<?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 Johan Uddståhl on Medium]]></title>
        <description><![CDATA[Stories by Johan Uddståhl on Medium]]></description>
        <link>https://medium.com/@baktakt?source=rss-53dc4300016d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*NGxlyp79Y7m1VMBmOqWGBA.jpeg</url>
            <title>Stories by Johan Uddståhl on Medium</title>
            <link>https://medium.com/@baktakt?source=rss-53dc4300016d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 01:55:37 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@baktakt/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[…when all we ever needed was a text box]]></title>
            <link>https://medium.com/@baktakt/when-all-we-ever-needed-was-a-text-box-c672c52a0dca?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/c672c52a0dca</guid>
            <category><![CDATA[visual-design]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[user-experience-design]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Wed, 01 Jan 2025 14:55:57 GMT</pubDate>
            <atom:updated>2025-05-27T07:57:24.258Z</atom:updated>
            <content:encoded><![CDATA[<h4>or how 2025 will be back to basics for the web</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hNC2HbsbmIXxrMKgE8a6sw.jpeg" /><figcaption>Welcome back, old friend</figcaption></figure><p>All of my articles are free to read, read for free by <a href="https://medium.com/@baktakt/when-all-we-ever-needed-was-a-text-box-c672c52a0dca?source=friends_link&amp;sk=f0df09e562c0936bf2ec992b86fb43e2">clicking this link</a>.</p><p>A few times each year, in discussions with fellow developer friends, we all agree we’re just building the same things over and over again, and they all have one thing in common: <strong>text inputs</strong>.</p><p>Now, as well as 20 years ago we have always been wizards of the text boxes, experts at turning HTML form elements into magical web experiences.</p><p>In this post I will try to look back at my career in coding text boxes, in search for the answer to the question:</p><blockquote>Why did we bother building all these fancy web interfaces, when all we ever needed was a text box?</blockquote><p>Hold on friends! Before you get upset, let me elaborate on that.</p><p>Recent developments, especially over the past few years, have enchanted us all by the seemingly magical powers of AI in general and LLMs (Large Language Models) in particular. New services specialized in different kinds of content generation are now being launched at light speed. <a href="https://chatgpt.com/">ChatGPT</a> for text, <a href="https://www.midjourney.com/">Midjourney</a> for images, <a href="https://suno.com/">Suno</a> for songs, <a href="https://github.com/features/copilot">Copilot</a> for coding just to name a few. Judging from the output from these services they would categorize as extremely powerful and complex systems. They create things in milliseconds that were not even possible just a few years back.</p><p>And they all have one thing in common. The main interaction with them is done via one of the <a href="https://developer.mozilla.org/en-US/docs/Web/HTML/Element/input/text">simplest types of HTML components we have</a>:</p><pre>&lt;input type=&quot;text&quot;&gt;</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kRwX58GqxRPB3IL2mMPvkA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q8r7OsQElIuKHEnpKwXN7A.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MjEjHXKfvMxEWCipPt9D1Q.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2PqNN5jYs5aYldqpC6dTNg.png" /><figcaption>See the pattern there, wizard?</figcaption></figure><p>This particular trend has occupied my thoughts for a while now.</p><p>What if users just want to use their natural language when interacting with systems, just as we expect when interacting with other humans?</p><p>Which opens for followup questions like:</p><ul><li>Did we deliberately make the web more complex than it needed to be?</li><li>Why did we spend decades adding complexity, when the most powerful tool we’ve ever built might just be a single text box?</li></ul><h3>Looking Back</h3><p>See, when I started coding, we had <strong>developers</strong>, or coders if you may. No backend, no frontend, hell I even heard of such a thing as an architect years into my career. We just did it all without bothering what we were called.</p><p>The interfaces were simple and often performed very simple tasks, like sending a mail or signing up to something. We used plain HTML, form tags and at best some CSS. Things were posted to some perl script at the backend via the action attribute.</p><pre>&lt;form action=&quot;/cgi-bin/script.cgi&quot; method=&quot;POST&quot;&gt;<br>  &lt;label for=&quot;name&quot;&gt;Name:&lt;/label&gt;<br>  &lt;input type=&quot;text&quot; id=&quot;name&quot; name=&quot;name&quot;&gt;<br>  &lt;input type=&quot;submit&quot; value=&quot;Submit&quot;&gt;<br>&lt;/form&gt;</pre><p>The interfaces were fugly and sometimes quite hard to use, but that didn’t matter. It worked, and that was enough for us.</p><h4>App #1: “CMS”</h4><p>For me it started with reading the CGI Bible back in 1996, building HTML forms posting stuff to a CGI backend. In my case I built something that today probably would be described as a CMS. But back then we hadn’t invented all those webzy cool acronyms we are so familiar with today. I just saw a problem that needed a solution. Nevertheless, it was just a simple <strong>form </strong>application where you entered text and pressed a button and things got posted.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/366/0*EoYgObfdtbgBah57.jpg" /><figcaption>618 pages of pure web magic</figcaption></figure><p>A few years pass and those simple text boxes turned into radio buttons and checkboxes, making it easier and faster to input text into the machine we now called the web.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/621/0*3rc-Mw02SIzMrBz7.gif" /></figure><h4>App #2: Peking Pizza</h4><p>So moving on to 2001, and university studies. This meant moving my beloved desktop PC from Gothenburg to Norrköping. A few weeks in me and some friends saw the urgent need of knowing where the best pizzas were made, so we created “Peking Pizza”, <a href="https://en.wikipedia.org/wiki/Norrk%C3%B6ping">named after the nickname of Norrköping</a>.</p><p>The website? A <strong>form</strong> application with <strong>checkboxes</strong> of pizza ingredients. The user selected what they wanted on a pizza and the web machine spat out the places where this particular pizza could be ordered, along with ratings for the pizza place etc.</p><p>Backend was made in PHP, but let’s not tell anyone.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/810/1*XzrTWCJSO87SzGXt3C01YQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/668/1*NTPfe5fjm0x7KgZsmzcGKA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*qFy947Ge9w7nYJ-g" /><figcaption>Yes, yes there was a UI, but would you <strong>look</strong> at those marvelous form inputs! All we ever needed…?</figcaption></figure><h4>App #3: Snapshot</h4><p>If I told you me and a university friend created Instagram two years before Instagram you would probably laugh. But I would argue this is exactly what we did. We had this crazy idea that with the new world spanning web technology people will probably want to share things online. For the world to see what their lives looked like.</p><p>The only problem was this was before fast web communication via phone, aka 3G, existed. So people had to send their images via MMS, which turned out to be a shit protocol that was too slow and too expensive to base a photo diary app on.</p><p>But we certainly built it, and in november 2004 we launched <a href="https://web.archive.org/web/20041127020155/http://snapshot.se/">snapshot.se</a> (thank you waybackmachine). Hell, we even went to the folks at the city officials advicing new companies and pitched the app. Do I need to say they did not quite get it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3jthGPO-Jq-kUpXzQBqJbA.png" /><figcaption>It was Instagram and Snapchat, two years early. Go figure…</figcaption></figure><p>The site was in Swedish at launch, but heres the start page text in english:</p><blockquote>At snapshot, we believe it should be easy to share your experiences. That’s why we created this site. Here, your friends or family can easily, with just a few clicks, see the things you want to share. It can be a funny monkey you managed to capture on camera in Thailand or your latest pike catch in the country. You choose what you want to share.