<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Jerry Kurian on Medium]]></title>
        <description><![CDATA[Stories by Jerry Kurian on Medium]]></description>
        <link>https://medium.com/@jerrykurian?source=rss-fda4240beb4b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*jKSGvmFxHmncyzPYJPUIvg.jpeg</url>
            <title>Stories by Jerry Kurian on Medium</title>
            <link>https://medium.com/@jerrykurian?source=rss-fda4240beb4b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 15:11:34 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@jerrykurian/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Voice Revolution: How AI is Transforming Human-Technology Interactions]]></title>
            <link>https://medium.com/@jerrykurian/the-voice-revolution-how-ai-is-transforming-human-technology-interactions-0e7d482ed7d6?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/0e7d482ed7d6</guid>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[voiceassitant]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Tue, 25 Mar 2025 09:36:43 GMT</pubDate>
            <atom:updated>2025-03-25T09:36:43.326Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CkDV_ukzAUFuecarVN2NwQ.jpeg" /></figure><p>The voice AI market is poised to grow multifold. As consumers become more comfortable with “conversing” with technology, their expectations from every technological interaction will keep evolving. Over the years, visual interface and touch have been the primary ways in which users interacted with systems, but with the emergence of AI all around us and people getting used to “prompting” to get things done, their expectations from interacting with technology are bound to evolve.</p><p>Voice is a fundamental way of human interaction with the world, and it has very few technical barriers. The lack of mature technology has kept voice from being the natural way of technical interaction, but this is changing. Recent advances in AI mean the capability of computer systems to understand natural language has evolved. Advances in voice synthesis technologies have also made the generated voice very natural, thus enhancing the acceptability of voice-based interaction.</p><p>As per market research, the voice AI market is growing at a CAGR of nearly 20%.</p><p><strong>Industries Impacted:</strong></p><p>Voice AI has already matured in the customer support industry and is being adopted for various use cases. Some of the use cases are as follows.<br><strong>Customer Service</strong>:<br>1. Interactive voice response for handling incoming customer calls. This is one of the most straightforward use cases that relies on simple voice commands for decision-making<br>2. Handling incoming customer calls in the front office and holding a conversation to help complete a transaction like order booking. The US QSR industry has been at the forefront of adopting voice AI technology in this manner.</p><p>Sales and marketing<br>1. Helping with meeting transcripts is another excellent use case where voice AI plays a significant role. Getting AI to listen to your meetings or sales calls can help determine the right actions.<br>2. Lead generation calls that dial potential leads and ask them screening questions before passing them on to the sales team is another way voice AI is helping the marketing team<br>3. Deepfakes and audio/video generation using various AI tools are now the marketing industry&#39;s default practices.</p><p>Technical Landscape:</p><p>There has been significant progress in Audio speech recognition (ASR) and voice AI&#39;s text-to-speech (TTS) aspects. Technology companies have been investing either horizontally (creating technologies that span across industries, typically core tech) or vertically (specializing in developing solutions for a particular industry or domain)</p><p>Building a high-quality ASR has been a significant challenge in the industry&#39;s growth. Companies like Google and Microsoft took the lead here, and their ASRs were the earliest solutions that relied on them. But over the years, with the advancements of AI, many players can provide very customized solutions that are much more effective. Companies like SoundHound, Deepgram, etc, have done an excellent job with it, and their solutions can be leveraged out of the box. OpenAI’s whisper AI is another perfect option for something that can work out of the box.<br>Many companies prefer to work with companies like Deepgram, as they provide the option to train custom models based on voice samples. This can be very useful for scenarios where the voice application could be in noisy areas. Sometimes, the best option is to just train for a location and its ambient noise, as that can work better than a generic ASR solution.