Pleasingly, CSX has a superior ROE than the average (14%) in the Transportation industry. That's what we like to see. With that said, a high ROE doesn't always indicate high profitability. A higher proportion of debt in a company's capital structure may ...

Should You Be Excited About CSX Corporation's (NASDAQ:CSX) 32% Return On Equity?

Canadian National Railway, Canadian Pacific Kansas City and CSX are part of the Zacks Industry Outlook article.

Zacks Industry Outlook Highlights Canadian National Railway, Canadian Pacific Kansas City and CSX

CSX Corp. (NASDAQ: CSX) has a higher price-to-earnings ratio of 15.52x compared to its average ratio, and the 36-month beta value for CSX is at 1.20. Analysts have varying views on the stock, with 16 analysts rating it as a “buy,

Should You Invest in CSX Corp. (CSX) Now?

CSX reported the ratification of a paid sick leave agreement with the Brotherhood of Railroad Signalman (BRS) that covers approximately 200 employees working on its Louisville and Nashville (L and N) property.

CSX and Brotherhood of Railroad Signalman sign paid sick leave agreement

Traffic congestion — including the state's plan to widen U.S. 15 though the city of Frederick — is expected to be a main topic when local officials meet with representatives from Maryland's Department of Transportation next week.

Meeting with county, state officials to look at county's transportation needs

WASHINGTON – Canadian Pacific Kansas City and CSX Transportation told the Surface Transportation Board today that there's no reason to grant Norfolk Southern's request for a delay in review of their proposals to create a new through route linking the ...

CPKC and CSX urge regulators to reject NS request for pause in review of Meridian & Bigbee transactions

WASHINGTON – Canadian Pacific Kansas City and CSX Transportation told the Surface Transportation Board today that there’s no reason to grant Norfolk Southern’s request for a delay in review of their proposals to create a new through route linking the Southeast and Mexico via Myrtlewood,

The CSX railroad crossing on Ulster Avenue (state Route 212) will close for repairs on Monday, Nov. 6, town police said Friday. The “anticipated timeframe” for the

CSX to close crossing on state Route 212 in Saugerties Nov. 6 through 8, police say

Norfolk Southern is asking the Surface Transportation Board to compel Canadian Pacific Kansas City and CSX to consolidate their applications on plans to create a Mexico-Southeast U.S. corridor. Doing so would provide greater insight on freight and passenger rail impacts,

Norfolk Southern wants close look at proposed CPKC-CSX Southeast corridor

The Virginia Railway Express plans to launch Saturday train service next year, which would be its first weekend operation in the system’s 31-year history.

VRE plans new weekend service, fare increases next year

SAN FRANSCISCO, United States, September 9th, 2024/Chainwire/--Innovative startup from Ethereum Core research, Google Brain, Stanford, MIT, and CMU aims to democratize on-device intelligence, empowering users with control over their personal AI, and receiving investment from a16z CSX, Hack VC, and investors from projects like Solana, Polygon, Near, Worldcoin, etc. PIN AI, a pioneering AI infrastructure company, announces $10 million in pre-seed funding to develop the world’s first open-source Personal Intelligence Network (PIN). A16z CSX, Hack VC, and notable investors, including Blockchain Builders Fund (Stanford Blockchain Accelerator), Illia Polosukhin (Transformer paper author; Founder, NEAR Protocol), Anagram/Lily Liu (President, SOL Foundation), Symbolic Capital (Co-Founder, Polygon), Evan Cheng (CEO, Mysten Labs/SUI), dcbuilder (Worldcoin Foundation), Foresight Ventures (parent company of the Block), Nomad Capital, Tim Shi (Co-Founder, Cresta), Ben Fisch (CEO, Espresso), Scott Moore (Co-Founder, Gitcoin), Alumni Ventures, and Dispersion Capital, have backed the project. Offering an open-source, web3-enabled alternative to Apple Intelligence, PIN AI's platform turns smartphones into privacy-focused AI personal assistants. By redirecting profits from users' data and attention, it empowers users to regain control and monetize their data. The platform leverages personal, contextual data and cryptography, deploying cutting-edge AI models on-device to handle tasks across apps—like shopping, ordering food, wealth management, and interacting with centralized exchanges, DeFi, and prediction markets. This AI-driven assistant ensures robust privacy and user-controlled data management, disrupting ultra-profitable models of web2 giants, including Apple's 30% app revenue cut (over $100 billion) and Google’s mobile ads and Android store revenue cuts. Davide Crapis, Co-Founder leading Protocol Research, said, "We’re building a movement towards an open-source future where personal AI assistants can work on PIN AI’s platform like smart contracts on Ethereum." He added, "PIN AI will take back the $100b+ profit from tech giants and return it to users, allowing them to control and monetize their data. Our platform offers access to a wider range of AI agents, developed by the open-source community, capable of handling tasks across popular apps." PIN AI's mission is to foster innovation for personal AI agents by offering access to personal, contextual data that reflects individual users' needs and preferences. Unlike closed ecosystems like Apple, PIN AI’s open platform connects privacy-protected user data via a Layer-2 blockchain. This enables more flexibility in AI application development without the constraints of traditional, closed systems. Bill Sun, Co-Founder and Chief Scientist, said, "On-device multi-modality models will revolutionize daily life. We are building a personal index for each user to create an on-device model that evolves via distributed training on the user's phone. Soon, users won’t need to open multiple apps to complete tasks. The personal AI assistant will understand preferences and manage tasks efficiently." PIN AI's assistant will connect users with apps and services bidding on its blockchain protocol to fulfill tasks like shopping or crypto activities. PIN AI shifts data monetization from big tech to users, allowing secure monetization of personal data. Users receive token incentives through data onboarding and intent fulfillment, only providing necessary data with matched Personal AI Agents, while maintaining control of their information. At launch, PIN AI partners with Worldcoin and is developing a front-end product similar to Siri, expanding its reach and enhancing user experience. Leading the team are co-founders Davide Crapis and Ben Wu. Crapis, formerly of Ethereum Core Research, leads Protocol Research. Ben Wu, heading Strategy, is an MIT graduate, Y Combinator alum, and serial entrepreneur. The technical leadership includes Bill Sun, a Stanford AI/Math PhD and early Google Brain researcher, as Chief AI Scientist, and Regan Peng, a CMU graduate and former lead at Didi Fintech and Yahoo Data Infra, as Founding Head of Engineering. PIN AI collaborates with a16z crypto research, Flashbots, Espresso Systems, and academics from Stanford, Columbia, and NYU. Ben Wu emphasized the importance of an open platform, saying, “The open internet has made large language models possible. We need to build an open platform for users on their trusted device, allowing access to their various data, making their Personal AI possible." Unlike Apple Intelligence, PIN AI's system can run on low-spec smartphones by dynamically shifting between edge AI (on-device) and server AI to optimize performance, ensuring wide accessibility. The funding will expand research, grow the team of AI and blockchain experts, and accelerate the deployment of PIN AI’s technology. The company will join the a16z CSX Fall 2024 cohort in New York City. About PIN AI PIN AI is developing an open-source personal AI operating system that leverages smartphone and app data to empower users with control and privacy. The company's approach aims to connect AI developers with users, promoting data sovereignty and developer empowerment. PIN AI collaborates with Ethereum Core Research and is committed to privacy and innovation. For more updates, visit pinai.io, X (formerly Twitter), Telegram, Discord, and Linkedin. Contact PIN AI contact@pinai.io This story was distributed as a release by Chainwire under HackerNoon’s Business Blogging Program. Learn more about the program here. SAN FRANSCISCO, United States, September 9th, 2024/Chainwire/--Innovative startup from Ethereum Core research, Google Brain, Stanford, MIT, and CMU aims to democratize on-device intelligence, empowering users with control over their personal AI, and receiving investment from a16z CSX, Hack VC, and investors from projects like Solana, Polygon, Near, Worldcoin, etc. PIN AI , a pioneering AI infrastructure company, announces $10 million in pre-seed funding to develop the world’s first open-source Personal Intelligence Network (PIN). A16z CSX, Hack VC, and notable investors, including Blockchain Builders Fund (Stanford Blockchain Accelerator), Illia Polosukhin (Transformer paper author; Founder, NEAR Protocol), Anagram/Lily Liu (President, SOL Foundation), Symbolic Capital (Co-Founder, Polygon), Evan Cheng (CEO, Mysten Labs/SUI), dcbuilder (Worldcoin Foundation), Foresight Ventures (parent company of the Block), Nomad Capital, Tim Shi (Co-Founder, Cresta), Ben Fisch (CEO, Espresso), Scott Moore (Co-Founder, Gitcoin), Alumni Ventures, and Dispersion Capital, have backed the project. PIN AI PIN AI Offering an open-source, web3-enabled alternative to Apple Intelligence, PIN AI's platform turns smartphones into privacy-focused AI personal assistants. By redirecting profits from users' data and attention, it empowers users to regain control and monetize their data. The platform leverages personal, contextual data and cryptography, deploying cutting-edge AI models on-device to handle tasks across apps—like shopping, ordering food, wealth management, and interacting with centralized exchanges, DeFi, and prediction markets. This AI-driven assistant ensures robust privacy and user-controlled data management, disrupting ultra-profitable models of web2 giants, including Apple's 30% app revenue cut (over $100 billion) and Google’s mobile ads and Android store revenue cuts. Davide Crapis, Co-Founder leading Protocol Research, said, "We’re building a movement towards an open-source future where personal AI assistants can work on PIN AI’s platform like smart contracts on Ethereum." Davide Crapis, Co-Founder leading Protocol Research, said, "We’re building a movement towards an open-source future where personal AI assistants can work on PIN AI’s platform like smart contracts on Ethereum." He added, "PIN AI will take back the $100b+ profit from tech giants and return it to users, allowing them to control and monetize their data. Our platform offers access to a wider range of AI agents, developed by the open-source community, capable of handling tasks across popular apps." He added, "PIN AI will take back the $100b+ profit from tech giants and return it to users, allowing them to control and monetize their data. Our platform offers access to a wider range of AI agents, developed by the open-source community, capable of handling tasks across popular apps." PIN AI's mission is to foster innovation for personal AI agents by offering access to personal, contextual data that reflects individual users' needs and preferences. Unlike closed ecosystems like Apple, PIN AI’s open platform connects privacy-protected user data via a Layer-2 blockchain. This enables more flexibility in AI application development without the constraints of traditional, closed systems. Bill Sun, Co-Founder and Chief Scientist, said, "On-device multi-modality models will revolutionize daily life. We are building a personal index for each user to create an on-device model that evolves via distributed training on the user's phone. Soon, users won’t need to open multiple apps to complete tasks. The personal AI assistant will understand preferences and manage tasks efficiently." PIN AI's assistant will connect users with apps and services bidding on its blockchain protocol to fulfill tasks like shopping or crypto activities. Bill Sun, Co-Founder and Chief Scientist, said, "On-device multi-modality models will revolutionize daily life. We are building a personal index for each user to create an on-device model that evolves via distributed training on the user's phone. Soon, users won’t need to open multiple apps to complete tasks. The personal AI assistant will understand preferences and manage tasks efficiently." PIN AI's assistant will connect users with apps and services bidding on its blockchain protocol to fulfill tasks like shopping or crypto activities. PIN AI shifts data monetization from big tech to users, allowing secure monetization of personal data. Users receive token incentives through data onboarding and intent fulfillment, only providing necessary data with matched Personal AI Agents, while maintaining control of their information. At launch, PIN AI partners with Worldcoin and is developing a front-end product similar to Siri, expanding its reach and enhancing user experience. Leading the team are co-founders Davide Crapis and Ben Wu. Crapis, formerly of Ethereum Core Research, leads Protocol Research. Ben Wu, heading Strategy, is an MIT graduate, Y Combinator alum, and serial entrepreneur. The technical leadership includes Bill Sun, a Stanford AI/Math PhD and early Google Brain researcher, as Chief AI Scientist, and Regan Peng, a CMU graduate and former lead at Didi Fintech and Yahoo Data Infra, as Founding Head of Engineering. PIN AI collaborates with a16z crypto research, Flashbots, Espresso Systems, and academics from Stanford, Columbia, and NYU. Ben Wu emphasized the importance of an open platform, saying, “The open internet has made large language models possible. We need to build an open platform for users on their trusted device, allowing access to their various data, making their Personal AI possible." Unlike Apple Intelligence, PIN AI's system can run on low-spec smartphones by dynamically shifting between edge AI (on-device) and server AI to optimize performance, ensuring wide accessibility. The funding will expand research, grow the team of AI and blockchain experts, and accelerate the deployment of PIN AI’s technology. The company will join the a16z CSX Fall 2024 cohort in New York City. About PIN AI PIN AI is developing an open-source personal AI operating system that leverages smartphone and app data to empower users with control and privacy. The company's approach aims to connect AI developers with users, promoting data sovereignty and developer empowerment. PIN AI collaborates with Ethereum Core Research and is committed to privacy and innovation. PIN AI PIN AI For more updates, visit pinai.io , X (formerly Twitter) , Telegram , Discord , and Linkedin . pinai.io pinai.io X (formerly Twitter) X (formerly Twitter) Telegram Telegram Discord Discord Linkedin Linkedin Contact PIN AI contact@pinai.io This story was distributed as a release by Chainwire under HackerNoon’s Business Blogging Program. Learn more about the program here. This story was distributed as a release by Chainwire under HackerNoon’s Business Blogging Program. Learn more about the program here . here here