</blockquote><blockquote>After you’ve registered here at snapshot, it’s just a matter of sending in pictures, and creating posts on snapshot is as easy as putting on your shoes.</blockquote><p>But now, to the point. The site was based on <strong>forms and text boxes </strong>where you could upload you images in an easy way, either by using the website or by MMS:ing or emailing (look, another text based application) an image to the system.</p><h4>App #4: Get on that bus</h4><p>Oh yes, enter the world of handheld web! Although testing this out early by connecting my Ericsson T39 to my pocket computer and browsing the web via GPRS, the handheld web was not really useful to access until Apple released the iPhone. Shortly following the first iPhone launch, Google launched their counterpart.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/0*IlOuWzvZUJsTp879" /></figure><p>And with that, I downloaded the Android SDK and began developing my very first app: Get on that bus.</p><p>Get on that bus was an app helping lazy people knowing when to leave your home to get in time for the bus or tram.</p><p>And yes, you guessed it. The main input method was the text box for finding from where you were leaving, and how many minutes left until you needed to walk, or most often run, to the bus.</p><h4>Hello Javascript — Goodbye Simplicity</h4><p>Most of my humble coding contribution to the world has been spent building web and mobile applications. As years have passed the user interfaces has gotten more and more complex, demanding more code on the frontend parts of the systems. This is where javascript came to the “rescue”.</p><p>Interfaces needed partial updates in real time, we invented words like infinite scroll, and every web page of any dignity had an image carousel part in the top of the page. And with this, code complexity moved to large extent from the backend to the frontend. When the script.js file, where we put all javascript into one single file became impossible to maintain we started seeing frontend build tools like grunt, gulp and webpack. Building frontend code! We laughed, complexity grew.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/924/0*rg20-IHDRhWPCBXm.png" /><figcaption><strong>Grunt</strong>, by far the coolest logo of the javascript bundling tools</figcaption></figure><p>Developers like me found themselves building web api’s to support the increasingly complex user interfaces. And with us spending time on the “backend” of things, a new type of developer arose:</p><p><strong>The Frontend Developer</strong>. These folks loved interfaces and things that moved. When you finished a signup form there were fireworks. When you typed something wrong in the text box a message magically appeared claiming you messed something up in your email address.</p><p>And then the styling. Suddenly those form elements we got so used to were boring and ugly. They needed style. And since CSS was too rudimentary we invented CSS extensions like SASS and LESS adding even more complexity to the already complex frontend applications.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dVQWfo5VTFF4T1m3" /><figcaption>Sass — Syntactically Awesome Stylesheet</figcaption></figure><p>With the increasing complexity the user interfaces got harder and harder to use, which of course was a huge problem for sites like e-com stores wanting to sell as much as possible to the users. This called for the rise of a few new roles in web development.</p><p>The <strong>User Experience Designer (UX) and the User Interface Designer (UI). </strong>UX Designers represented and empathized with the users of the complex interfaces we created. They liked the users so much they would even offer themselves to do interviews and and test interfaces with them. Sometimes face to face! With their newly found knowledge and user insights, the interfaces created by the User Interface Designers and the Frontend Developers could be made a bit less confusing to the users.</p><p>As 2016 turned 2017 I actually wrote a post on the challenges of getting all of these roles working together in teams. It’s quite a good side story so if you need a pause from my ranting here, I’d recommend you <a href="https://medium.com/wearehumblebee/2017-a-love-story-f501bf3c730b">read it</a>.</p><p><a href="https://medium.com/wearehumblebee/2017-a-love-story-f501bf3c730b">2017 — a love story for the web</a></p><p>As things continued to evolve we even started fiddling around with the way things are rendered in the web browser, and the DOM (Document Object Model) became the main character. When engineers at Facebook grew tired of the problems attached to building user interfaces at scale, they created React, and with that introduced the Virtual DOM, an in-memory representation of the real DOM. Instead of directly manipulating the DOM (a slow process), React updates the Virtual DOM, calculates the most efficient way to apply changes, and updates only the necessary parts of the real DOM. This approach dramatically improved performance, especially for dynamic and interactive applications.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*eNgVGrRdFw3pYOa0.png" /><figcaption>React — the answer to user interfaces at scale?</figcaption></figure><p>With this came also the pattern to build web applications with components, that made it easier to re-render parts of a web page and maintaining code updates in complex user interfaces.</p><h4>At the crossroads?</h4><p>Continuing along this path for years, we now find ourselves with large design systems, consisting of hundreds of complex components, and millions of lines of code. Heck<strong>, </strong>we even have <a href="https://thedesignsystem.guide/blog/a-guide-for-calculating-design-system-costs">guides</a> on how to calculate design system costs, to be able to choose the right design system, or worst case, build your own. And they all serve one purpose; Helping the users making sense of an over engineered world wide web<strong>.</strong></p><p>So what do you think? Do I have a case here, stating that maybe we have created a web that just might be a bit over engineered? A bit too complex?</p><p>And maybe, just maybe, it is not only in response to user insights, but also because we like to over complicate things?</p><p>We are at the crossroads and here are my thoughts on what 2025 and coming years have in store for us.</p><h3><strong>2025 — Back to Basics</strong></h3><p>Just looking at the new tools released the past year I think it’s clear the web is heading back to simplicity. LLMs (Large Language Models) and AI-driven applications is now showing how to use <strong>natural language</strong> as an interface, reducing complex designs to a single input. You guessed it: the <strong>text box</strong>.</p><p>So what does this mean for the future of the web?</p><h4>Major Changes In Roles And Responsibilities</h4><p><strong>Developers as Problem Solvers</strong></p><p>Much like in the early days of web development, when developers needed to solve problems regardless of language, frontend or backend, developers in 2025 will focus less on building intricate user interfaces and more on solving problems through functionality and efficiency. This will most probably change the developer role.</p><p>I think we will see an evolution of what we usually refer to as micro service architectures, where these micro services now instead become agents, powered by large language models, that have a clear purpose and give specific context and capabilities to the general LLMs.</p><p>Developers need to be aware of what services to use, what to build ourselves and what to buy by the token. The move to api based systems has already pushed us in this direction and this is the next big leap. We will be the jigsaw puzzle masters, putting different already made pieces together into complete systems that will serve the users with relevant data. Here, programming language matters less. You will use what’s relevant in every specific case. And with tools like Copilot and others now widely available, this is a problem of the past. Forget long hours of online courses, instead focus on solving real problems together with your new smarter best friend.