:20:09 GMT</pubDate>
            <atom:updated>2018-03-21T06:20:09.389Z</atom:updated>
            <content:encoded><![CDATA[<p>Anaconda, Tensorflow and Keras, that is all you need to create machine learning (ML) programs. Without coding skills, without ML understanding, just copy and paste.</p><p>What does this mean, that almost anyone who has computer and can read English can do ML programs? Will there be abundant flow of companies entering ML market, with no understanding of data, algorithms and domain?</p><p>Anaconda can be understood as platform which includes all needed python libraries for machine learning calculations, so user don’t need to collect and install them one by one separately. It also includes Conda, which is library and version management tool. It automatically tracks new versions of installed libraries and also their dependencies. Convenient, just install Anaconda, then you are halfway.</p><p>Next you need to install Tensorflow, it is lower level ML command tool. It includes ready made functions to make most basic ML commands like create neural network. But to make things even easier, you need to install Keras, it is higher level ML command tool. It works above Tensorflow and uses it’s lower level commands, wraps and bundle them, so, for example, you can create deep neural network with three commands, neat, hah?</p><p>Okay, now tool are setup, now you need to do Python coding? No, not really, copy -&gt; paste is enough, and then just modify parameters. One of the best place to start is here <a href="https://machinelearningmastery.com/python-machine-learning-mini-course/">https://machinelearningmastery.com/python-machine-learning-mini-course/</a></p><p>So, do these tools, which make starting ML very easy, mean that anyone from the street, who can read English, has access to computer and wants to, can start selling machine learning projects? And then failing, of course, which would be bad for real ML companies. Some kind of ML snake oil sellers? Answer is NO.</p><p>If mastering these tools, or even algorithms, would be enough, it would mean same that if you can keep hammer in your hand and even know which way to hit with it, it still doesn’t mean that you could build a proper house.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=557efb518009" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[End of touch?]]></title>
            <link>https://medium.com/@pauli.isoaho/end-of-touch-a5ad867a2d47?source=rss-265e035d125------2</link>
            <guid isPermaLink="false">https://medium.com/p/a5ad867a2d47</guid>
            <category><![CDATA[voice-recognition]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[voice-control]]></category>
            <category><![CDATA[nlp]]></category>
            <dc:creator><![CDATA[Pauli Isoaho]]></dc:creator>
            <pubDate>Sat, 10 Mar 2018 10:10:53 GMT</pubDate>
            <atom:updated>2018-03-10T10:10:53.547Z</atom:updated>
            <content:encoded><![CDATA[<p>2007 world changed and never game back again, Apple published it first phone with touch control. After that touch user interface changed the rules of phone game, leading to market exit for a previous market dominant player. Touch user interface changed the game because it was very intuitive way to use phone, it felt like it should had always been so. Now, 2018, things are going to change again.</p><p>Thanks to progress in machine learning the accuracy and speed of Natural Language Processing (NLP) has reached that level that it can be used in phones. Both Apple IoS and Android offer full voice control of phone and also dictating messages.</p><p>Is voice user interface intuitive? Yes, would you like to chat or talk with friend who is sitting next to you? Is it more intuitive that touch user interface? Yes, it will change paradigm and soon we will think that, it should had always been so.</p><p>I use my Android phone with voice almost all the time, and only what stops me using it ALL the time, is shame. Speaking to phone in public feels still awkward, but of course still will change soon when more and more people are using voice user interface.</p><p>Of course, voice user interface doesn’t stop to phones. Homebots, like Amazon Echo, Google home and Apple Home bot, can already used in many ways and soon they can be used like in science fiction movies, you come home and say “warm up my meal, open my favourite channel etc.” But the difference is that they don’t have touch user interfaces.</p><p>We can see also multiple devices. like printers, home appliances, cars etc, which are getting now touch user interface, will be just temporary, and over ruled by voice user interface. This strong suggest that voice control is superior and more intuitive that touch control.</p><p>In my voice user interface testing there is only one black area, and it is laptops, or computers in general. Voice recognitions is not yet in that level that I could use windows 10 and it’s application only by voice, actually it is so bad that I don’t use it now after testing period. So there is work to do, and it is not only normal NLP challenge but also the logic. When you say “comma”, do you mean that there should be “,” or world “comma”, this might sound trivial, but you get what I mean. Like I’m writing this using 300 years old invention, keyboard. Although, the promise is clear, soon you can dictate your “writings”, calculate your spreadsheets and so on with voice.</p><p>Voice control is coming everywhere, where nowadays you seen any user interface, is it ticket machine, ATM, phone, TV etc. it will be voice controlled tomorrow. Like in Shanghai subway station where you just <em>say </em>destination and it will print you ticket. Touch is dead, voice is here!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a5ad867a2d47" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI tax]]></title>