PIN AI Secures $10M From a16z CSX, Hack VC And All-Star Angels

ፒን AI ከ16z CSX፣ Hack VC እና ከኮከብ መላእክቶች 10ሚ.

PIN AI تحصل على 10 ملايين دولار من a16z CSX وHack VC وAll-Star Angels

PIN AI a16z CSX, হ্যাক ভিসি এবং অল-স্টার অ্যাঞ্জেলস থেকে $10M সুরক্ষিত করে

PIN AI zajišťuje 10 milionů dolarů od a16z CSX, Hack VC a All-Star Angels

PIN AI obtiene 10 millones de dólares de a16z CSX, Hack VC y All-Star Angels

PIN AI מאבטח 10 מיליון דולר מ-a16z CSX, Hack VC ו-All-Star Angels

PIN AI、a16z CSX、Hack VC、オールスターエンジェルスから1,000万ドルを調達

PIN AI užtikrina 10 mln. USD iš „a16z CSX“, „Hack VC“ ir „All-Star Angels“

PIN AI Segura $10M A16z CSX, Hack VC Y All-Star Angels

PIN AI siguron 10 milion dollarë nga a16z CSX, Hack VC dhe All-Star Angels

PIN AI Inalinda $10M Kutoka kwa a16z CSX, Hack VC na All-Star Angels

PIN AI Yi Tiyisisa $10M Ku Suka Eka a16z CSX, Hack VC Na Tintsumi Ta Tinyeleti Hinkwato

The Future of Education and AI: Beyond Traditional Degrees with Marc Andreessen & Ben Horowitz

শিক্ষার ভবিষ্যত এবং এআই: মার্ক অ্যান্ড্রেসেন এবং বেন হোরোভিটজের সাথে ঐতিহ্যগত ডিগ্রির বাইরে

Die Zukunft von Bildung und KI: Jenseits traditioneller Abschlüsse mit Marc Andreessen und Ben Horowitz

El futuro de la educación y la IA: más allá de las titulaciones tradicionales con Marc Andreessen y Ben Horowitz

L'avenir de l'éducation et de l'IA : au-delà des diplômes traditionnels avec Marc Andreessen et Ben Horowitz

शिक्षा और एआई का भविष्य: मार्क आंद्रेसेन और बेन होरोविट्ज़ के साथ पारंपरिक डिग्री से परे

教育と AI の未来: マーク・アンドリーセンとベン・ホロウィッツによる従来の学位を超えて

교육과 AI의 미래: Marc Andreessen 및 Ben Horowitz와 함께하는 전통적인 학위를 넘어서

O futuro da educação e da IA: além dos diplomas tradicionais com Marc Andreessen e Ben Horowitz

Будущее образования и искусственного интеллекта: за пределами традиционных степеней с Марком Андриссеном и Беном Горовицем

Eğitimin ve Yapay Zekanın Geleceği: Marc Andreessen ve Ben Horowitz ile Geleneksel Derecelerin Ötesinde

Tương lai của Giáo dục và AI: Vượt xa các bằng cấp truyền thống với Marc Andreessen & Ben Horowitz