</p><p>Framing, understanding, dissecting and then solving problems has always been the key skill for a good developer, and tomorrows landscape will put even more emphasis on this.</p><p>Us developers will need to put a lot of work into how AI-related systems are built, tested and deployed. We will be responsible for making sure these systems are secure to use, user data is handled correctly and that users understand the systems, not to mention all the ethical aspects of utilising AI systems. It will require a lot of upskilling, so make sure you use every hour you can to educate yourself. Try as many tools as you can. Read about the latest advancements. But most importantly; just build stuff.</p><p>I promise you, it will be fun!</p><p><strong>Designers as Empathy Architects</strong></p><p>I know, cheesy, but bear with me. The move towards text-based, conversational systems reduces visual complexity but amplifies the need for strategic, empathetic design.</p><p><em>User Interface Designers</em> will need to transition from creating decorative interfaces to becoming architects of meaningful and efficient interactions. This involves focusing on the clarity and intuitiveness of conversational flows, ensuring that users can navigate systems seamlessly through natural language, both with text and voice. Designers will need to create interactions that are not only functional but also engaging and human-like.</p><p>I believe designers will play a major role in complementing text or speech based systems with contextual interfaces, navigating, guiding and making sense of responses and generated content.</p><p><strong>User Experience Designers</strong> will need to learn how natural language input changes the way we interact with systems. In 2025 I think we will see a lot of new voice to text enabled applications. UX designers will need to learn how to create user interfaces that encourages the user to stay off the keyboard, and instead rely on using voice. Yes, we’ve had assistants like Siri and Alexa for years, but it isn’t until now we are starting to get close to having actual conversations with such systems, and this opens up a pandoras box of new possibilities, where UX can play a crucial part in guiding the users.</p><p>All in all, the<a href="https://medium.com/wearehumblebee/2017-a-love-story-f501bf3c730b"> things I wrote in early 2017</a> still applies. To succeed we need to work in teams, cross functional teams. Developers need to understand, value and respect the empathetic approach of the UX and UI designers, and the designer need to understand the complexity that goes into developing a system, especially an AI powered one.</p><h3><strong>Wrapping up</strong></h3><p>So did we overcomplicate the web?</p><p>Absolutely — but not without reason. Every evolution, from static HTML to complex JavaScript frameworks, was a response to a challenge or a trend.</p><p>But to some extent I do think much of the complexity was unnecessary. The basic goal — helping users get what they need — is achievable with far less.</p><p>Users want interactions that feel intuitive and human, not cluttered menus and overly styled interfaces. Text based tools and image generators available today shows clearly that simplicity is a key element. Instead of creating complex UI components and workflows, we’re moving toward enabling <strong>conversations</strong> with systems. A simple text box, paired with powerful AI, might just be the most reasonable way to get things done.</p><p><strong>TLDR;</strong></p><p>In 2025, the web will increasingly rely on:</p><ul><li><strong>Simple Inputs</strong>: Text boxes, voice commands, and other natural interfaces.</li><li><strong>AI-Driven Logic</strong>: Backends that understand and execute complex tasks without requiring complex UI.</li><li><strong>Scalable Solutions</strong>: Systems that adapt to user needs without requiring new interfaces for every problem.</li></ul><p>This all means a lot of challenges for us working in this industry. We will need to navigate a world where new tools and services are released every day. All claiming to have the power to optimize and revolutionize, yes even replace us. In some cases this will be true, and we need to find the sweet spots where human hands will still be needed.</p><p>And oh, the text box, that beautiful little HTML element, is making a well deserved return.</p><h3>And maybe, just maybe, all we ever needed <strong>was</strong> a text box — to cast our next spell.</h3><p>So a happy new year fellow web loving friends! I surely hope 2025 has a lot to offer, and I look forward to what we all can create together with the extreme advancements in technology we get to experience. Please don’t be shy to show and tell the world about the things you create. I know I will, and I have quite a lot of ideas that deserves to be tested!</p><p>It’s a good time to be a developer! 2025 — Bring it!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c672c52a0dca" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Adventures from Tokyo to Tech Frontlines]]></title>
            <link>https://blog.knowcast.ai/ai-unplugged-adventures-from-tokyo-to-tech-frontlines-404772e5f97c?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/404772e5f97c</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Tue, 10 Dec 2024 13:58:03 GMT</pubDate>
            <atom:updated>2024-12-10T14:02:15.136Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="An exploration of AI usage and its impact across various sectors" src="https://cdn-images-1.medium.com/proxy/0*lgaRPU2hIg77tgj8" /></figure><p>This week’s edition explores the diverse utilization of artificial intelligence, spanning travel, military, education, and corporate sectors. We embark on a journey through Japan aided by an AI travel companion, delve into OpenAI’s latest technological advancements, and uncover intriguing developments in military AI systems. Additionally, we explore how humanitarian organizations and major companies like Google and Adobe collaborate with AI to enhance functionalities and streamline processes. Join us as we unveil how AI reshapes industries and lives globally.</p><h4>Discover Tokyo with Your AI Travel Ally</h4><p><em>Published 09 December, 2024</em></p><p>Join an adventurous 48-hour journey in Tokyo with a digital companion by your side. Explore how AI can transform travel experiences and bring a unique perspective to solo adventures.</p><p><a href="https://www.wired.com/story/48-hours-in-tokyo-with-my-ai-travel-companion/">Read the story</a></p><h4>Generate Videos Effortlessly with OpenAI’s Sora</h4><p><em>Published 09 December, 2024</em></p><p>Dive into OpenAI’s newest tool, Sora, which allows users to create videos effortlessly. This piece uncovers how this innovation is shaping the landscape of video production during the company’s “shipmas” launch event.</p><p><a href="https://www.technologyreview.com/2024/12/09/1108309/how-to-use-sora-openais-new-video-generating-tool/">Read the story</a></p><h4>Humanitarian Efforts Enhanced by AI</h4><p><em>Published 10 December, 2024</em></p><p>A transformative wave is sweeping through humanitarian organizations as they harness AI to drive positive change and extend their reach where it’s needed most.</p><p><a href="https://www.wired.com/story/humanitarian-organizations-artificial-intelligence/">Read the story</a></p><h4>Surge in Developers Embracing Generative AI</h4><p><em>Published 02 December, 2024</em></p><p>Over 140,000 developers engaged in a five-day intensive course on generative AI, showcasing the growing interest and demand in mastering this innovative technology.</p><p><a href="https://blog.google/technology/developers/google-kaggle-genai-intensive/">Read the story</a></p><h4>Google Gemini’s Email Summarization in Gmail: A Must-Have?</h4><p><em>Published 09 December, 2024</em></p><p>Explore how AI can revolutionize your email management in Gmail, with Google Gemini offering to simplify your inbox and boost productivity.</p><p><a href="https://www.wired.com/story/google-gemini-summarize-emails-in-gmail/">Read the story</a></p><h4>Adobe and AWS Join Forces for Cutting-Edge AI Solutions</h4><p><em>Published 09 December, 2024</em></p><p>Discover Adobe’s major expansion with AWS, focusing on enhancing AI-driven customer experiences and agile management of data with groundbreaking personalization tools.</p><p><a href="https://venturebeat.com/ai/adobe-partners-with-amazon-web-services-to-supercharge-ai-driven-customer-experiences/">Read the story</a></p><p>This week’s selection of articles traverses the dynamic landscape of artificial intelligence, demonstrating its diverse applications and impacts across various fields. As we keep an eye on these developments, we thank you for joining us on this journey. Have a wonderful weekend and stay tuned for next week’s exciting insights!</p><p>BR, The Weekly Buzz</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=404772e5f97c" width="1" height="1" alt=""><hr><p><a href="https://blog.knowcast.ai/ai-unplugged-adventures-from-tokyo-to-tech-frontlines-404772e5f97c">Adventures from Tokyo to Tech Frontlines</a> was originally published in <a href="https://blog.knowcast.ai/">The Knowcast</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[New Developments in Generative Capabilities]]></title>
