<?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 Pat D&#39;Ambrosio on Medium]]></title>
        <description><![CDATA[Stories by Pat D&#39;Ambrosio on Medium]]></description>
        <link>https://medium.com/@pdambros?source=rss-f2632ab25e43------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*Wy6TiE7EGeqHvwdoQ__cvg.jpeg</url>
            <title>Stories by Pat D&amp;#39;Ambrosio on Medium</title>
            <link>https://medium.com/@pdambros?source=rss-f2632ab25e43------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 22:03:19 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@pdambros/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[Machines Are Hijacking Writers]]></title>
            <link>https://medium.com/@pdambros/machines-are-hijacking-writers-997db76ba119?source=rss-f2632ab25e43------2</link>
            <guid isPermaLink="false">https://medium.com/p/997db76ba119</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[future-of-writing]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[writing-tips]]></category>
            <dc:creator><![CDATA[Pat D'Ambrosio]]></dc:creator>
            <pubDate>Thu, 17 Sep 2020 18:25:23 GMT</pubDate>
            <atom:updated>2020-09-17T18:25:23.476Z</atom:updated>
            <content:encoded><![CDATA[<p>Or, Are they?</p><h3>~Patricia D’Ambrosio</h3><p>~ 6 min read</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*chh80hleXEVXEqjQr31LjQ.jpeg" /></figure><h3>Machine writing is here and your writing’s already adapting</h3><p>I love writing and crafting its elements: finding the best words, structuring the sentence, thinking about grammar and meaning…. I come from a family of writers, so perhaps writing is my lifeblood. Either way, it’s also my trade. It got me high marks in my Master’s program and today, it pays the bills: I’m a UX and Technical Writer. As such I develop scripts for videos and animated gifs, am a content strategist and have written marketing copy….</p><p>My tech writing began in search and went on into text analytics and AI. I’ve been a fan of both SEO and analytics ever since. But I never expected that technology might put my job at risk.</p><p>In his 10/15/2019 New Yorker article entitled <em>The Next Word: Where will predictive text take us?</em> John Seabrook asks whether machines will someday replace writers. He digs deep into a question anyone who’s used Google’s Smart Compose, introduced in May 2018, might ask.</p><p>How does Google’s Smart Compose predict the end of my sentences?</p><p>Google’s artificial intelligence (AI) running in the background makes predictions based on its millions of worldwide users. As you’re entering your sentence, Google’s AI is supplying light gray text based on what other users have entered. Like what you see? Hit Tab, write your next sentence and see its predictive text….</p><p>Which also leads me to wonder how Google determines which Smart Reply buttons pop up. How does Google know what might be an appropriate response to each email I receive?</p><p>Are these short, bland, depersonalized, and perhaps inconsiderate replies always appropriate? This is especially true when the email to which I’m responding is lengthily and detailed.</p><h3>Personal style is identifiable</h3><p>Predicting the end of my sentences is one thing; writing articles, Help, or UX text is another. As I write, I structure my sentences, choose punctuation and words — <a href="https://www.lexico.com/en/definition/function_word">function words</a>, which contribute to sentence syntax instead of meaning, in particular…I’m creating a writing style. This is all according to a 1/16/2013 The Sydney Morning Herald article by Gemma Khaicy entitled <a href="https://www.smh.com.au/technology/why-hackers-should-be-afraid-of-how-they-write-20130116-2csdo.html"><em>Why hackers should be afraid of how they write</em></a>. Each person’s writing style is uniquely different. So different that the science of stylometry has been used for hundreds of years to confirm the authorship of works like Shakespeare’s plays, Homer’s <em>Iliad</em>…. Now stylometry can take a 5,000-word sample and with 80% accuracy track any of us on the web regardless of IP address….</p><p>Sounds scary. Maybe not? As I slowly adapt to hitting Tab to complete my sentences, my emails lose personal sentiment. They conform to what the general user, which Google has defined, writes. As more tools like these are unleashed, perhaps we’ll all become writing machines producing standardized pieces. In the best-case scenario: Editors instead of writers?</p><h3>Unique Human Capabilities</h3><p>IBM’s Speech by Crowd, which according to Mr. Seabrook, was developed by Noam Slonim an Israeli Research Fellow, processed almost 2,000 essays. Mr. Seabrook opined that the articles could have been written, less formulaically, by seventh graders.