教育和人工智能的未来：马克·安德森 (Marc Andreessen) 和本·霍洛维茨 (Ben Horowitz) 的超越传统学位

Hackernoon Gets the low-down from Amir Khan, Chief Executive Officer and Founder of Alkira. Network Cloud pioneer Alkira® has announced a close collaboration with the Microsoft for Startups program, a select group of emerging businesses hand-picked by Microsoft for the benefits they offer to Microsoft Azure customers in their journey to the cloud. The select start-ups receive growth partnership and ecosystem support from Microsoft including access to technical, sales and marketing opportunities and leadership. Alkira also announced that the Alkira Cloud Services Exchange (CSX) – the core of the company’s  Network Cloud platform – is now available on the Azure Marketplace. The Alkira Network Cloud platform, which reduces the time, cost and complexity of connectivity cloud environments, provides Azure customers with increased performance, manageability and security when deploying Windows workloads in the Azure cloud. Microsoft Azure customers can access Alkira’s solution in Azure, or in a multi-cloud network alongside other best-of-breed products in the Azure Marketplace. Alkira CSX is delivered as a secure cloud networking infrastructure as-a-service (CNaaS) that dynamically scales according to customers’ needs and reduces the time taken to deploy cloud networks from months to minutes. Mr Khan, thank you for answering our questions on this news story. Firstly, can you tell us about Alkira? "Alkira was founded by my brother, Atif Khan, who is the company CTO, and me as CEO. We have so far raised $76M from Sequoia Capital, Kleiner Perkins, Koch Disruptive Technologies, LLC, and GV Capital - (formerly Google Ventures). Alkira developed the Network Cloud which enables enterprises to build and deploy a global unified Network Cloud that delivers network connectivity, advanced network and services, and end-to-end visibility and governance. Alkira CSX seamlessly and securely connects branches, data centres, remote users and clouds." "Alkira’s Network Cloud simplifies a customer’s journey to, across and between clouds, providing enhanced routing and security features and operations, and massive simplification of the total enterprise corporate and cloud networks. Alkira’s Network Cloud reduces the time to add Azure from months or years to minutes and hours. Is there an overarching purpose of the service? The time, cost and complexity of building networks in the cloud with end-to-end security and management visibility are major challenges holding back digital transformation. Our service was developed to help businesses overcome the trade-offs, complexity and long lead times associated with cloud networking. The Alkira Network Cloud platform accelerates the deployment and simplifies the management of cloud workloads, leaving customers free to focus on the objectives of their business." How big a deal is the Microsoft deal? "Microsoft recognizes that businesses are increasingly using multiple public and private clouds and sees Alkira as the best multi-cloud network and security solution. We’re excited to be deepening our relationship with Microsoft through the Microsoft for Startups program and to be able to offer those benefits to the growing number of organizations seeking transformational solutions from the Azure Marketplace. Microsoft customers who access Alkira’s Network Cloud through Azure Marketplace benefit from reduced procurement overheads, simplified billing and quicker deployment, as well as the assurance of Microsoft certification. Alkira is one of a handful of strategic collaborators in Microsoft’s start-up program and the only one focused on cloud." Why would Microsoft back a multi-cloud proposition? I can think of two reasons. First, the Alkira Network Cloud benefits Azure customers even if they never leave the Microsoft cloud environment. Networking in any cloud is challenging and the Alkira solution hides the complexities of routing and network management in Azure, making it easier to deploy and manage application workloads, integrate cloud and on-prem services and monitor network traffic from end to end. Second, Microsoft recognizes that enterprise customers are adopting multi-cloud strategies. It can resist that trend or work with it. Flexera’s State of the Cloud 2020 Report says the average enterprise connects to 4.4 clouds and that 93% profess to a multi-cloud strategy. The endorsement of Alkira suggests that Microsoft sees the value in enabling its customers’ digital strategies, not trying to hold them back. Enterprises often connect to different cloud environments to meet different geographic needs or to take advantage of features from a particular provider. But because different clouds implement networking in different ways, it’s much less common to see these clouds working together as a single, seamless environment for enterprise services. Alkira’s Network Cloud platform enables enterprises to build true multi-cloud networks in a matter of hours or days that would take months or years and millions of dollars in development costs using conventional approaches. and hides the differences between cloud environments, allowing security, disaster recovery and other critical functions to be supported across the entire network with the same levels of visibility and management control traditionally associated with on-prem systems. Its solution abstracts the complexities Whether customers rely on a single cloud provider or multiple providers, to enabling the flexibility, scalability and resilience that enterprises need not just to avoid disaster but to ensure that their businesses triumph. the network is the key Was the OVHCloud fire a reminder of why companies should go multi-cloud? "Absolutely. This fire at a major French cloud provider in March took out millions of websites and underlines the risks for businesses putting all of their eggs in one basket. The fire at a facility owned by OVHCloud in March s for government, banks and retailers. Nearly 2% of all websites on the .fr domain (the top-level web domain in France) were affected. Some customers had to wait a month before the company was able to bring servers back up and restore access to their applications and workloads and it has been reported that some data may be permanently lost. took out online service The moral of the story is not that single or multi-cloud is better, but that regardless of where applications are hosted, processing is taking place or data is stored, enterprises need to maintain control of the network that keeps critical business services running." For more information on Alkira: www.alkira.com

Alkira Teams Up with Microsoft to Transform the Cloud Networking Journey to Azure

I have been haunted and daunted by Javascript for almost a decade i.e. 10+ Years. These years do make a history for me in programming. Today while teaching one of my newly made friends on a quick course to programming, I stumbled upon some things that I should share and that is Javascript. Let us try to understand it with some examples. Introduction and History The web has been inclined towards Javascript which made its cover in 1996. A company by the name of Netscape which was famous for its speedy browsers in the mid 90s submitted it to the ECMA International to produce a standardized language which other browsers could implement. I still remember the awe at Netscape Navigator in the 90s. Today we know an entity by the name of Mozilla, a portmanteau of Mosaic and Godzilla. The first version was Mosiac Netscape 0.9, released in 1994. Implementation of Javascript in Browser Code Lets create a folder on your desktop by the name of noobJS and make an index.html file in it. For HTML boilerplate code, I will be using http://htmlshell.com/ Noob JS Hello Noobs! <!DOCTYPE html> < > html < > head < = > meta charset "UTF-8" < = = > meta http-equiv "X-UA-Compatible" content "IE=edge,chrome=1" < = = > meta name "viewport" content "width=device-width,initial-scale=1" < > title </ > title </ > head < > body < > h3 </ > h3 </ > body </ > html Now if you double click the file, it will run in the browser. Button Click Through Vanila JS <body> <button id= >Click< script> Hello Noobs! < > h3 </ > h3 "myFirstButton" /button>

    <script>
        document.getElementById('myFirstButton').onclick = function(e) {
            console.log("Yo Noobs!");
        }
    </ </ > body To start Javascript, we open a "script" tag and then close it with the same tag name. Everything in between is Javascript. "document " is a reference to the current screen which is in view. Now using the getElementById function for document means that Javascript is going to search all elements that have an ID of myFirstButton. An element can be anything from a button or a H3 or any HTML Tag which can be given an ID just by specifying id in it. Now if you right click the browser and click inspect, we will have a console where we can print information using the console.log() function. Now refresh on the browser and click the button. What is JQuery? JQuery is javascript just making it easier to write and more easier to use. It is like re-arranging the same language so that I can make it shorter, more abbreviated hence faster to implement. I would like to metaphorize it as typing Laugh Out Loud again and again but I would prefer using Lol and expect the reader to understand it. Lets write the same click function in JQuery. We will first import JQuery using a CDN (Content Delivery Network). Head over to and click on minified jQuery Core 3.4.1 https://code.jquery.com/ <head> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1">
        <meta name="viewport" content="width=device-width,initial-scale=1">
        <title>Noob JS</title>

        <script src="https://code.jquery.com/jquery-3.4.1.min.js"
            integrity="sha256-CSXorXvZcTkaix6Yvo6HppcZGetbYMGWSFlBw8HfCJo=" crossorigin="anonymous"></script>
    </head> < = > meta charset "UTF-8" Add it to the Head of the index.html file <body>
    ...
    <button id= >JQuery: Click< body> "mySecondBUtton" /button>
    ...
</ < > script .getElementById( ).onclick = { .log( );
}

$( ).click( { .log( );
}); document 'myFirstButton' ( ) function e console "Yo Noobs!" '#mySecondBUtton' ( ) => e console "Not a Noob Anymore!! Heh!!" </ > script Now JQuery once imported initializes with the tag of "$" and then comes the brackets which in quotes becomes a selector from the document. If with context to javascript this is getElementBy, it can either be a class with a dot or a id with a Hash. Therefore, if you compare it is way shorted just like LOLs. Making a GET Call using Vanila JS Lets try to fetch some data through a Get Call, so that we are now communicating in the Javascript. First we need some Mock API and we have a tool for it. Head over to and scroll down to the body. We will write a simple response in JSON (Javascript Object Notation). https://www.mocky.io/ Now click on Generate and it will give you an endpoint for the Get Call. Next create a new button and call it in the script tab <body>
...
<button id= >Javascript: Get Data< body> "myFirstCall" /button>
...
</ < > script ...
document.getElementById( ).onclick = { xhttp = XMLHttpRequest();
    xhttp.onreadystatechange = { ( .readyState == && .status == ) { .log( .responseText);
        }
    };
    xhttp.open( , , );
    xhttp.send();                
}
... 'myFirstCall' ( ) function e var new ( ) function if this 4 this 200 console this "GET" "http://www.mocky.io/v2/5da666b83400001528633045" true </ > script Javascript has an class that handles all the API in the form of an object. This object allows transfer of data between a web browser and a web sever. We have created an object as var xhttp. Next we are listening if there is a change of state which detects whether the call has been run or not. this.readyState is 4 when everything is OK, which puts in a nice check so that during the call we don't break the code and response of 200 means all is good that the webserver responded successfully with some data. this.responseText contains the required data that we want to attain. Next we open the connection and then we send the request. Lets refresh the page and click the button. As you can see we now have the response with all the wiggly brackets. Let convert the response into an object by parsing the JSON. ( .readyState == && .status == ) { .log( .parse( .responseText));            
} if this 4 this 200 console JSON this As you can see by just using JSON.parse() we are able to convert the response from wiggly to usable object. To get the message you can use the following code now; .parse( .responseText).msg JSON this Making a Get Call using JQuery Lets do the same thing with JQuery <body>
...
<button id= >JQuery: Get Data< body> "mySecondCall" /button>
...
</ < > script ...
$( ).click( {
    $.ajax({ : , : , : , : { .log(data);
        }, : { .log(data);
        }
    });
});
... '#mySecondCall' ( ) => e type 'GET' dateType 'JSON' url 'http://www.mocky.io/v2/5da66a18340000a99963304c' success ( ) function data console error ( ) function error console </ > script Jquery has a ajax function for making communication with a web server. Ajax stands for Advanced Java and XML. You provide it with the type of call, type of response and the url and it handles the states in simple readable language. As you can see the same code has been made minimal with the same type of response. Conclusion Going through Vanilla JS really makes up concept to the what is happening first hand. I do recommend everyone to at least go through it so that a better level of understanding can be made but do acknowledge on the other hand the importance of saving time and hassle. Re-inventing the wheel does not bring much change and innovation but in todays time utilizing libraries to speed up work and save time is what one should prefer. Complete Code ~ index.html Noob JS Hello Noobs! Javascript: Click JQuery: Click Javascript: Get Data JQuery: Get Data <!DOCTYPE html> < > html < > head < = > meta charset "UTF-8" < = = > meta http-equiv "X-UA-Compatible" content "IE=edge,chrome=1" < = = > meta name "viewport" content "width=device-width,initial-scale=1" < > title </ > title < = = = > script src "https://code.jquery.com/jquery-3.4.1.min.js" integrity "sha256-CSXorXvZcTkaix6Yvo6HppcZGetbYMGWSFlBw8HfCJo=" crossorigin "anonymous" </ > script </ > head < > body < > h3 </ > h3 < = > button id "myFirstButton" </ > button < = > button id "mySecondBUtton" </ > button < > hr < = > button id "myFirstCall" </ > button < = > button id "mySecondCall" </ > button < > script .getElementById( ).onclick = { xhttp = XMLHttpRequest();
                xhttp.onreadystatechange = { ( .readyState == && .status == ) { .log( .parse( .responseText));            
                    }
                };
                xhttp.open( , , );
                xhttp.send();                
            } .getElementById( ).onclick = { .log( );
            }