            <link>https://blog.knowcast.ai/new-developments-in-generative-capabilities-f4db3766511d?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/f4db3766511d</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[large-language-models]]></category>
            <category><![CDATA[generative-ai-use-cases]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Sun, 20 Oct 2024 18:40:33 GMT</pubDate>
            <atom:updated>2024-10-20T18:40:33.961Z</atom:updated>
            <content:encoded><![CDATA[<p>This week, we’re excited to bring you the latest updates on artificial intelligence’s fast-paced progression in generative capabilities. We’ve curated a selection of articles that showcase AI’s potential in transforming industries and improving productivity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5Pf5QJUEnFkDORGxlraWEg.png" /></figure><h4>Filmmakers Look to Big Tech for AI Possibilities,.Meta</h4><p><em>October 17, 2024</em></p><p>Amazon and Meta are leading the charge in enabling filmmakers to unfold new possibilities with AI technology. Join the revolution and discover what’s in store for the filmmaking industry.</p><p><a href="https://www.wired.com.story/filmmakers-are-worried-about-ai-big-tech-wants-them-to-see-whats-possible/">Read the story</a></p><h4>Transforming Software Development with Gen AI</h4><p><em>October 17, 2024</em></p><p>Generative AI is transforming the software development lifecycle, promising increased productivity and efficiency. Dive in to learn more about this game-changing technology.</p><p><a href="https://www.technologyreview.com/2024/10/17/1105295/transforming-software-with-generative-ai/">Read the story</a></p><h4>Meta Boosts Material Discovery with Altruistic AI Data Release</h4><p><em>October 18, 2024</em></p><p>Meta is accelerating the quest to discover new materials with a massive data share. This breakthrough could redefine the scientific landscape, and we’ve got the details.</p><p><a href="https://www.technologyreview.com/2024/10/18/1105880/the-race-to-find-new-materials-with-ai-needs-more-data-meta-is-giving-massive-amounts-away-for-free/">Read the story</a></p><h4>Spirit LM: Meta’s Expressive AI-Powered Tool that Seamlessly Unites Speech and Text</h4><p><em>October 19, 2024</em></p><p>Introducing Spirit LM, a revolutionary AI tool that perfects speech generation by incorporating emotional cues. Get ready to take communication to the next level.</p><p><a href="https://venturebeat.com/ai/meta-introduces-spirit-lm-open-source-model-that-combines-text-and-speech-inputs-outputs/">Read the story</a></p><h4>Take Control of AI-Powered Podcasts with Google’s NotebookLM Update</h4><p><em>October 17, 2024</em></p><p>Google has upped the ante on AI-generated podcasts with NotebookLM. Customizing generated content means you’re only limited by your imagination.</p><p><a href="https://www.wired.com/story/google-notebooklm-customize-ai-podcast/">Read the story</a></p><p>That’s a wrap for this edition of The Weekly Buzz! Thank you for joining us on this fascinating journey through AI’s frontier. Have a fantastic weekend, and we look forward to catching you next week! BR, The Weekly Buzz</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f4db3766511d" width="1" height="1" alt=""><hr><p><a href="https://blog.knowcast.ai/new-developments-in-generative-capabilities-f4db3766511d">New Developments in Generative Capabilities</a> was originally published in <a href="https://blog.knowcast.ai/">The Knowcast</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Nobel Prize, Mental Health and Ethics]]></title>
            <link>https://blog.knowcast.ai/ai-breakthroughs-innovation-and-ethical-concerns-904b27a15501?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/904b27a15501</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Sat, 12 Oct 2024 08:02:09 GMT</pubDate>
            <atom:updated>2024-10-12T08:05:30.134Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="AI innovation and ethics" src="https://cdn-images-1.medium.com/proxy/0*0HxkngWNcdszjcaM" /></figure><h4>Google DeepMind leaders share Nobel Prize in Chemistry for Protein Prediction AI</h4><p><em>09 October, 2024</em></p><p>The Royal Swedish Academy of Sciences has awarded half the 2024 prize in chemistry to Demis Hassabis for their pioneering work on AI-powered protein prediction. This breakthrough has far-reaching implications for a wide range of industries, including healthcare and biotechnology.</p><p><a href="https://www.technologyreview.com/2024/10/09/1105335/google-deepmind-wins-joint-nobel-prize-in-chemistry-for-protein-prediction-ai/">Read the story</a></p><h4>OpenAI’s GPT Store Has Left Some Developers in the Lurch</h4><p><em>11 October, 2024</em></p><p>OpenAI had promised payouts to creators of custom GPT models, but several developers are now being left in dismay as they search for alternative revenue streams. This episode raises concerns about the uncertainties faced by developers relying on AI platforms.</p><p><a href="https://www.wired.com/story/openai-gpt-store/">Read the story</a></p><h4>AI Startups Revolutionizing Mental Health Care</h4><p><em>10 October, 2024</em></p><p>A new wave of AI-powered startups is aiming to improve mental health care, demonstrating the vast potential of AI in this field. Startups are leveraging AI to proactively address mental health issues, mitigating current shortcomings in the care continuum.</p><p><a href="https://blog.google/outreach-initiatives/entrepreneurs/ai-startups-mental-health/">Read the story</a></p><h4>Amazon’s Rufus AI Shopping Assistant Now Lets Some Shoppers Check Price History</h4><p><em>10 October, 2024</em></p><p>The introduction of price history checks for select shoppers promises transparency in Amazon’s retail environment. Rufus, Amazon’s ChatGPT competitor, intends to tackle this core task through seamless integration, enriching userShopping and offering deal hunters valuable insights.</p><p><a href="https://www.wired.com/story/amazon-rufus-price-check/">Read the story</a></p><h4>How Should We Feel About Ring?</h4><p><em>10 October, 2024</em></p><p>The home security camera brand has harnessed AI technologies to better facilitate footage reviewing. This sparks pressing questions — is the integration of AI in security tools more effective security or subtle surveillance harassment?</p><p><a href="https://www.wired.com/story/gadget-lab-podcast-661/">Read the story</a></p><p>This week’s selection of articles showcases significant improvements, interesting intersections of human and artificial intelligence, and ethical complexities that demand discussion. As AI infiltrates our daily lives and our workplaces, these conversations become vital.</p><p>We appreciate your readership and look forward to your reactions! It’s nearing the weekend; fill it with exciting discussions about AI. Better yet, look forward to another dose of insightful selections next week in The Weekly Buzz.</p><p>BR, The Weekly Buzz</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=904b27a15501" width="1" height="1" alt=""><hr><p><a href="https://blog.knowcast.ai/ai-breakthroughs-innovation-and-ethical-concerns-904b27a15501">Nobel Prize, Mental Health and Ethics</a> was originally published in <a href="https://blog.knowcast.ai/">The Knowcast</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Debates, Innovations, and Ethical Dilemmas]]></title>