</p><p>From a solution perspective, some companies take a vertical approach. Some of the leading names here are Conversenow.ai and HiAuto, which specialize in servicing the QSR industry. The QSR industry has been quick in adopting voice AI due to the nature of this industry, which sees very high attrition and low margins. Training an AI is a much better option than training human operators. However, the need of the QSR industry to take orders makes the solution very complicated. Voice AI combined with the ability to understand specific menus and their variations means that a focused approach could yield better outcomes. The challenge here is not just to understand what the user is saying but to respond so that the interaction can be completed by placing an order. The problem is even more severe in drive-through situations, which have noisy stations and hardware complexities. Voice AI players in this domain don’t just have to solve the technical complexities of ASR and TTS but also manage the menu and integrate it with various POS systems.</p><p>Conclusion:</p><p>Voice interaction technologies like Alexa and Siri have existed for over a decade. However, combining technical advancements and increasing user comfort with natural interactions is now driving the proper adoption of voice AI. The emergence of large language models (LLMs) has significantly enhanced AI’s ability to understand and process natural language, enabling the development of sophisticated, transactional voice-based solutions. As users become more accustomed to conversing with AI systems like ChatGPT and Claude, the demand for voice AI across various interaction channels will continue to grow.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0e7d482ed7d6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Streaming responses from OpenAI using Java]]></title>
            <link>https://blog.devgenius.io/streaming-responses-from-openai-using-java-1a05ce2b15d7?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/1a05ce2b15d7</guid>
            <category><![CDATA[java]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[stream]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Tue, 28 Jan 2025 08:35:08 GMT</pubDate>
            <atom:updated>2025-02-07T15:28:19.522Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IBpFEyKFhrOpunOqPZQf4g.jpeg" /></figure><p>Continuing with developing the<a href="https://medium.com/@jerrykurian/how-to-use-function-calling-with-openai-and-java-5af27ba66a41"> Java library</a> for OpenAI, I will share how you can handle OpenAI chat completion response as a stream in this post.</p><p>OpenAI supports streaming through the standard Chat Completion API by passing a property “stream: true”</p><pre>curl https://api.openai.com/v1/chat/completions \ -H &quot;Content-Type: application/json&quot; -H &quot;Authorization: Bearer $OPENAI_API_KEY&quot; \<br>  -d &#39;{<br>    &quot;model&quot;: &quot;gpt-4o&quot;,<br>    &quot;messages&quot;: [<br>      {<br>        &quot;role&quot;: &quot;developer&quot;, &quot;content&quot;: &quot;You are a helpful assistant.&quot;<br>      },<br>      {<br>        &quot;role&quot;: &quot;user&quot;, &quot;content&quot;: &quot;Hello!&quot;<br>      }<br>    ],<br>    &quot;stream&quot;: true<br>  }&#39;</pre><h4>Calling the chat function for streaming</h4><p>When using the Java library to call the chat completion API, the ChatRequest class contains a <em>stream </em>property<em>.</em> Setting that to true will enable responses to be streamed back.</p><pre>public class ChatRequest {<br>    @JsonProperty(&quot;stream&quot;)<br>    private Boolean stream;</pre><p>To handle the streaming response, the Java library supports a special method for initiating the request. The <em>com.genaipeople.openai.Chat </em>class has a method called <em>stream </em>that responds with a <em>Flow.Publisher</em> object that can be used to handle the streaming response.</p><pre> public Flow.Publisher&lt;ChatResponse&gt; stream(ChatRequest request) {<br>        request.setStream(true);</pre><p>The following test case demonstrates the use of the stream API</p><pre>void testStreamingImageChatCompletionResponse() throws JsonProcessingException, ExecutionException, InterruptedException {<br>        List&lt;Message&gt; messages = new ArrayList&lt;&gt;();<br>        messages.add(new Message(&quot;You are a helpful assistant.&quot;, Role.system));<br>        messages.add(new Message(&quot;What&#39;s in this image?