            $( ).click( { .log( );
            });

            $( ).click( {
                $.ajax({ : , : , : , : { .log(data);
                    }, : { .log(data);
                    }
                });
            }); document 'myFirstCall' ( ) function e var new ( ) function if this 4 this 200 console JSON this "GET" "http://www.mocky.io/v2/5da66a18340000a99963304c" true document 'myFirstButton' ( ) function e console "Yo Noobs!" '#mySecondBUtton' ( ) => e console "Not a Noob Anymore!! Heh!!" '#mySecondCall' ( ) => e type 'GET' dateType 'JSON' url 'http://www.mocky.io/v2/5da66a18340000a99963304c' success ( ) function data console error ( ) function error console </ > script </ > body </ > html Github https://github.com/th3n00bc0d3r/vanilajsvsjquery

Vanilla JS to Jquery: The Noobs Guide

How to Build a Multi-Lingual Chatbot with IBM Translation API and Ably

The layers and pieces of Cake When you don’t know the details behind a technology, it’s easy to dismiss it as magic, but if you dare to look behind the curtain — it might not be that complicated or foreign to what you already know. In this blog post, I’ll go through some of the pieces that make up the open source build orchestration tool . Cake I’ll show you how you can reuse pieces of Cake with different environments and even languages, I’ll go into detail on some parts and glance over others. If there’s any part you would want me to go into more detail with please let me now! I will provide links to the examples and resources at the end of this blog post. “It’s just C# and .Net” This is a statement I have often made as Cake scripts provide a superset of C# — which means anything you can do in C# you can do Cake. The Cake DSL via its aliases merely provide a shortcut to make APIs and tools easier to consume, but you can always opt to like “regular” code reference assemblies, write methods and classes. An example of this could be working with JSON, in your console or MVC app it’s likely that you would use JSON.Net to serialize/deserialize your JSON from/to .NET objects and doing the same with a Cake script doesn’t differ much from plain vanilla C# Basically, only difference is the preprocessor, which fetches and references an assembly from NuGet and the lack of need to wrap in the code in a namespace/class/method. #addin Being “just” C# and .NET truly means that you’ll always have the safety net of .NET, C# and the vast number of resources available for those pieces of technology. The pieces of Cake Cake is distributed in two flavors, one for the full classic .NET which also works with Mono and a version for .NET Core — the new cross platform framework/runtime from Microsoft. The most common way to obtain the Cake tool is via the and NuGet packages ( ). Cake Cake.CoreCLR Cake is also available via Chocolatey, Homebrew, GitHub and official docker containers are on their way Cake / Cake.CoreCLR NuGet package The NuGet packages contain all dependencies needed to execute under respective runtime, but the “Magic” is in three assemblies and a console application Cake.exe / Cake.dll — “The Tool” To name a few things it handles Compilation of scripts Argument parsing Console logging Composition of modules/assemblies Basically, it’s what glues everything together and provides the “Cake experience” Cake.Core — “The heart” Provides things like Core abstractions/interfaces/attributes used by Cake Tool/Addins/Modules ( ) this is the only assembly needed to reference when creating a Cake addin/module DSL Parsing and transpiling/codegen to C# that Roslyn understands and can compile Base tool implementation ) (tool resolution, execution, settings etc. Cake Task runtime ( ) task runner/dependency graph/setup & teardown etc. Cake.Common — “The doer” This assembly provides almost 400 convenient aliases/extension methods for things like Tools ( ) MSBuild, NuGet, .NET Core CLI, code signing, etc. APIs ( ) compression, project file parsing, text templating, HTTP, etc. Build system support ( AppVeyor, VSTS, Jenkins, TeamCity, MyGet, etc.) Cake.NuGet — “The Fetcher” Cake module that enables fetching dependencies from NuGet for i.e. preprocessor directives like , and . #addin #tool #load Cake composition — “It’s just C# and .Net part deux” The assemblies , and are all available up on NuGet.org targeting both .NET Full/Desktop and .NET Core. Cake.Core Cake.Common Cake.NuGet This means you can reference and leverage parts/most of the work and effort that’s been put into Cake with almost any .NET application, environment or platform — it’s just standard C# code and assemblies! That said Cake relies heavily on dependency injection and has over time been refactored into custom modules — so how it all fits together can be a bit daunting for someone that’s new to the project. A rough basic example of doing this using an Inversion of Control container like AutoFac could look something like this: As the observant might see from the comment there’s a few cases where the implementation currently resides in ( ), the interfaces they implement exist in Core so you can implement and provide your own implementation or depending on what parts of Cake you reuse you might not need them ( ). Cake.exe/Cake.dll this might be something we’ll look at refactoring in the future for unit testing we provide Cake.Testing which provides fake context and file system, environment abstractions for tool testing, etc. Unit testing Cake addins / modules might be a good topic for another blog post — please let me know if you reckon that’s the case Proof of concept custom script host So, to do your own custom host for your own “build script” implementation, you currently need to implement a few interfaces: , , , and , because these implementations as mentioned earlier currently resides in , but all other are available in ready for reuse in any .NET project. ICakeArguments ICakeConfiguration ICakeLog ICakeReportPrinter IScriptHost Cake.exe/Cake.dll Cake.Core To demonstrate this, I’ve created the assembly called , compiled for both .NET and .NET Core, which means it could be used most places .NET is available today ( ). “Proof of concept, in no way official, don’t use in production, just to see how the sausage is made, etc.” Cake.Bridge binary is up on NuGet and source on GitHub, I’ll provide links to all resources at end of this post provides a static class which provides easy access to working with the Cake task runner and the ( ), created with the goal to demonstrate an easy way to reuse Cake assemblies from any .NET language and not only with C# which is what Cake supports out of the box. Cake.Bridge CakeBridge ICakeContext which is what all Cake aliases/methods extend What you won’t get is anything the Cake DSL provides and no addin nor module support. To illustrate this I’ve authored a few code snippets on how using Cake from a few different .NET languages using the assembly, the code snippets are quick and rough proof of concepts to mostly prove it can be done and obviously more refinement needs to be done for them to be more idiomatically correct. Cake.Bridge disclaimer There often exist more native, prominent and widely used build systems already. PowerShell Even if perhaps mainly the sysadmin language of choice — PowerShell still is a .NET based language and its ability to utilize .NET makes it a very powerful scripting language as anything not provided native by the language often can be solved by reaching out to the framework it’s running on. It also means that you can use Cake from it: Visual Basic You can’t talk about .NET languages without mentioning Visual Basic and now when it’s also joining the .NET Core party too it’s getting up to date with the times. And obviously you can Bake with Cake using some Basic ( there’s no VB.NET interactive console that I know of so I created a simple VB.NET Core console app) F# F# is a .NET language which makes it fully possible to use Cake with F#.Full disclaimer though, as I’m not very proficient with the F# language and wanted something that somewhat looked like F# I called on the help and assistance of my friend to do a quick port from C# to F# and this was the result: Mårten Rånge Cake / C# For reference below is the script that was used for inspiration when porting to other languages/runtimes, there’s a little less bootstrapping involved but besides that they’re very similar: Conclusion Even though there might seem to be some Magic around Cake and it’s DSL it’s mostly just .NET and C#, having skills in those areas — are skills you can use with Cake scripts or when using Cake assemblies in some other way. If you have a different need from what the official supported way offers, you could still have a piece of the Cake. Resources Cake.Bridge GitHub: https://github.com/devlead/Cake.Bridge NuGet: https://www.nuget.org/packages/Cake.Bridge PowerShell proof of concept https://github.com/devlead/PsPoC Visual Basic proof of concept https://github.com/devlead/VBPoC F# interactive script proof of concept https://github.com/devlead/FsxPoC C# interactive proof of concept https://github.com/devlead/CsxPoC Cake Website: http://cakebuild.net GitHub: https://github.com/cake-build/cake NuGet: https://www.nuget.org/profiles/cake-build

Dispelling the magic!