            <link>https://blog.knowcast.ai/debates-innovations-and-ethical-dilemmas-57a6b892416c?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/57a6b892416c</guid>
            <category><![CDATA[ethics]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Mon, 30 Sep 2024 17:56:50 GMT</pubDate>
            <atom:updated>2024-09-30T17:56:50.159Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pnAOi6GfJy2fjkTw-uNbmw.png" /></figure><p>This week, we explore the transformative power of AI, LLMs, and Generative AI. From debunking myths about AI’s potential to categorizing hateful content, we delve into the various facets of AI’s influence on humanity.</p><h4>No, Sam Altman, AI Won’t Solve All of Humanity’s Problems</h4><p><em>27 September, 2024</em></p><p>Reevaluating the role of AI in human behavior, the OpenAI CEO’s claims on AI’s grandeur are revisited.</p><p><a href="https://www.wired.com/story/plaintext-no-sam-altman-ai-wont-solve-all-of-humanitys-problems/">Read the story</a></p><h4>Want AI that flags hateful content? Build it</h4><p><em>25 September, 2024</em></p><p>Humane Intelligence challenges developers to create a computer vision model that tracks hateful image-based propaganda online.</p><p><a href="https://www.technologyreview.com/2024/09/25/1104519/ai-models-hate-imagery-humane-intelligence-bounty-competition/">Read the story</a></p><h4>A tiny new open-source AI model performs as well as powerful big ones</h4><p><em>25 September, 2024</em></p><p>The Allen Institute for Artificial Intelligence releases a family of open-source multimodal language models that rival top proprietary models.</p><p><a href="https://www.technologyreview.com/2024/09/25/1104465/a-tiny-new-open-source-ai-model-performs-as-well-as-powerful-big-ones/">Read the story</a></p><h4>Onboarding the AI workforce: How digital agents will redefine work itself</h4><p><em>29 September, 2024</em></p><p>By 2030, AI will evolve from a back-office assistant to a front-line collaborator in enterprise. Embracing it will be crucial for success.</p><p><a href="https://venturebeat.com/ai/onboarding-the-ai-workforce-how-digital-agents-will-redefine-work-itself/">Read the story</a></p><h4>Why AI is a know-it-all know nothing</h4><p><em>28 September, 2024</em></p><p>Today’s AI systems lack reasoning behind their outputs, rendering them untrustworthy.</p><p><a href="https://venturebeat.com/ai/why-ai-is-a-know-it-all-know-nothing/">Read the story</a></p><h4>China’s Plan to Make AI Watermarks Happen</h4><p><em>27 September, 2024</em></p><p>China introduces regulations to define AI content labeling and combat AI-generated disinformation.</p><p><a href="https://www.wired.com/story/china-wants-to-make-ai-watermarks-happen/">Read the story</a></p><h4>185 real-world gen AI use cases from the world’s leading organizations</h4><p><em>24 September, 2024</em></p><p>Google Cloud highlights the diverse applications of generative AI across industries.</p><p><a href="https://blog.google/products/google-cloud/gen-ai-business-use-cases/">Read the story</a></p><h4>An ‘iPhone of AI’ Makes No Sense. What Is Jony Ive Really Building?</h4><p><em>28 September, 2024</em></p><p>LoveFrom partners with OpenAI to develop AI devices that are less socially disruptive than the iPhone.</p><p><a href="https://www.wired.com/story/jony-ive-iphone-of-ai/">Read the story</a></p><p>So let’s see what this week has in store for us in the AI field. Stay tuned!</p><p>BR, The Weekly Buzz</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=57a6b892416c" width="1" height="1" alt=""><hr><p><a href="https://blog.knowcast.ai/debates-innovations-and-ethical-dilemmas-57a6b892416c">Debates, Innovations, and Ethical Dilemmas</a> was originally published in <a href="https://blog.knowcast.ai/">The Knowcast</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Voice Clones to World Builders]]></title>
            <link>https://blog.knowcast.ai/ai-adventures-from-voice-clones-to-world-builders-bbdf61a22070?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/bbdf61a22070</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Sun, 15 Sep 2024 21:50:07 GMT</pubDate>
            <atom:updated>2024-09-17T12:08:11.937Z</atom:updated>
            <content:encoded><![CDATA[<h3>From Voice Clones to World Builders and Fact-Checking Models</h3><figure><img alt="Exploring the latest in AI advancements" src="https://cdn-images-1.medium.com/proxy/0*XXDOqti2x08yscqn" /></figure><p>Enough reading for today? <a href="https://drive.google.com/file/d/1IxcGr1319cH_c5SsyqNLjJ9qtjxEy9lk/view?usp=drive_link">Listen to the podcast</a> instead.</p><p>This week’s newsletter dives into groundbreaking advancements in AI technology. From Amazon’s new feature allowing audiobook narrators to create AI voice clones to OpenAI’s latest model solving complex problems, the AI landscape is rapidly evolving. Stanford’s Fei-Fei Li introduces a startup aimed at teaching AI systems about physical reality, while Google tackles fact-checking in large language models. We also explore how AI is being used to combat conspiracy theories, enhance study tools, and its impact on the fast fashion industry.</p><h4>Amazon’s Audiobook Narrators Can Now Make Their Own AI Voice Clones</h4><p><em>14 September, 2024</em></p><p>Amazon has unveiled a new feature allowing audiobook narrators to create their own AI voice clones. This development is part of a broader trend, including Meta’s AI activities in the UK, Google’s passport digitization, Chrome’s tab synchronization, and Apple’s new iPhone announcements.</p><p><a href="https://www.wired.com/story/audible-audiobook-narrators-ai-voice-clones/">Read the story</a></p><h4>OpenAI’s New Model ‘Strawberry’ Solves Complex Problems Step by Step</h4><p><em>12 September, 2024</em></p><p>OpenAI introduces ‘Strawberry,’ an AI model that demonstrates the need for more than just scale to advance AI capabilities. This model excels in solving difficult problems through a step-by-step approach, showcasing a new dimension in AI problem-solving.</p><p><a href="https://www.wired.com/story/openai-o1-strawberry-problem-reasoning/">Read the story</a></p><h4>The Visionary World of AI: Fei-Fei Li’s New Startup</h4><p><em>13 September, 2024</em></p><p>Stanford computer scientist Fei-Fei Li is launching a startup focused on teaching AI systems deep knowledge of physical reality. This innovative venture is attracting significant investment and aims to revolutionize how AI interacts with the real world.</p><p><a href="https://www.wired.com/story/plaintext-the-godmother-of-ai-wants-everyone-to-be-a-world-builder/">Read the story</a></p><h4>Google Introduces Fact-Checking Tool for AI Responses</h4><p><em>12 September, 2024</em></p><p>Google has released a new tool designed to enable large language models to fact-check their responses. This tool addresses the issue of AI “hallucinations,” making AI-generated information more reliable and trustworthy.</p><p><a href="https://www.technologyreview.com/2024/09/12/1103926/googles-new-tool-lets-large-language-models-fact-check-their-responses/">Read the story</a></p><h4>AI Chatbots: A New Weapon Against Conspiracy Theories</h4><p><em>12 September, 2024</em></p><p>Researchers from MIT Sloan and Cornell University have discovered that AI chatbots can effectively persuade people to stop believing in conspiracy theories. This breakthrough offers a new approach to combating the spread of harmful misinformation online.</p><p><a href="https://www.technologyreview.com/2024/09/12/1103930/chatbots-can-persuade-people-to-stop-believing-in-conspiracy-theories/">Read the story</a></p><h4>Listen to Conversations About Your Sources with NotebookLM</h4><p><em>11 September, 2024</em></p><p>Google’s NotebookLM now features an audio player that allows users to listen to conversations about their sources, such as Google Slides and PDFs. This new functionality aims to enhance the way users engage with and understand their materials.</p><p><a href="https://blog.google/technology/ai/notebooklm-audio-overviews/">Read the story</a></p><h4>DataGemma: Tackling AI Hallucinations with Real-World Data</h4><p><em>12 September, 2024</em></p><p>Google’s DataGemma initiative uses real-world data to address the issue of AI hallucinations. By grounding AI models in reality, this project aims to improve the accuracy and reliability of AI-generated content.