&quot;, <br>                &quot;https://miro.medium.com/v2/resize:fit:828/format:webp/1*LLxq7oaQj8dmPW3wKZTMJA.jpeg&quot;, <br>                Role.user));<br>        ChatRequest request = new ChatRequest(messages, MODEL);<br><br>        Flow.Publisher&lt;ChatResponse&gt; stream = chat.stream(request);<br>        Object lock = new Object();<br>        stream.subscribe(new Flow.Subscriber&lt;ChatResponse&gt;() {<br>            @Override<br>            public void onSubscribe(Flow.Subscription subscription) {<br>                subscription.request(Long.MAX_VALUE);<br>                System.out.println(&quot;Subscription requested&quot;);<br>            }<br><br>            @Override<br>            public void onNext(ChatResponse item) {<br>                Choice choice = item.getChoices().get(0);   <br>                Delta delta = choice.getDelta();<br>                if (choice.getFinishReason() == null) {<br>                    System.out.println(&quot;Item: &quot; + delta.getContent().getContent());<br>                }<br>            }<br><br>            @Override<br>            public void onError(Throwable throwable) {<br>                throwable.printStackTrace();<br>                fail(&quot;Stream failed in Test&quot;);<br>            }<br><br>            @Override<br>            public void onComplete() {<br>                System.out.println(&quot;Stream completed in Test&quot;);<br>            }<br>        });</pre><h3>Sample Application</h3><p>Here is a link to a sample application that uses the library and its streaming feature to build a sample application: <a href="https://github.com/GenAI-People/java-ai-client">https://github.com/GenAI-People/java-ai-client</a></p><p>The library code can be cloned from <a href="https://github.com/GenAI-People/openai-java">https://github.com/GenAI-People/openai-java</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1a05ce2b15d7" width="1" height="1" alt=""><hr><p><a href="https://blog.devgenius.io/streaming-responses-from-openai-using-java-1a05ce2b15d7">Streaming responses from OpenAI using Java</a> was originally published in <a href="https://blog.devgenius.io/">Dev Genius</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to use function calling with OpenAI and Java]]></title>
            <link>https://blog.devgenius.io/how-to-use-function-calling-with-openai-and-java-5af27ba66a41?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/5af27ba66a41</guid>
            <category><![CDATA[java]]></category>
            <category><![CDATA[llm-function-calling]]></category>
            <category><![CDATA[openai]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Fri, 24 Jan 2025 03:23:06 GMT</pubDate>
            <atom:updated>2025-02-07T15:36:23.453Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LGMAKyLyAV04XHcepeVbeA.jpeg" /></figure><p>Continuing to build on the <a href="https://medium.com/gopenai/chatting-about-an-image-using-java-and-openai-42d64a8d4a7b">Java library</a> I have been creating for OpenAI APIs, I have now added support for Function Calling.</p><p>Function calling allows you to use LLM purely for natural language understanding while you can get your application code to act on the request, fetch data, and respond or take any action the user has asked for.</p><p>A function is a tool that OpenAI supports that can assist in taking actions that LLM would like the host application to perform.</p><pre>curl https://api.openai.com/v1/chat/completions \ -H &quot;Content-Type: application/json&quot; \ -H &quot;Authorization: Bearer $OPENAI_API_KEY&quot; \<br>-d &#39;{<br>  &quot;model&quot;: &quot;gpt-4o&quot;,<br>  &quot;messages&quot;: [<br>    {<br>      &quot;role&quot;: &quot;user&quot;,<br>      &quot;content&quot;: &quot;What&#39;\&#39;&#39;s the weather like in Boston today?&quot;<br>    }<br>  ],<br>  &quot;tools&quot;: [<br>    {<br>      &quot;type&quot;: &quot;function&quot;,<br>      &quot;function&quot;: {<br>        &quot;name&quot;: &quot;get_current_weather&quot;,<br>        &quot;description&quot;: &quot;Get the current weather in a given location&quot;,<br>        &quot;parameters&quot;: {<br>          &quot;type&quot;: &quot;object&quot;,<br>          &quot;properties&quot;: {<br>            &quot;location&quot;: {<br>              &quot;type&quot;: &quot;string&quot;,<br>              &quot;description&quot;: &quot;The city and state, e.g. San Francisco, CA&quot;<br>            }          },<br>          &quot;required&quot;: [&quot;location&quot;]<br>        }<br>      }<br>    }<br>  ],<br>  &quot;tool_choice&quot;: &quot;auto&quot;<br>}&#39;</pre><h4>Passing Function parameters</h4><p>I aimed to make the Java library follow as much of Java standards as possible. In the case of a function, the API expects the type of parameters to be passed in. I tried to make it work by allowing the developer to define this as Java generic types, but I came across some limitations to make this happen. Could someone suggest a way to overcome the limitations?</p><p>The class that represents a function is as follows. It’s a tool that takes parameters. The API requires multiple parameters to be passed in, and the ObjectType represents an object whose properties are the parameters.