Using Azure Functions to automate your pull requests validation If you have an open source project, the more you can automate around the pull requests review process the better, at least that’s what I believe. So if you go through your checklist for expectations you have some things are obvious continuous integration which there already exist good services for, but some of your checks could be unique and currently manual — one my just recent automated check is the example in this post. The Example The example explain in this blog post will be validating so each commit in a pull request comes from an known GitHub user, it’s a quite easy check, but also easy to miss when reviewing yourself as the pull request itself will always originate from an known GitHub user — so just by looking at the pull request on GitHub it’s not an obvious issue. The reason to do this could be i.e. if you want to automate attribution of authors in release notes, validate CLAs and so on. The reason for this occurring often isn’t malicious intent, but rather that a to GitHub unknown email address had been used to do the commit, i.e. issuers corporate email address. The easiest fix is just to add all email addresses under their GitHub Settings/Emails, the other option is to amend the commit with the desired email address and do force push with the new commit. The Concept FaaS — Functions as a Service — is very well suited for these kind of scenarios as you don’t need a web site/server up and running 24/7, but instead FaaS enables you to have code that’s only up and running when invoked — where the idea is paying only for what you use when you use it. Roughly the flow will be GitHub triggers a webhook call to a function, the function will parse out what’s needed from the GitHub payload and place it on an Azure Storage Queue, the queue will trigger a function that does the validation and reports back status to the pull request using the GitHub rest API. New Function I usually use the Azure Functions CLI scaffolding, write my functions in VS Code and then deploy via KUDU or CI/CD of choice. But the Azure portal is quite capable too and can be a quick way to get started and it’s fully possible do get the code to source control later ( ) source control should always be the way when in production, but when getting started, prototyping and playing around the portal is actually quite productive Azure Functions has a ready template to quickly get started with a GitHub webhooks. This template will give you a good starting point to dynamically parse the data posted from GitHub, making easy to support most scenarios. Adding a queue So now we got data setup to come in to the service and got the ability to parse the incoming data. Now we just want to quickly cherry pick what’s needed out of the GitHub payload, push it to a queue for later processing and tell GitHub we got it. Go to “Integrate” to add a queue output via the portal, click “+ New Output”, in this case we want an Azure Storage Queue You then give the “queue output” a parameter name which will be the name of the output parameter in your function, then name of the queue to push messages to and last choose the where your storage queue connection string resides.Hit save and our queue parameter binding is ready. appsetting Filling the queue Now we just need add it as an output parameter to our function so we can assign content to it: As you might see here using output parameter you lose the async/await capabilities, if you don’t need multiple outputs and just want to return“200 — OK” if processed and queued successfully, then you can keep using async/await by going under “Integrate” removing the HTTP result binding and setting the queue to bind to the return value of the function. So a complete function that parses and queues while still supporting async/await could look something like this: Typed parameter A neat thing is that the queue item can be changed to custom class and then it will automatically serialize and deserialize the JSON string.So lets create an class and to keep it nice and tidy we’ll add it to it’s own file.Hit “View Files”->”+Add”, then type in the filename hit enter. The class to our payload would look something like below We can then in our function use the directive like , it’s available to us in our function and we can change our function method signature to a typed return value and end up with something like below: run.csx #load #load "GitHubPayload.csx" So what does the above do? Not much and a lot, it parses the GitHub ’s action, URL to post status to and URL to fetch pull request commits. As we bound the return value to the queue it’ll be automatically serialized as JSON and queued. pull request event Pop the queue Now we need something to pop the queue and do the actual work. To get up and running quickly Azure Functions actually offers an scaffolding option that will get you up and running quickly. If you go to Integrate and the queue output we previously created there’s a “Create a new function triggered by this output” action . This will pre-select a C# Queue Trigger, fill in the queue name and the storage account connection used. All you need to do his hit create, but you probably want to name your function more meaningful than too. QueueTriggerCSharp1 The function crated will look something like below. It gets triggered when something gets enqueued and the dequeued value is what you get in and the template example just outputs it to the trace log. Parse & Validate Also the dequeue trigger can be typed by reusing the class we previous created, this way Azure Functions will automatically deserialize the queued JSON into an typed object we can easily work with. The above code will validate that it’s an event/action we’re interesting and make sure we’ve got status API URL needed to report back status and commits API URL needed to fetch pull request commit details. GitHub authorization Now we got queue item parsed and ready, next step is to fetch the commits from GitHub, before we can do that we need to be able to authenticate against them. GitHub has several ways of against their APIs, one way is thru personal access tokens, which is what I’ll be using for this blog post. authenticating You’ll find your personal access tokens by going to “Settings” and then click on “Personal access tokens” in the left menu. Then click the “Generate Token” button, give the token a description, choose the required access, there’s loads of available permissions, but for this post “repo:status” is the only access needed Hit “Generate token” and you’ll be presented with the newly generated token. App settings We don’t want to store our token in our source code, fortunately Azure Functions just as the other Azure app services provide means to store app settings. You find it under “Function app settings” -> “Configure app settings” There you have a list of key value pairs, which you can add, delete and modify both app settings and connection strings. All app settings & connection strings can be accessed as environment variables, which makes it easy to access regardless of which of the languages supported by Azure Functions you choose to use. Connection strings environment variables are prefixed with it’s provider, i.e. if you choose the custom provider it’ll be prefixed , so if you enter a variable named you’ll access it as . CUSTOMCONNSTR_ GITHUB_TOKEN CUSTOMCONNSTR_GITHUB_TOKEN Calling GitHub So now we got our token and app setting in place we want to and iterate over pull request commits, validate each and then to if it passed or failed validation. The .NET is used to call to the GitHub API, for convenience and to keep it nice and tidy I’ve wrapped it in two helper methods and , both utilizing the same method for common headers and authentication, all placed in it’s own file get post HttpClient GetObjectAsync PostObjectAsJsonAsync HttpClientHelper.csx Finishing the puzzle Putting all the pieces together we end up with a function like this And we now we’re feature complete, ready to queue GitHub webhook request, validate input, validate and report that commits has an registered GitHub user as author. Configuring webhook Now we just need to configure GitHub to call our function when a pull request is created or modified. This is done under repository (or organization) settings “Webhooks”->”Add webhook” You can under your GitHub webhook function find the URL and secret to be used. Enter these as Payload URL, Secret and select as Content type application/json Choose to select individual events and select only the pull request event then just hit “Add webhook” and your done. Testing first PR Now we create our first PR to the repository we’ve configured our webhook on. If all goes to plan, you should see that 1 author validation has been performed and as I’m a known GitHub user the test passes and all’s OK. Now we’ll add a commit from an unknown user, easiest way to test this is by overriding the author when doing the commit. From command line that would look something like this git commit -a -m "Unknown update" --author "John < >" John@unknown.doe.com Then we push our changes, it’ll trigger the GitHub pull request event and our functions synchronize Pressing the details link will take us to the commit displaying the unknown author. Conclusion It probably took a lot longer to read this post than it would take to get the functions up and running — as with many things it’s easy when you know it. It really lets you focus on what you want to solve and as I’ve shown combining with third party APIs is super simple which makes it really powerful. Working with queues is a breeze and for incoming web hooks that’s a great approach as you just receive and acknowledge — doing the actual work in a separate function, you can hook up functions to a poisonous message queue for each queue making it a very robust solution — without turning it complex. too On a final note extending your GitHub pull request review process with small automatic checks is a great way to not only save you time, but also speed up the communication towards your contributors — and the sky truly is the limit on what you can check! You can also do bot like things like certain maintainer comments triggering a integration test, creating new issues, a Microsoft Teams message, etc. So I think for these scenarios the functions as a service approach fits like a glove! What do you think? Would love to hear your feedback! The Code The complete code for the functions in this blog post can be found on GitHub _Contribute to OnlyKnownCommitters development by creating an account on GitHub._github.com azurevoodoo/OnlyKnownCommitters Previous posts If you want to learn more Azure Functions please checkout some of the previous posts I’ve done on the topic! _Breathe life into your static website using Azure Functions_hackernoon.com Bring static to life using serverless⚡ _Using Azure functions to hook up Azure app service deployment notifications to a Microsoft Team channel_medium.com ⚡Azure Deployment -> Microsoft Teams _Why should JavaScript developers have all the fun?_medium.com Going serverless with PowerShell