            <link>https://medium.com/@pauli.isoaho/ai-tax-c34d4ddd4b79?source=rss-265e035d125------2</link>
            <guid isPermaLink="false">https://medium.com/p/c34d4ddd4b79</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[robot-tax]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai-tax]]></category>
            <dc:creator><![CDATA[Pauli Isoaho]]></dc:creator>
            <pubDate>Mon, 19 Feb 2018 20:32:20 GMT</pubDate>
            <atom:updated>2018-02-19T20:32:20.220Z</atom:updated>
            <content:encoded><![CDATA[<p>There have been a lot of talk about robot tax, meaning that when robots are used then there could be some kind of penalty tax. This idea has been shot down many times, but discussion still continues.</p><p>But now back to AI tax. AI is threatening take more and more humans jobs, leaving hundreds of thousands people to without work. Unemployment costs are skyrocketing, so where to get money? If AI is replacing workers maybe AI need to pay taxes?</p><p>How this could be done? One choice is to calculate how many employees are replaced and take their tax amount, which then the company providing that AI would pay. Or maybe we could legalise Ai, making it as citizen, so then we could tax it directly? Or we could tax more AI companies, because they would make enormous money selling these AI systems.</p><p>Nice ideas, but all doomed. In order AI tax to work all countries in the whole world should agree on it, and that has never worked and never will be. If any country would slip away from this agreement, they could get massive competitive advantage. So not going to work!</p><p>Then what is going to happen? This, AI companies and their owners will make billions after billions, unemployment is skyrocketing, and then when there is risk that whole society collapse, then there will be AI tax.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c34d4ddd4b79" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Books review]]></title>
            <link>https://medium.com/@pauli.isoaho/books-review-c0851d69e49b?source=rss-265e035d125------2</link>
            <guid isPermaLink="false">https://medium.com/p/c0851d69e49b</guid>
            <category><![CDATA[book-review]]></category>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[book-recommendations]]></category>
            <category><![CDATA[human-behavior]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Pauli Isoaho]]></dc:creator>
            <pubDate>Thu, 01 Feb 2018 09:42:27 GMT</pubDate>
            <atom:updated>2018-02-01T09:42:27.733Z</atom:updated>
            <content:encoded><![CDATA[<p>Here are some latest books I red, they are roughly in time order</p><p><strong>Guns, Germs and Steel</strong>, Jared Diamomd<br><a href="https://en.wikipedia.org/wiki/Guns,_Germs,_and_Steel">https://en.wikipedia.org/wiki/Guns,_Germs,_and_Steel</a><br>Stops racists bullshit about white “race” supremacy. Only bad thing that I red it just now, not 10 years ago. Books name could be also “Trade, writing and agriculture” because these things differentiated area and nations which went to supremacy and other’s not.</p><p><strong>Machine Learning</strong>, Ethem Alpaydin<br><a href="https://mitpress.mit.edu/books/machine-learning-1">https://mitpress.mit.edu/books/machine-learning-1</a><br>If you want to understand how AI works, but don’t want to kill your excitement with horrible math, then this one is for you!</p><p><strong>101 Very Finnish problem</strong><br><a href="https://www.gummerus.fi/fi/kirja/9789512408597/101-very-finnish-problems/">https://www.gummerus.fi/fi/kirja/9789512408597/101-very-finnish-problems/</a><br>Nice stories about Finns, outside a box</p><p><strong>Sapiens</strong>: A Brief History of Humankind, Yuval Harari<br><a href="https://en.wikipedia.org/wiki/Sapiens:_A_Brief_History_of_Humankind">https://en.wikipedia.org/wiki/Sapiens:_A_Brief_History_of_Humankind</a><br>How we came here, where we are today? Extremely well written book about history of human, even this is not a history book.</p><p><strong>Thinking, Fast and Slow</strong>, Daniel Kahneman<br><a href="https://en.wikipedia.org/wiki/Thinking,_Fast_and_Slow">https://en.wikipedia.org/wiki/Thinking,_Fast_and_Slow</a><br>If Yuval wrote about how we came here, then Daniel writes how we think, and that picture is not rosy. After reading this book, you world is not same anymore.</p><p><strong>Rise of the Robots</strong>: Technology and the Threat of a Jobless Future, Martin Ford<br><a href="https://en.wikipedia.org/wiki/Rise_of_the_Robots_(book">https://en.wikipedia.org/wiki/Rise_of_the_Robots_(book</a>)<br>Already classic in robot time but excellent book about how robots will change society. Has also answer to question why Trump won president election.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c0851d69e49b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>