</p><p><a href="https://blog.google/technology/ai/google-datagemma-ai-llm/">Read the story</a></p><h4>AI’s Role in Shein’s Environmental Impact</h4><p><em>14 September, 2024</em></p><p>Shein, the fast fashion giant, has become the industry’s largest polluter, nearly doubling its emissions in 2023. AI has played a significant role in this increase, exacerbating the environmental impact of an already unsustainable industry.</p><p><a href="https://www.wired.com/story/shein-is-officially-the-biggest-polluter-in-fast-fashion-ai-is-making-things-worse/">Read the story</a></p><p>In summary, this week’s articles highlight the rapid advancements and diverse applications of AI technology. From enhancing digital interactions to addressing significant societal issues, AI continues to shape our world in profound ways. Thank you for reading The Weekly Buzz. Have a great weekend, and we look forward to sharing more exciting updates with you next week.</p><p>BR, The Weekly Buzz</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bbdf61a22070" width="1" height="1" alt=""><hr><p><a href="https://blog.knowcast.ai/ai-adventures-from-voice-clones-to-world-builders-bbdf61a22070">From Voice Clones to World Builders</a> was originally published in <a href="https://blog.knowcast.ai/">The Knowcast</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Berries, Mysteries, and AI: A Juicy Week in Tech!]]></title>
            <link>https://blog.knowcast.ai/ai-breakthroughs-and-innovations-this-week-in-tech-759ae8104fad?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/759ae8104fad</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Fri, 13 Sep 2024 13:53:09 GMT</pubDate>
            <atom:updated>2024-09-17T12:17:23.880Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lDOyzduYEeNZ627qgKiBug.jpeg" /></figure><p>This week, we saw significant advancements in AI technology, from OpenAI’s new problem-solving model to Apple’s integration of AI features in the new iPhone 16. We also explored the role of AI in solving mysteries and the importance of addressing AI hallucinations. Catch up on the latest developments in the world of tech!</p><h4>Strawberry: The AI Model That Solves Difficult Problems Step by Step</h4><p><em>12 September, 2024</em></p><p>The ChatGPT maker reveals details of what’s officially known as OpenAI o1, which shows that AI needs more than scale to advance.</p><p><a href="https://www.wired.com/story/openai-o1-strawberry-problem-reasoning/">Read the story</a></p><h4>The AI Twist in Reddit’s Celebrity Number Six Mystery</h4><p><em>13 September, 2024</em></p><p>This week, Reddit solved the long-running Celebrity Number Six mystery, but it almost got derailed thanks to accusations that the missing piece was made with AI.</p><p><a href="https://www.wired.com/story/reddit-celebrity-number-six-ai/">Read the story</a></p><h4>The Future of AI: What to Expect from Apple’s iPhone 16</h4><p><em>12 September, 2024</em></p><p>This week, we talk about the new iPhone 16 and its many AI tricks. Some of them are impressive, some of them are meh, all of them are very Apple.</p><p><a href="https://www.wired.com/story/gadget-lab-podcast-658/">Read the story</a></p><h4>DataGemma: A Step Towards Addressing AI Hallucinations</h4><p><em>12 September, 2024</em></p><p>DataGemma is a new initiative that uses real-world data to address AI hallucinations, a crucial step towards developing more accurate and reliable AI models.</p><p><a href="https://blog.google/technology/ai/google-datagemma-ai-llm/">Read the story</a></p><p>Thanks for reading this week’s Buzz Weekly! We hope you enjoyed the latest updates from the world of tech. Have a great weekend, and we’ll see you next week!</p><p>BR, The Buzz News Crew</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=759ae8104fad" width="1" height="1" alt=""><hr><p><a href="https://blog.knowcast.ai/ai-breakthroughs-and-innovations-this-week-in-tech-759ae8104fad">Berries, Mysteries, and AI: A Juicy Week in Tech!</a> was originally published in <a href="https://blog.knowcast.ai/">The Knowcast</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Breakthroughs and Blunders]]></title>
            <link>https://blog.knowcast.ai/ai-breakthroughs-and-blunders-this-week-in-buzz-weekly-fe753e1441f0?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/fe753e1441f0</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Tue, 20 Aug 2024 11:57:25 GMT</pubDate>
            <atom:updated>2024-11-01T11:21:50.823Z</atom:updated>
            <content:encoded><![CDATA[<p>This week we see both remarkable advancements and surprising setbacks. From AI models that can detect diseases based on coughs to AI-generated images that can deceive even the most discerning eyes, we’ve got it all covered.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dlIMm3R8zMVK5chc2PUWFA.png" /></figure><h4>Why AI Sometimes Gets Wrong — and Big Strides to Address It</h4><p><em>16 July, 2024</em></p><p>The post explores the limitations of AI and the significant progress being made to overcome them.</p><p><a href="https://news.microsoft.com/source/features/ai/why-ai-sometimes-gets-it-wrong-and-big-strides-to-address-it/">Read the story</a></p><h4>Trump Shares AI-Generated Images Claiming Swifties Are Supporting Him</h4><p><em>19 August, 2024</em></p><p>A surprising example of AI-generated images being used to spread misinformation.</p><p><a href="https://www.wired.com/story/trump-shares-ai-generated-taylor-swift/">Read the story</a></p><h4>This AI Model is Helping Researchers Detect Disease Based on Coughs</h4><p><em>19 August, 2024</em></p><p>A remarkable breakthrough in AI-assisted disease detection.</p><p><a href="https://blog.google/technology/health/ai-model-cough-disease-detection/">Read the story</a></p><h4>Researchers Have Ranked AI Models Based on Risk — and Found a Wild Range</h4><p><em>15 August, 2024</em></p><p>A crucial study that highlights the need for regulations to prevent AI misbehavior.</p><p><a href="https://www.wired.com/story/ai-models-risk-rank-studies/">Read the story</a></p><h4>New Tools for Innkeepers, Weavers, and Florists: How AI is Reaching the Farthest Edges of Ireland</h4><p><em>17 June, 2024</em></p><p>A heartwarming story of how AI is empowering small businesses in rural Ireland.</p><p><a href="https://news.microsoft.com/source/emea/features/leaving-no-one-behind-ireland-ai-tools-innkeepers-weavers-florists/">Read the story</a></p><p>Thanks for reading this week’s Buzz Weekly! We hope you enjoyed the fascinating stories about AI. Have a great weekend, and we’ll see you next week!</p><p>BR, The Buzz News Crew</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fe753e1441f0" width="1" height="1" alt=""><hr><p><a href="https://blog.knowcast.ai/ai-breakthroughs-and-blunders-this-week-in-buzz-weekly-fe753e1441f0">AI Breakthroughs and Blunders</a> was originally published in <a href="https://blog.knowcast.ai/">The Knowcast</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Breakthroughs and Innovations: This Week in Tech]]></title>
            <link>https://blog.knowcast.ai/ai-breakthroughs-and-innovations-this-week-in-tech-486f9351616e?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/486f9351616e</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[large-language-models]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Sat, 03 Aug 2024 19:15:32 GMT</pubDate>