GitHub statuses made easy with FaaS

Amazon has been salivating at the multitrillion-dollar United States healthcare market since 2018. Although Amazon Health played a vital role in public health during the pandemic, most still don’t view it as a disruptor. A new chapter in the American conglomerate’s move into the industry is upon us — Health Condition Programs. How does this initiative fit into the tech titan’s grand healthcare plan? It’s anybody’s guess. I can only speculate about the countless scenarios surrounding the interplay between Health Condition Programs and Amazon’s diverse solutions. However, the real question we should all be asking is — will it buck the trend of the organization’s cybersecurity, patient safety, and privacy scandals? A Treasure Trove of Patient Data On paper, Health Condition Programs aim to encourage millions of Americans to maximize their unused healthcare features. This mission has merit because over one-third of workers don’t fully understand the employee benefits they signed up for. Many American adults might be unaware that they can receive care to manage chronic conditions — such as diabetes and hypertension — between doctor visits at no extra cost to them. Enrolling in these programs allows insured individuals to maximize their coverage and squeeze more value out of the rising premiums or contributions they shoulder. In practice, Health Condition Programs enable the Andy Jassy-led all-encompassing empire to grow its patient data repository by tens of millions. This offering’s coverage check involves eligibility verification, necessitating personal data collection. Enrollees must share their legal name, date of birth, employer’s name, insurer's name, email address, and other sensitive information. This will determine whether their insurance plan or employee benefits cover health programs Amazon’s partners offer. After confirming an applicant is covered, Amazon Health directs the patient to third-party digital health companies — the actual virtual care providers — to complete registration and program enrollment. While Amazon may say it stops collecting personal data after a coverage check, it may still follow patients beyond that point. This Big Tech firm is a known participant in server-to-server tracking, so I would be more surprised if it doesn’t monitor the rest of an enrollee’s telehealth journey through a backdoor. Connected devices are notable program features that covered individuals can access. Continuous glucose monitor sensors, blood sugar meters, blood pressure monitors, and smart scales generate personalized quantitative data that forms a part of medical records and is personal. This is protected health information tech conglomerates could leverage to enrich their subsidiaries. A Slowly Expanding Network As disturbing as it is to think that Amazon could acquire bits and pieces of millions of medical records under the table through its Health Condition Programs, this multinational’s reach is proportional to its network’s size. As of October 2024, only two digital health companies have participated — Omada Health and Talkspace. Omada specializes in diabetes prevention, weight management, and hypertension care, while Talkspace delivers virtual therapy and psychiatric care. Between the two, insured Americans can choose from only three programs. It’s unclear why Amazon Health could only ink two partnerships in nine months — a departure from the online retail behemoth’s “Get Big Fast” mantra. I bet telehealth providers would move mountains to be in this exclusive club, so Amazon must have been overly selective about its partners for reasons unknown. Only time will tell if Amazon will bring more digital health companies into its fold sooner rather than later. A safe guess is that it would prioritize partner diversity to prevent cannibalization and harvest data with variety. A New Piece to the Amazon Health Care Strategy Puzzle Amazon plays the long game. For all we know, it might envision itself as the savior of health care — the ultimate medical data synthesizer the sector sorely needs. It’s impossible to connect all the dots when many of them haven’t come to light. Still, recognizing this offering’s potential value to this tech colossus isn’t brain surgery. If Health Condition Programs gain currency and the number of participating digital health companies on the network balloons, the company will better understand our health needs. These insights can be the rising tide that lifts the boats in the Amazon Health sub-ecosystem and other related subsidiaries. A quick thought experiment would reveal that more program enrollees could increase Amazon One Medical’s in-person visits. Health Condition Programs almost always require a prior medical diagnosis. Health care professionals on the network — health coaches, licensed therapists and psychiatrists, nutritionists, dieticians, and personal care specialists — could help train AWS HealthScribe with large volumes of recordings. The aim would be to churn out error-free transcripts and clinical notes consistently. The company could expand Alexa’s footprint by integrating its partners’ services into the voice assistant. It’s a tried-and-true move, as evidenced by the partnership between Amazon and Teladoc Health in 2022. Prescriptions from digital health companies can also boost Amazon Pharmacy’s sales. Analyzing mountains of medical data can shed light on the usual indicators of events that can endanger patients’ lives. Tech-forward hospitals have taken this route. They proved it’s feasible to predict signs of upcoming acute events — which increase the chances of patient mortality by three to four times — early enough to address them before they happen. Amazon can do the same thing for people living with chronic diseases. It can mobilize Whole Foods to make healthier food options more widely available to those who need them most. Accessible healthy grocery items empower chronically ill individuals to meet their unique nutritional needs and keep symptoms manageable. Are Health Condition Programs a Scandal Waiting to Happen? Amazon’s venture into health care has been riddled with controversies. In 2023, hackers successfully victimized more than 19,000 PillPack customers, stealing their names, residential addresses, phone numbers, and email addresses. About 3,600 of the impacted accounts included prescription information. That same year, senators grilled the company for forcing patients to waive their right to health information privacy to enroll in Amazon Clinic. In mid-2024, Amazon found itself in hot water for allegedly compromising the safety of One Medical patients. Incidents of awful misjudgments by staff assisting customers by phone have risen since the subsidiary entrusted the job to inadequately trained agents. Amazon has failed to give a good account of itself in the sector on various occasions. At this rate, I feel like someone would expose a dubious practice underpinning Health Condition Programs any moment now. Can We Trust Amazon’s Health Condition Programs? Criticisms toward this initiative are unsubstantiated and premature. However, it would be naive to expect Amazon not to supercharge its health data collection prowess through coverage checks and its partners when the alternative does nothing to improve other parts of its business. We’ll just have to wait and see whether this Amazon Health offering’s practical benefits outweigh its perceived privacy drawbacks — but let’s not hold our breaths. Amazon has been salivating at the multitrillion-dollar United States healthcare market since 2018. Although Amazon Health played a vital role in public health during the pandemic, most still don’t view it as a disruptor. A new chapter in the American conglomerate’s move into the industry is upon us — Health Condition Programs. How does this initiative fit into the tech titan’s grand healthcare plan? It’s anybody’s guess. I can only speculate about the countless scenarios surrounding the interplay between Health Condition Programs and Amazon’s diverse solutions. However, the real question we should all be asking is — will it buck the trend of the organization’s cybersecurity, patient safety, and privacy scandals? A Treasure Trove of Patient Data On paper, Health Condition Programs aim to encourage millions of Americans to maximize their unused healthcare features. This mission has merit because over one-third of workers don’t fully understand the employee benefits they signed up for. one-third of workers don’t fully understand one-third of workers don’t fully understand Many American adults might be unaware that they can receive care to manage chronic conditions — such as diabetes and hypertension — between doctor visits at no extra cost to them. Enrolling in these programs allows insured individuals to maximize their coverage and squeeze more value out of the rising premiums or contributions they shoulder. In practice, Health Condition Programs enable the Andy Jassy-led all-encompassing empire to grow its patient data repository by tens of millions. This offering’s coverage check involves eligibility verification, necessitating personal data collection. Enrollees must share their legal name, date of birth, employer’s name, insurer's name, email address, and other sensitive information. This will determine whether their insurance plan or employee benefits cover health programs Amazon’s partners offer. After confirming an applicant is covered, Amazon Health directs the patient to third-party digital health companies — the actual virtual care providers — to complete registration and program enrollment. While Amazon may say it stops collecting personal data after a coverage check, it may still follow patients beyond that point. This Big Tech firm is a known participant in server-to-server tracking , so I would be more surprised if it doesn’t monitor the rest of an enrollee’s telehealth journey through a backdoor. a known participant in server-to-server tracking a known participant in server-to-server tracking Connected devices are notable program features that covered individuals can access. Continuous glucose monitor sensors, blood sugar meters, blood pressure monitors, and smart scales generate personalized quantitative data that forms a part of medical records and is personal. This is protected health information tech conglomerates could leverage to enrich their subsidiaries. forms a part of medical records forms a part of medical records A Slowly Expanding Network As disturbing as it is to think that Amazon could acquire bits and pieces of millions of medical records under the table through its Health Condition Programs, this multinational’s reach is proportional to its network’s size. As of October 2024, only two digital health companies have participated — Omada Health and Talkspace. Omada specializes in diabetes prevention, weight management, and hypertension care, while Talkspace delivers virtual therapy and psychiatric care. Between the two, insured Americans can choose from only three programs. It’s unclear why Amazon Health could only ink two partnerships in nine months — a departure from the online retail behemoth’s “Get Big Fast” mantra. I bet telehealth providers would move mountains to be in this exclusive club, so Amazon must have been overly selective about its partners for reasons unknown. Only time will tell if Amazon will bring more digital health companies into its fold sooner rather than later. A safe guess is that it would prioritize partner diversity to prevent cannibalization and harvest data with variety. A New Piece to the Amazon Health Care Strategy Puzzle Amazon plays the long game. For all we know, it might envision itself as the savior of health care — the ultimate medical data synthesizer the sector sorely needs. It’s impossible to connect all the dots when many of them haven’t come to light. Still, recognizing this offering’s potential value to this tech colossus isn’t brain surgery. If Health Condition Programs gain currency and the number of participating digital health companies on the network balloons, the company will better understand our health needs. These insights can be the rising tide that lifts the boats in the Amazon Health sub-ecosystem and other related subsidiaries. A quick thought experiment would reveal that more program enrollees could increase Amazon One Medical’s in-person visits. Health Condition Programs almost always require a prior medical diagnosis. Health care professionals on the network — health coaches, licensed therapists and psychiatrists, nutritionists, dieticians, and personal care specialists — could help train AWS HealthScribe with large volumes of recordings. The aim would be to churn out error-free transcripts and clinical notes consistently. The company could expand Alexa’s footprint by integrating its partners’ services into the voice assistant. It’s a tried-and-true move, as evidenced by the partnership between Amazon and Teladoc Health in 2022. Prescriptions from digital health companies can also boost Amazon Pharmacy’s sales. partnership between Amazon and Teladoc Health partnership between Amazon and Teladoc Health Analyzing mountains of medical data can shed light on the usual indicators of events that can endanger patients’ lives. Tech-forward hospitals have taken this route. They proved it’s feasible to predict signs of upcoming acute events — which increase the chances of patient mortality by three to four times — early enough to address them before they happen. by three to four times by three to four times Amazon can do the same thing for people living with chronic diseases. It can mobilize Whole Foods to make healthier food options more widely available to those who need them most. Accessible healthy grocery items empower chronically ill individuals to meet their unique nutritional needs and keep symptoms manageable. Are Health Condition Programs a Scandal Waiting to Happen? Amazon’s venture into health care has been riddled with controversies. In 2023, hackers successfully victimized more than 19,000 PillPack customers, stealing their names, residential addresses, phone numbers, and email addresses. About 3,600 of the impacted accounts included prescription information. 3,600 of the impacted accounts 3,600 of the impacted accounts That same year, senators grilled the company for forcing patients to waive their right to health information privacy to enroll in Amazon Clinic. In mid-2024, Amazon found itself in hot water for allegedly compromising the safety of One Medical patients. Incidents of awful misjudgments by staff assisting customers by phone have risen since the subsidiary entrusted the job to inadequately trained agents. Amazon has failed to give a good account of itself in the sector on various occasions. At this rate, I feel like someone would expose a dubious practice underpinning Health Condition Programs any moment now. Can We Trust Amazon’s Health Condition Programs? Criticisms toward this initiative are unsubstantiated and premature. However, it would be naive to expect Amazon not to supercharge its health data collection prowess through coverage checks and its partners when the alternative does nothing to improve other parts of its business. We’ll just have to wait and see whether this Amazon Health offering’s practical benefits outweigh its perceived privacy drawbacks — but let’s not hold our breaths.