</p><p>While Mr. Seabrook says the AI used both rules and deep learning to summarize them into a pair of pro and con articles, Jesus Rodriguez in his Inside Project Debater Speech by Crowd Medium cites different analytics. (This might be attributed to the specific use case for Mr. Seabrook’s New Yorker article?)</p><p>Specifically, Mr. Rodriguez, points to three capabilities once thought to be unique to humans:</p><p>● Concept abstraction: Able to abstract concepts from text, even when these concepts aren’t labelled data.</p><p>● Argumentative content search: Find claims in large text corpuses. Relying on deep neural networks they were able to identify claims sentences that affirm or deny a specific assertion.</p><p>● Listening comprehension: Able to listen to audio streams without using text and mark the arguments used by speakers.</p><p>Currently, the Washington Post’s Jeremy Gilbert, according to Mr. Seabrook, oversees the Post’s robotic newshound Heloiograf. Heliograf hasn’t replaced any freelancers to-date, but it’s great at collecting data. The machine takes the data, puts it into a spreadsheet and determines the key play in a sports game or a key election issue. Heliograf wraps this all in appropriately generated language. For example, one of its reports stated:</p><p><em>The Yorktown Patriots triumphed over the visiting Wilson Tigers <br> in a close game on Thursday, 20–14.</em></p><h3>Benefits for all?</h3><p>For Mr. Seabrook, automated writers such as GPT-2 aren’t all substandard or unacceptable. Consider folks for whom the dominant language is secondary. Since 2016, he says, Google Translate has begun to replace human translators in areas such as medicine. (My American son, Paul D’Ambrosio PhD., running a Chinese Philosophy translation company in China, has joked about Google translation being a helpful partner at times and for specific texts.) That’s because it works using neural nets, which like the human neural network learns recognition through trial and error. This is a huge leap from machine learning, which is a labor-intensive process.</p><p>And what about people who need to write memos, email, and reports every day? These non-professional writers might find that their lives improve with automated writing. I’d speculate that these writing types might not always be the final product, but they’d reduce the workload, provide a basic, editable piece.</p><p>Turning these benefits to my own work as a Help writer. Writing product Help is a highly formulaic process. Sentence structures are often defined, function words are repetitive, and descriptions are few to non-existent. As an example, consider the <a href="https://slack.com/help/articles/205239997-Pin-messages-and-files#ios-1">following steps from the Slack Help Cente</a>r:</p><p>Steps, such as those above, use highly repetitive language. Coupled with screen recognition, These steps might someday be automatically reproducible leaving me and my colleagues to edit and add caveats.</p><h3>Harm</h3><p>Mr. Seabrook focuses much of his article on GPT-2 created by Open AI, which is what he calls “a kind of supercharged version of Smart Compose.” It runs on vast neural networks and according to is chief scientist, Ilya Sutskever, “It’s [works] like <em>alchemy</em>.” (p. 55) But GPT-2 can get a second round of training on top of its general language skills. For example, if you feed it a bunch of comments from a site like Amazon, it can spit out phony customer reviews.</p><p>Fake news? The potential exists for it to post divisive political information on an industrial scale — much cheaper he says, than paying Russian college students. He notes that these students don’t have the command of the English language that the machine does.</p><h3>The bottom line</h3><p>Machine learning has been around for awhile and today it’s possible for an algorithm to generate a distinct style and narrative ideas. Algorithms can create a blueprint for a story Twinkle, Twinkle that was good enough to be published in <a href="https://www.technologyreview.com/2020/06/17/1003665/i-used-an-algorithm-to-help-me-write-a-story-heres-what-i-learned/?truid=dd5a8f98d1e3feef08af79efcba67fe1&amp;utm_source=the_download&amp;utm_medium=email&amp;utm_campaign=the_download.unpaid.engagement&amp;utm_term=non-subs&amp;utm_content=08-13-2020">Wired</a>. Stephen Marche used the SciFiQ program helped him do this. I wonder if we could create a TechQ program to run for technical and UX writing? Would it understand voice and tone? Could we add branding, minimalism, a dose of humor where appropriate and let’s not forget branding and usability.</p><p>Perhaps my writing will change and become better using an AI generated draft…or perhaps, like the automated email responses I avoid, AI won’t have more than a shiny new object impact. Either way, it’s up to me to stay up with user trends, know my audience, and write well.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=997db76ba119" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>