            <atom:updated>2024-09-17T12:17:53.733Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QomIqyW97Z2c6bhYKhqJ4Q.png" /></figure><p>This week, we saw significant advancements in AI technology, from tamperproofing open source language models to building AI supercomputers. We also explored the importance of a silicon revolution in the US and how AI can reduce traffic congestion and emissions.</p><h4>A New Trick to Block the Misuse of Open Source AI</h4><p><em>02 August, 2024</em></p><p>Researchers have developed a way to prevent open source large language models from being coaxed into harmful activities, ensuring a safer AI future.</p><p><a href="https://www.wired.com/story/center-for-ai-safety-open-source-llm-safeguards/">Read the story</a></p><h4>Tesla Dojo: Elon Musk’s Big Plan to Build an AI Supercomputer, Explained</h4><p><em>03 August, 2024</em></p><p>Dojo, Tesla’s custom-built supercomputer, is designed to train its “Full Self-Driving” neural networks, pushing the boundaries of AI capabilities.</p><p><a href="https://techcrunch.com/2024/08/03/tesla-dojo-elon-musks-big-plan-to-build-an-ai-supercomputer-explained/">Read the story</a></p><h4>To Lead in AI, the US Needs a Silicon Revolution</h4><p><em>01 August, 2024</em></p><p>US Commerce Department undersecretary Laurie E. Locascio emphasizes the need for new chip manufacturing techniques to maintain America’s lead in AI innovation.</p><p><a href="https://www.wired.com/story/laurie-locascio-interview-ai-silicon-revolution/">Read the story</a></p><h4>Intel Is Cutting More Than 15,000 Jobs Despite Getting Billions From the US Government</h4><p><em>01 August, 2024</em></p><p>Despite receiving significant government funding, Intel is slashing 15 percent of its workforce, highlighting the challenges faced by the chipmaker.</p><p><a href="https://www.wired.com/story/intel-job-cuts/">Read the story</a></p><h4>How Google uses AI to Reduce Stop-and-Go Traffic on Your Route — and Fight Fuel Emissions</h4><p><em>29 July, 2024</em></p><p>Google’s AI-powered project, Greenlight, aims to reduce traffic congestion and emissions by optimizing traffic light timings.</p><p><a href="https://blog.google/outreach-initiatives/sustainability/google-ai-project-greenlight/">Read the story</a></p><p>Thank you for reading this week’s Buzz Weekly! We hope you enjoyed the latest AI breakthroughs and innovations. Have a great weekend, and we’ll see you next week!</p><p>BR, The Buzz News Crew</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=486f9351616e" width="1" height="1" alt=""><hr><p><a href="https://blog.knowcast.ai/ai-breakthroughs-and-innovations-this-week-in-tech-486f9351616e">AI Breakthroughs and Innovations: This Week in Tech</a> was originally published in <a href="https://blog.knowcast.ai/">The Knowcast</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Automating a Newsroom Crew with AI Agents and CrewAI]]></title>
            <link>https://medium.com/@baktakt/automating-a-newsroom-crew-with-ai-agents-and-crewai-0f102b4cf6a6?source=rss-53dc4300016d------2</link>
            <guid isPermaLink="false">https://medium.com/p/0f102b4cf6a6</guid>
            <category><![CDATA[crew-ai]]></category>
            <category><![CDATA[digital]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Johan Uddståhl]]></dc:creator>
            <pubDate>Sat, 03 Aug 2024 17:12:07 GMT</pubDate>
            <atom:updated>2024-11-01T12:14:29.876Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OEzYQjgCzqXYrUsc1vmqCg.png" /></figure><p>This is the first post in a series around AI Agents, automation and Large Language Models.</p><p>The code for the crew is available in <a href="https://github.com/baktakt/buzz_weekly_crew/tree/main">github</a> and will be continuously updated. Contributions and pull requests are appreciated.</p><h3>No Time to Waste</h3><p>Ah yes, vacation time in Sweden. Time to switch off, relax, and enjoy lazy days in the sun. Read a book on a towel at the beach. A non-business book that doesn’t aim to make you a better programmer, business leader, or human being. Just a regular fictional book….</p><p>Who am I kidding? As a tech nerd, I finally have time to dive into those interesting topics, blog posts, YouTube videos, and news articles that flashed by in my online feeds while I was busy with PowerPoints and Teams meetings.</p><p>This year started with client discussions on Robotic Process Automation (RPA), which involves automating repetitive business tasks with computer programs. I explored a few frameworks, but most were enterprise-grade, complex, expensive, and felt outdated, like ERP systems from the ’90s.</p><p>I needed something more lightweight, fun, and preferably open-source. I had a specific problem to solve with automation. Reading a few blog posts, a new concept emerged: <strong>AI Agents</strong>.</p><p>Before diving into the problem I wanted to solve and my approach, let’s take a step back.</p><h3>AI Agents, You Say?</h3><p>In the context of AI Agents, an agent is defined as an autonomous unit programmed to:</p><ul><li>Perform tasks</li><li>Make decisions</li><li>Communicate with other agents</li></ul><p>Think of an agent as a team member performing a specific task. For instance, a newspaper crew might include:</p><ul><li><strong>A journalist</strong> finding and writing stories</li><li><strong>• A photographer</strong> creating images</li><li><strong>• An editor</strong> selecting and editing stories</li><li><strong>• A social media expert</strong> pushing content and driving traffic</li></ul><p>The team operates like a real news desk, tasked with creating, say, five interesting tech stories for the front page. The team moves from agent to agent, performing necessary tasks to deliver the final output.</p><p><strong>Sounds Like Good Ol’ </strong><a href="https://en.wikipedia.org/wiki/Robotic_process_automation"><strong>RPA</strong></a><strong>, Johan?</strong></p><p>Indeed. The difference is <a href="https://en.wikipedia.org/wiki/Large_language_model">Large Language Models</a>. Since the introduction of OpenAI’s ChatGPT in 2022, LLMs have shown remarkable abilities in understanding and generating human-like text, making decisions, utilizing tools, writing and executing code, and maintaining context — capabilities previously unattainable with predefined RPA actions.</p><p>The real power lies in combining these powerful LLMs and assigning tasks to the most suitable models. Imagine assembling a news desk crew where each agent possesses more knowledge than an entire traditional newspaper organization.</p><p>You see where I’m going? You’ve not just created a news desk team; you’ve assembled <strong>the world’s most exceptional news desk crew,</strong> with each agent possessing more knowledge than an entire traditional newspaper organization could ever dream of.</p><h3>Problem statement</h3><p>So what sparked my interest in this technology? Well, it’s all in the start of this article. Working 100% as a partner at digital service studio <a href="https://www.humblebee.se/">Humblebee</a>, while also driving a tech startup, Modly Studios, leaves very little room to read up on all the latest news and development within Artificial Intelligence, an area that I love and cannot get enough of. I needed a way of getting information served to me. Quality content, served in a quick readable format and filtered to my specific interest and needs. If only I had a news desk crew available to me 24/7, watching the internet for those nuggets of information I need. If only….</p><blockquote>The AI Agent Crew should provide a solution for professionals with limited time, to stay updated on the latest developments in Artificial Intelligence. This solution should deliver high-quality, concise, and filtered AI news tailored to the user’s specific interests and needs. The AI Agent Crew should function as a 24/7 news desk, continuously monitoring the internet and curating relevant information, ensuring users receive timely and pertinent updates without the need for extensive personal research.</blockquote><h3>CrewAI — an AI Agent Framework</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jJ0WF3odgbPD33Oz.png" /></figure><p>I think it was the name that caught my eye at first. It’s a fun creative name and sparked the idea of creating my own crew. So I started digging into the documentation, firing up example code and fiddling around with the code.</p><p><a href="https://docs.crewai.com/">CrewAI’s docs</a> are easy to navigate and read, so I really recommend a visit if you want to dig deeper, but here are the core concepts of how it works.</p><p><strong>Agents</strong></p><p>We covered the basics of Agents above and in rewAI an agent can be defined in either code or yaml. In my code examples I have consistently used yaml to define agents and tasks.</p><p><strong>Tasks</strong></p><p>In the crewAI framework, tasks are specific assignments completed by agents, including details such as descriptions, responsible agents, required tools, and more, accommodating various action complexities.