Did You Know About Amazon’s Health Condition Programs? What This Means for Users

Elon Musk v OpenAI, Court Filing, retrieved on April 30, 2024, is part of HackerNoon’s Legal PDF Series. You can jump to any part of this filing here. This part is 5 of 29. B. Altman Induces Musk to Back OpenAI, Inc. 68. From the start, Altman courted Musk by presenting himself as sharing Musk’s well-known concerns over the threat posed by AI/AGI. Altman, an experienced tech player, feigned altruism to convince Musk into giving him free start-up capital and recruiting top AI scientists to develop technological assets from which Defendants would stand to make billions. 69. Altman began by testing the waters. In early March 2015, he approached Musk to help draft an open letter to the U.S. Government emphasizing the need for regulation to ensure the safe creation of AI. Musk agreed, and the two began preparing the open letter and approaching Musk’s influential contacts in the technology and AI sectors about signing the letter, which was published on October 28, 2015. 70. Sensing opportunity, Altman suggested to Musk on May 25, 2015 that they endeavor to beat Google in the race to develop AGI. He wrote that he’d “[b]een thinking a lot about whether it’s possible to stop humanity from developing AI. I think the answer is almost definitely not. If it’s going to happen, it seems like it would be good for someone other than Google to do it first.” Altman proposed they start an AI “Manhattan Project” and, to win Musk’s backing, offered to “structure it so that the tech belongs to the world via some sort of nonprofit but the people working on it get startup-like compensation if it works. Obviously we’d comply with/aggressively support all regulation.” Still noncommittal, Musk merely responded: “Probably worth a conversation.” 71. To convince Musk of his sincerity, Altman promised that he too would have skin in the game and would make meaningful financial contributions to the non-profit. It has since been revealed, however, that Altman grossly inflated and misrepresented his actual contributions, which pale in comparison to what he had promised. 72. A month later on June 24, 2015, Altman tried again, this time wooing Musk with a detailed proposal for a new AI lab: “The mission would be to create the first general AI [AGI] and use it for individual empowerment—ie [sic], the distributed version of the future that seems the safest. More generally, safety should be a first-class requirement.” “The technology would be owned by the foundation and used ‘for the good of the world[.]’” This time Musk agreed. 73. Soon thereafter, Altman recruited Stripe’s CTO Gregory Brockman who helped him seal the deal. 74. Altman’s plan worked. In November 2015, Musk agreed to commit funding and help recruit the top scientists necessary to make Altman’s project a success provided that—as Altman and Brockman had repeatedly promised— OpenAI, Inc. would be a non-profit devoted to developing AI/AGI for the benefit of humanity and would accomplish this mission by (i) distributing its research and technology openly, preventing its concentration, and (ii) focusing on safety, not profits. Indeed, to celebrate what he was led to believe was their mission, Musk named the endeavor—“OpenAI.” 75. Altman moved fast. Just a month later on December 8, 2015, a Certificate of Incorporation for OpenAI, Inc. was filed with the Delaware Secretary of State, which reaffirmed Altman and Brockman’s promises to Musk: This Corporation shall be a nonprofit corporation organized exclusively for charitable and/or educational purposes within the meaning of section 501(c)(3) of the Internal Revenue Code of 1986, as amended, or the corresponding provision of any future United States Internal Revenue law. The specific purpose of this corporation is to provide funding for research, development and distribution of technology related to artificial intelligence. The resulting technology will benefit the public and the corporation will seek to open source technology for the public benefit when applicable. The corporation is not organized for the private gain of any person. . . . The property of this corporation is irrevocably dedicated to the[se] purposes . . . and no part of the net income or assets of this corporation shall ever inure to the benefit of any director, officer or member thereof or to the benefit of any private person (emphasis added). 76. OpenAI, Inc. was publicly announced on December 11, 2015 and leveraged Musk’s name by making him co-chair of its Board of Directors (“Board”) alongside Altman, with Brockman as the CTO. The promotional announcement published on OpenAI’s website further touted: “OpenAI is a nonprofit artificial intelligence research company [whose] goal is to advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return. Since our research is free from financial obligations, we can better focus on a positive human impact.” 77. Around this same time, Altman and Brockman also began to broadly advertise their new endeavor, specifically promoting its humanitarian purpose on their respective social media accounts, which reach many millions of followers. Continue Reading Here. About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings.
This court case retrieved on August 05, 2024, deadline.com is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. Elon Musk v OpenAI, Court Filing, retrieved on April 30, 2024, is part of HackerNoon’s Legal PDF Series. You can jump to any part of this filing here. This part is 5 of 29. Elon Musk v OpenAI, Court Filing, retrieved on April 30, 2024, is part of HackerNoon’s Legal PDF Series . You can jump to any part of this filing here . This part is 5 of 29. HackerNoon’s Legal PDF Series here B. Altman Induces Musk to Back OpenAI, Inc. 68. From the start, Altman courted Musk by presenting himself as sharing Musk’s well-known concerns over the threat posed by AI/AGI. Altman, an experienced tech player, feigned altruism to convince Musk into giving him free start-up capital and recruiting top AI scientists to develop technological assets from which Defendants would stand to make billions. 69. Altman began by testing the waters. In early March 2015, he approached Musk to help draft an open letter to the U.S. Government emphasizing the need for regulation to ensure the safe creation of AI. Musk agreed, and the two began preparing the open letter and approaching Musk’s influential contacts in the technology and AI sectors about signing the letter, which was published on October 28, 2015. 70. Sensing opportunity, Altman suggested to Musk on May 25, 2015 that they endeavor to beat Google in the race to develop AGI. He wrote that he’d “[b]een thinking a lot about whether it’s possible to stop humanity from developing AI. I think the answer is almost definitely not. If it’s going to happen, it seems like it would be good for someone other than Google to do it first.” Altman proposed they start an AI “Manhattan Project” and, to win Musk’s backing, offered to “structure it so that the tech belongs to the world via some sort of nonprofit but the people working on it get startup-like compensation if it works. Obviously we’d comply with/aggressively support all regulation.” Still noncommittal, Musk merely responded: “Probably worth a conversation.” 71. To convince Musk of his sincerity, Altman promised that he too would have skin in the game and would make meaningful financial contributions to the non-profit. It has since been revealed, however, that Altman grossly inflated and misrepresented his actual contributions, which pale in comparison to what he had promised. 72. A month later on June 24, 2015, Altman tried again, this time wooing Musk with a detailed proposal for a new AI lab: “The mission would be to create the first general AI [AGI] and use it for individual empowerment—ie [sic], the distributed version of the future that seems the safest. More generally, safety should be a first-class requirement.” “The technology would be owned by the foundation and used ‘for the good of the world[.]’” This time Musk agreed. 73. Soon thereafter, Altman recruited Stripe’s CTO Gregory Brockman who helped him seal the deal. 74. Altman’s plan worked. In November 2015, Musk agreed to commit funding and help recruit the top scientists necessary to make Altman’s project a success provided that—as Altman and Brockman had repeatedly promised— OpenAI, Inc. would be a non-profit devoted to developing AI/AGI for the benefit of humanity and would accomplish this mission by (i) distributing its research and technology openly, preventing its concentration, and (ii) focusing on safety, not profits. Indeed, to celebrate what he was led to believe was their mission, Musk named the endeavor—“OpenAI.” 75. Altman moved fast. Just a month later on December 8, 2015, a Certificate of Incorporation for OpenAI, Inc. was filed with the Delaware Secretary of State, which reaffirmed Altman and Brockman’s promises to Musk: This Corporation shall be a nonprofit corporation organized exclusively for charitable and/or educational purposes within the meaning of section 501(c)(3) of the Internal Revenue Code of 1986, as amended, or the corresponding provision of any future United States Internal Revenue law. The specific purpose of this corporation is to provide funding for research, development and distribution of technology related to artificial intelligence. The resulting technology will benefit the public and the corporation will seek to open source technology for the public benefit when applicable. The corporation is not organized for the private gain of any person. . . . The property of this corporation is irrevocably dedicated to the[se] purposes . . . and no part of the net income or assets of this corporation shall ever inure to the benefit of any director, officer or member thereof or to the benefit of any private person (emphasis added). The specific purpose of this corporation is to provide funding for research, development and distribution of technology related to artificial intelligence. The resulting technology will benefit the public and the corporation will seek to open source technology for the public benefit when applicable. The corporation is not organized for the private gain of any person. . . . The property of this corporation is irrevocably dedicated to the[se] purposes . . . and no part of the net income or assets of this corporation shall ever inure to the benefit of any director, officer or member thereof or to the benefit of any private person 76. OpenAI, Inc. was publicly announced on December 11, 2015 and leveraged Musk’s name by making him co-chair of its Board of Directors (“Board”) alongside Altman, with Brockman as the CTO. The promotional announcement published on OpenAI’s website further touted: “OpenAI is a nonprofit artificial intelligence research company [whose] goal is to advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return. Since our research is free from financial obligations, we can better focus on a positive human impact.” 77. Around this same time, Altman and Brockman also began to broadly advertise their new endeavor, specifically promoting its humanitarian purpose on their respective social media accounts, which reach many millions of followers. Continue Reading Here. Continue Reading Here . Here About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case retrieved on August 05, 2024, deadline.com is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case retrieved on August 05, 2024, deadline.com is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. deadline.com