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_DJlKpM7Cj3tnAizDkyruQ.png" /></figure><h4>Passing Functions</h4><p>To use “function calling” using the completion API, a ChatRequest needs to be created with the set of messages and tools added.</p><p>The tools can be added to the ChatRequest in multiple ways. Suppose you want to pass a function <em>get_current_weather </em>with its user message and tools then it can be done as follows.</p><pre>List&lt;Message&gt; messages = Arrays.asList(new Message(&quot;What&#39;\\&#39;&#39;s the weather like in Boston today?&quot;, Role.user));<br>ChatRequest request = new ChatRequest(messages, MODEL);<br>Function function = new Function(&quot;get_current_weather&quot;, &quot;Get the current weather in a given location&quot;);<br>function.addParameter(String.class, &quot;location&quot;, &quot;The city and state, e.g. San Francisco, CA&quot;, null, true);<br>function.addParameter(String.class, &quot;unit&quot;, &quot;The unit of temperature to use, either &#39;celsius&#39; or &#39;fahrenheit&#39;&quot;, <br>Arrays.asList(&quot;celsius&quot;, &quot;fahrenheit&quot;), false);<br>request.setTools(Arrays.asList(function));<br>request.setToolChoice(ToolChoice.auto.toString());<br><br>CompletableFuture&lt;ChatResponse&gt; futureResponse = chat.complete(request);<br>ChatResponse actualResponse = futureResponse.get();</pre><p>As a Java developer, I would love to avoid passing in the types as Java is a typed language. If you wish to have the types inferred based on the type of the argument, then you can use the following approach</p><pre>Function function = new Function(&quot;get_product_recommendations&quot;, <br>        &quot;Searches for products matching certain criteria in the database&quot;);<br>    <br>// Add categories array parameter<br>function.addParameter(new Object() {<br>    @PropertyDetails(description = &quot;categories that could be a match&quot;,<br>        enumValues = {&quot;coats &amp; jackets&quot;, &quot;accessories&quot;, &quot;tops&quot;, <br>                    &quot;jeans &amp; trousers&quot;, &quot;skirts &amp; dresses&quot;, &quot;footwear&quot;})<br>    public List&lt;String&gt; categories;<br>    <br>    @PropertyDetails(description = &quot;The color of the product to be recommended&quot;,<br>        enumValues = {&quot;black&quot;, &quot;white&quot;, &quot;brown&quot;, &quot;red&quot;, &quot;blue&quot;, &quot;green&quot;,<br>                    &quot;orange&quot;, &quot;yellow&quot;, &quot;pink&quot;, &quot;gold&quot;, &quot;silver&quot;})<br>    public List&lt;String&gt; colors;<br>    <br>    @PropertyDetails(description = &quot;keywords that should be present in the item title or description&quot;)<br>    public List&lt;String&gt; keywords;<br><br>    @PropertyDetails(description = &quot;price range for the products to be recommended&quot;)<br>    public PriceRange price_range;<br>    <br>    @PropertyDetails(description = &quot;The maximum number of products to return, use 5 by default if nothing is specified by the user&quot;)<br>    public Integer limit;<br>}.getClass(), &quot;parameters&quot;, null, null, true);</pre><p>The anonymous Object class and its properties allow you to set the function parameters much more Java-like. You can also pass the details to prompt using the <em>PropertyDetails </em>annotation.</p><p>You can also create a separate class and pass its instance if you do not wish to use an anonymous class.</p><pre>class PriceRange {<br>    @PropertyDetails(description = &quot;The minimum price to search for&quot;)<br>    private Double min;<br>    @PropertyDetails(description = &quot;The maximum price to search for&quot;)<br>    private Double max;<br>    @PropertyDetails(description = &quot;The currencies to search for&quot;, <br>        enumValues = {&quot;USD&quot;, &quot;EUR&quot;, &quot;GBP&quot;, &quot;JPY&quot;, &quot;AUD&quot;, &quot;CAD&quot;, &quot;CHF&quot;, &quot;NZD&quot;, &quot;SEK&quot;, &quot;NOK&quot;, &quot;DKK&quot;, &quot;PLN&quot;, &quot;CZK&quot;, &quot;HUF&quot;, &quot;RON&quot;, &quot;BGN&quot;, &quot;HRK&quot;, &quot;RUB&quot;, &quot;TRY&quot;, &quot;ZAR&quot;, &quot;BRL&quot;, &quot;MXN&quot;, &quot;INR&quot;, &quot;IDR&quot;, &quot;THB&quot;, &quot;MYR&quot;, &quot;PHP&quot;, &quot;SGD&quot;, &quot;HKD&quot;, &quot;KRW&quot;, &quot;TWD&quot;, &quot;MYR&quot;, &quot;PHP&quot;, &quot;SGD&quot;, &quot;HKD&quot;, &quot;KRW&quot;, &quot;TWD&quot;})<br>    private List&lt;String&gt; currencies;<br><br>    @PropertyDetails(description = &quot;The items to return&quot;)<br>    private List&lt;Item&gt; products;<br>}</pre><h4>Handling Response With Functions</h4><p>The chat completion API will respond with the function that needs to be invoked along with the arguments to the function. This is inferred from the prompt and returned. The response handler must identify the function to be called and then invoke it.