</p><p>Tasks can be collaborative, involving multiple agents working together. This collaboration is managed through task properties and orchestrated by the Crew’s process, enhancing teamwork and efficiency.</p><pre>research_ai_trends_task:<br>  description: &gt;<br>    Conduct thorough research to identify the latest trends and breakthroughs <br>    in AI technology.<br>    <br>    Utilize various online resources, including research papers, tech blogs, <br>    and news articles, to gather a comprehensive overview.<br>  expected_output: &gt;<br>    A detailed report summarizing the top 5 current trends in AI, <br>    including key players, emerging technologies, and potential impacts on the industry.</pre><p><strong>Tools</strong></p><p>A tool in CrewAI is a skill or function that agents can utilize to perform various actions. This includes tools from the <a href="https://github.com/joaomdmoura/crewai-tools">crewAI Toolkit</a> and <a href="https://python.langchain.com/docs/integrations/tools">LangChain Tools</a>, enabling everything from simple searches to complex interactions and effective teamwork among agents.</p><p>For solving my problems I needed to build a few specific tools of my own:</p><ul><li>A tool to extract the content of a list of articles found in a specific RSS feed</li></ul><pre>@tool<br>    def get_articles_from_feeds(rss_feeds_str: str, take: int = 3) -&gt; str:</pre><ul><li>Tool to create a draft post in my medium account</li></ul><pre>@tool<br>    def create_medium_draft_post(title: str, content: str, tags: list=None):</pre><p>With the core building blocks crewAI provides an intuitive and fun way of defining your crew and the tasks you want it to perform. Under the hood, the framework then helps with context, delegation, logging, error handling and execution in a very neat way.</p><p>One of the funniest thing with exceuting a crew is turning on verbose logging, and watching when the agents reason among themselves, handing over responsibility and completing tasks. The most entertaining runs are those that completely spirals out of control and you end up with unexpected results.</p><h3>Models, models, models</h3><p>A thing that has really caught my interest is the fierce race for building performant LLM models. In one ring corner, the closed source models from OpenAI, <a href="https://openai.com/index/hello-gpt-4o/">GPT-4o</a> etc. On the other side, the vast amount of open source models like <a href="https://llama.meta.com/">Facebooks Llama3</a>, <a href="https://blog.google/technology/developers/google-gemma-2/">Googles Gemma 2</a> and <a href="https://mistral.ai/technology/#models">Mistral</a>.</p><p>The fact that the tech giants are on both sides of this pushes the development with a speed I never thought would be possible, and it seems new models are released every day (hence the problem statement above).</p><p>For this solution, I wanted to explore the possibilities with open source models. Both for the possibility to see what is in them and to be able to tweak models to my specific use case, but also because they offer smaller models available for running on my local computer. This would enable my crew to live entirely inside my computer, executing tasks while I am in Teams meetings and making Powerpoints. This thought is oddly satisfying to me.</p><h3>Hack away</h3><p>So, when understanding the core concepts I started hacking away with finding the perfect crew for my problem statement. I decided that the platform for publishing news would be Medium. This is where the articles found by the crew would end up. I also knew early on that I wanted to select the specific feeds that the crew should choose it’s articles from, so I defined this as an input parameter to the crew.</p><p>For some reason, probably me being a developer at heart, I wanted the crew to pass json documents between each other.</p><p>It also felt comforting to not let the models decide entirely what sources to use. Therefore I added an input parameter where I could define a list of rss feeds that the crew could choose content from. This is quite a big boundary that probably limits the models more than necessary, and I might remove that boundary when I’ve found the perfect models.</p><p>After a lot of token spending and trial and error, I finally started getting post drafts up on Medium that looked good, with links and descriptions correctly written. I created a new publication to host the articles and used ChatGPT to find a catchy name. Still a lot of things to tweak, I need a better logo, some intro image and text. But nevertheless, I give you “<a href="https://blog.knowcast.ai/">Knowcast</a>” — Updates from a World in Change</p><p><a href="https://blog.knowcast.ai/">The Knowcast</a></p><h3>Learnings so far</h3><p>As I move my way forward I like to document some of the key learnings. Here are a few.</p><p><strong>Models matter</strong></p><p>Trying out the smaller models, like Llama 3 8B-models turned out to be entertaining. But it never created the results I was after. Specifically the model had problems with executing the tools available to the agents. Switching to a larger model like GPT-4o or LLama 3 70B worked a lot better.</p><p><strong>Pricing</strong></p><p>OpenAI’s API is great. But it can also be expensive, when using many tokens. And for personal use this can seem a bit scary, since you can’t really control the token usage the agents will need to complete the tasks. GPT-4o performed really well, though. So if you’re using the company card, just go my friend.</p><p>I decided to try out <a href="http://groq.com/">Groq</a>, and their API Service <a href="https://console.groq.com/login">groqcloud</a>, with the Llama3 70B model. This has so far worked out really great and my crew is delivering good results 9/10 times.</p><p><strong>Types</strong></p><p>I wrote my code in Python, which is quite new to me, so I forgot to define the type for the input parameters for one of the tools, which gave really bad results. So pro-tip: Don’t forget to define the types.</p><pre>def get_article_links(from_url: str, take: int) -&gt; list:</pre><p><strong>Utilise AI when coding</strong></p><p>ChatGPT has some great GPTs to utilise for help with crewAI. I used <a href="https://chatgpt.com/g/g-qqTuUWsBY-crewai-assistant">CrewAI Assistant</a> when needing help on agent definitions, coding tools etc.</p><h3>What’s next</h3><p>My crew now produces a decent output when run, but there are things to improve:</p><ul><li><strong>Post layout</strong>. I want to tweak the prompts to make the formatter agent and the formatter_task create a better layout.</li><li><strong>Images</strong>. I would love to automate inspiring images that could make the post more interesting. Preferably with Midjourney.</li><li><strong>Input feeds</strong>. Tweak the input links to provide more specific news articles.</li><li><strong>Social media</strong>. I will create an instagram account and see if I can add a social media agent to the mix, creating content to push the articles with images and texts on Instagram and X.</li><li><strong>Medium publication</strong>. I want to improve the publication logo, images and description. Help is appreciated.</li><li><strong>Multi-model agents.</strong> I want to try to use different models for different agents. Maybe one model can be fine tuned to produce the best post layout, one is better at choosing the right articles and one for proof reading etc. This would also allow me to dig deeper into fine tuning open source models, another topic I need to get into.</li></ul><p>The code for my specific crew is available in <a href="https://github.com/baktakt/buzz_weekly_crew/tree/main">github</a>, and I welcome contributions to making this the best news crew in the world.</p><p>As crewAI turned out to be such an interesting framework, I reached out to the creators and signed up for their SaaS solution for running crews, <a href="https://www.crewai.com/crewaiplus">crewAI+</a>. I am testing it out as we speak, and I will cover some details in my next post about AI Agents.</p><p>As I continue to develop my crew, I will also post new learnings and ideas about coming iterations. Join the ride by following me here and on <a href="https://www.linkedin.com/in/johanuddstahl/">LinkedIn</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0f102b4cf6a6" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>