Elon Musk Alleges OpenAI Was a Bait-and-Switch Scheme

Authors:
(1) Shengqiong Wu, NExT++, School of Computing, National University of Singapore;
(2) Hao Fei ,from NExT++, School of Computing at the National University of Singapore, serves as the corresponding author: haofei37@nus.edu.sg.
(3) Leigang Qu, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;;
(4) Wei Ji, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;;
(5) Tat-Seng Chua, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;. Table of Links Abstract and 1. Introduction
2 Related Work3 Overall Architecture4 Lightweight Multimodal Alignment Learning
5 Modality-switching Instruction Tuning
5.1 Instruction Tuning
5.2 Instruction Dataset
6 Experiments
6.1 Any-to-any Multimodal Generation and 6.2 Example Demonstrations
7 Conclusion and References 2 Related Work Cross-modal Understanding and Generation Our world is replete with multimodal information, wherein we continuously engage in the intricate task of comprehending and producing cross-modal content. The AI community correspondingly emerges varied forms of cross-modal learning tasks, such as Image/Video Captioning [99, 16, 56, 56, 27, 49], Image/Video Question Answering [94, 90, 48, 98, 3], Text-to-Image/Video/Speech Synthesis [74, 30, 84, 23, 17, 51, 33], Image-to-Video Synthesis [18, 37] and more, all of which have experienced rapid advancements in past decades. Researchers have proposed highly effective multimodal encoders, with the aim of constructing unified representations encompassing various modalities. Meanwhile, owing to the distinct feature spaces of different modalities, it is essential to undertake modality alignment learning. Moreover, to generate high-quality content, a multitude of strong-performing methods have been proposed, such as Transformer [82, 101, 17, 24], GANs [53, 7, 93, 110], VAEs [81, 67], Flow models [73, 6] and the current state-of-the-art diffusion models [31, 64, 57, 22, 68]. Especially, the diffusion-based methods have recently delivered remarkable performance in a plethora of cross-modal generation tasks, such as DALL-E [66], Stable Diffusion [68]. While all previous efforts of cross-modal learning are limited to the comprehension of multimodal inputs only, CoDi [78] lately presents groundbreaking development. Leveraging the power of diffusion models, CoDi possesses the ability to generate any combination of output modalities, including language, images, videos, or audio, from any combination of input modalities in parallel. Regrettably, CoDi might still fall short of achieving human-like deep reasoning of input content, with only parallel cross-modal feeding&generation. Multimodal Large Language Models LLMs have already made profound impacts and revolutions on the entire AI community and beyond. The most notable LLMs, i.e., OpenAI’s ChatGPT [59] and GPT4 [60], with alignment techniques such as instruction tuning [61, 47, 104, 52] and reinforcement learning from human feedback (RLHF) [75], have demonstrated remarkable language understanding and reasoning abilities. And a series of open-source LLMs, e.g., Flan-T5 [13], Vicuna [12], LLaMA [80] and Alpaca [79], have greatly spurred advancement and made contributions to the community [109, 100]. Afterward, significant efforts have been made to construct LLMs dealing with multimodal inputs and tasks, leading to the development of MM-LLMs. On the one hand, most of the researchers build fundamental MM-LLMs by aligning the well-trained encoders of various modalities to the textual feature space of LLMs, so as to let LLMs perceive other modal inputs [35, 109, 76, 40]. For example, Flamingo [1] uses a cross-attention layer to connect a frozen image encoder to the LLMs. BLIP-2 [43] employs a Q-Former to translate the input image queries to the LLMs. LLaVA [52] employs a simple projection scheme to connect image features into the word embedding space. There are also various similar practices for building MM-LLMs that are able to understand videos (e.g., Video-Chat [44] and Video-LLaMA [103]), audios (e.g., SpeechGPT [102]), etc. Profoundly, PandaGPT [77] achieves a comprehensive understanding of six different modalities simultaneously by integrating the multimodal encoder, i.e., ImageBind [25]. Nevertheless, these MM-LLMs all are subject to the limitation of only perceiving multimodal data, without generating content in arbitrary modalities. To achieve LLMs with both multimodal input and output, some thus explore employing LLMs as decision-makers, and utilizing existing off-theshelf multimodal encoders and decoders as tools to execute multimodal input and output, such as Visual-ChatGPT [88], HuggingGPT [72], and AudioGPT [34]. As aforementioned, passing messages between modules with pure texts (i.e., LLM textual instruction) under the discrete pipeline scheme will inevitably introduce noises. Also lacking comprehensive tuning across the whole system significantly limits the efficacy of semantics understanding. Our work takes the mutual benefits of both the above two types, i.e., learning an any-to-any MM-LLM in an end-to-end manner. This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. Authors: (1) Shengqiong Wu, NExT++, School of Computing, National University of Singapore; (2) Hao Fei ,from NExT++, School of Computing at the National University of Singapore, serves as the corresponding author: haofei37@nus.edu.sg. (3) Leigang Qu, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;; (4) Wei Ji, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;; (5) Tat-Seng Chua, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;. Authors: Authors: (1) Shengqiong Wu, NExT++, School of Computing, National University of Singapore; (2) Hao Fei ,from NExT++, School of Computing at the National University of Singapore, serves as the corresponding author: haofei37@nus.edu.sg. (3) Leigang Qu, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;; (4) Wei Ji, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;; (5) Tat-Seng Chua, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;. Table of Links Abstract and 1. Introduction 2 Related Work3 Overall Architecture4 Lightweight Multimodal Alignment Learning 5 Modality-switching Instruction Tuning 5.1 Instruction Tuning 5.2 Instruction Dataset 6 Experiments 6.1 Any-to-any Multimodal Generation and 6.2 Example Demonstrations 7 Conclusion and References Abstract and 1. Introduction Abstract and 1. Introduction 2 Related Work 3 Overall Architecture 4 Lightweight Multimodal Alignment Learning 2 Related Work 3 Overall Architecture 4 Lightweight Multimodal Alignment Learning 5 Modality-switching Instruction Tuning 5.1 Instruction Tuning 5.1 Instruction Tuning 5.2 Instruction Dataset 5.2 Instruction Dataset 6 Experiments 6.1 Any-to-any Multimodal Generation and 6.2 Example Demonstrations 6.1 Any-to-any Multimodal Generation and 6.2 Example Demonstrations 7 Conclusion and References 7 Conclusion and References 2 Related Work Cross-modal Understanding and Generation Our world is replete with multimodal information, wherein we continuously engage in the intricate task of comprehending and producing cross-modal content. The AI community correspondingly emerges varied forms of cross-modal learning tasks, such as Image/Video Captioning [99, 16, 56, 56, 27, 49], Image/Video Question Answering [94, 90, 48, 98, 3], Text-to-Image/Video/Speech Synthesis [74, 30, 84, 23, 17, 51, 33], Image-to-Video Synthesis [18, 37] and more, all of which have experienced rapid advancements in past decades. Researchers have proposed highly effective multimodal encoders, with the aim of constructing unified representations encompassing various modalities. Meanwhile, owing to the distinct feature spaces of different modalities, it is essential to undertake modality alignment learning. Moreover, to generate high-quality content, a multitude of strong-performing methods have been proposed, such as Transformer [82, 101, 17, 24], GANs [53, 7, 93, 110], VAEs [81, 67], Flow models [73, 6] and the current state-of-the-art diffusion models [31, 64, 57, 22, 68]. Especially, the diffusion-based methods have recently delivered remarkable performance in a plethora of cross-modal generation tasks, such as DALL-E [66], Stable Diffusion [68]. While all previous efforts of cross-modal learning are limited to the comprehension of multimodal inputs only, CoDi [78] lately presents groundbreaking development. Leveraging the power of diffusion models, CoDi possesses the ability to generate any combination of output modalities, including language, images, videos, or audio, from any combination of input modalities in parallel. Regrettably, CoDi might still fall short of achieving human-like deep reasoning of input content, with only parallel cross-modal feeding&generation. Cross-modal Understanding and Generation Multimodal Large Language Models LLMs have already made profound impacts and revolutions on the entire AI community and beyond. The most notable LLMs, i.e., OpenAI’s ChatGPT [59] and GPT4 [60], with alignment techniques such as instruction tuning [61, 47, 104, 52] and reinforcement learning from human feedback (RLHF) [75], have demonstrated remarkable language understanding and reasoning abilities. And a series of open-source LLMs, e.g., Flan-T5 [13], Vicuna [12], LLaMA [80] and Alpaca [79], have greatly spurred advancement and made contributions to the community [109, 100]. Afterward, significant efforts have been made to construct LLMs dealing with multimodal inputs and tasks, leading to the development of MM-LLMs. Multimodal Large Language Models On the one hand, most of the researchers build fundamental MM-LLMs by aligning the well-trained encoders of various modalities to the textual feature space of LLMs, so as to let LLMs perceive other modal inputs [35, 109, 76, 40]. For example, Flamingo [1] uses a cross-attention layer to connect a frozen image encoder to the LLMs. BLIP-2 [43] employs a Q-Former to translate the input image queries to the LLMs. LLaVA [52] employs a simple projection scheme to connect image features into the word embedding space. There are also various similar practices for building MM-LLMs that are able to understand videos (e.g., Video-Chat [44] and Video-LLaMA [103]), audios (e.g., SpeechGPT [102]), etc. Profoundly, PandaGPT [77] achieves a comprehensive understanding of six different modalities simultaneously by integrating the multimodal encoder, i.e., ImageBind [25]. Nevertheless, these MM-LLMs all are subject to the limitation of only perceiving multimodal data, without generating content in arbitrary modalities. To achieve LLMs with both multimodal input and output, some thus explore employing LLMs as decision-makers, and utilizing existing off-theshelf multimodal encoders and decoders as tools to execute multimodal input and output, such as Visual-ChatGPT [88], HuggingGPT [72], and AudioGPT [34]. As aforementioned, passing messages between modules with pure texts (i.e., LLM textual instruction) under the discrete pipeline scheme will inevitably introduce noises. Also lacking comprehensive tuning across the whole system significantly limits the efficacy of semantics understanding. Our work takes the mutual benefits of both the above two types, i.e., learning an any-to-any MM-LLM in an end-to-end manner. This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. available on arxiv

Stories, News & Mentions about CSX

Csx’s Stories on HackerNoon

Csx's News & Mentions