</p><pre>CompletableFuture&lt;ChatResponse&gt; responseFuture = conversation.sendMessage(message);<br>  responseFuture.thenAccept((response) -&gt; {<br>      if(response != null) {<br>          for(Choice choice : response.getChoices()) {<br>              if(choice.getFinishReason().equals(FinishReason.tool_calls.toString())) {<br>                  for(ToolCall toolCall : choice.getMessage().getToolCalls()) {<br>                      handleFunctionCall(toolCall.getFunction());<br>                  }<br>              }<br>          }<br>      }<br>}).exceptionally((exception) -&gt; {<br>    return null;<br>});<br><br>public void handleFunctionCall(Function function) {<br>    switch(function.getName()) {<br>        case &quot;lookup_doctor&quot;:<br>            Map&lt;String, Object&gt; doctorDetails = (Map&lt;String, Object&gt;) function.getParameters().get(&quot;doctorDetails&quot;);<br>            handleLookupDoctor(doctorDetails);<br>            break;<br>        case &quot;schedule_appointment&quot;:<br>            Map&lt;String, Object&gt; appointmentDetails = (Map&lt;String, Object&gt;) function.getParameters().get(&quot;appointmentDetails&quot;);<br>            handleScheduleAppointment(appointmentDetails);<br>            break;<br>        default:<br>            System.out.println(&quot;Please let me know how I can help you&quot;);<br>            break;<br>    }<br>}</pre><h4>Sample Application</h4><p>Here is a link to a sample application that uses the library and its function-calling feature to build an appointment booking application: <a href="https://github.com/GenAI-People/java-ai-client">https://github.com/GenAI-People/java-ai-client</a></p><p>The library code can be cloned from <a href="https://github.com/GenAI-People/openai-java">https://github.com/GenAI-People/openai-java</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5af27ba66a41" width="1" height="1" alt=""><hr><p><a href="https://blog.devgenius.io/how-to-use-function-calling-with-openai-and-java-5af27ba66a41">How to use function calling with OpenAI and Java</a> was originally published in <a href="https://blog.devgenius.io/">Dev Genius</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Chatting about an Image using Java and OpenAI]]></title>
            <link>https://blog.gopenai.com/chatting-about-an-image-using-java-and-openai-42d64a8d4a7b?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/42d64a8d4a7b</guid>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[java]]></category>
            <category><![CDATA[genai]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Thu, 03 Oct 2024 08:01:44 GMT</pubDate>
            <atom:updated>2024-10-07T10:12:26.512Z</atom:updated>
            <content:encoded><![CDATA[<p>Have you ever wanted to enable your users to pass in an image as input and then have your app take action based on that? OpenAI enables you to do this using the completion API.</p><p>The chat completion API allows you to pass an image URL as an input and an accompanying text prompt in which you can ask for details about the image and get the details back as a response.</p><pre>The API<br>curl https://api.openai.com/v1/chat/completions</pre><pre>Passing your input messages to the API<br>&quot;messages&quot;: [<br>      {<br>        &quot;role&quot;: &quot;user&quot;,<br>        &quot;content&quot;: [<br>          {<br>            &quot;type&quot;: &quot;text&quot;,<br>            &quot;text&quot;: &quot;What&#39;\&#39;&#39;s in this image?&quot;<br>          },<br>          {<br>            &quot;type&quot;: &quot;image_url&quot;,<br>            &quot;image_url&quot;: {<br>              &quot;url&quot;: &quot;https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg&quot;<br>            }<br>          }<br>        ]<br>      }<br>    ]</pre><pre>The response from OpenAI<br>{<br>  &quot;choices&quot;: [{<br>    &quot;index&quot;: 0,<br>    &quot;message&quot;: {<br>      &quot;role&quot;: &quot;assistant&quot;,<br>      &quot;content&quot;: &quot;\n\nThis image shows a wooden boardwalk extending through a lush green marshland.&quot;,<br>    },<br>    &quot;logprobs&quot;: null,<br>    &quot;finish_reason&quot;: &quot;stop&quot;<br>  }]<br>}</pre><p>The above code snippets show the API and its usage.</p><h4>Using Java to call the API</h4><p>As discussed in my earlier <a href="https://medium.com/@jerrykurian/building-a-java-library-for-openai-apis-bca43cc2ed01">blog</a>, I am developing a Java-based client library to support OpenAI APIs. I will gradually add support for all the APIs and recently added support for getting responses about images.</p><p>The <em>content</em> field of the <em>messages </em>in OpenAI can either contain a simple String message or an array of objects, depending upon whether you wish to do a text or image completion.</p><p>Since both use the same API, in my Java library, I have chosen to make it easy for the API caller by providing two different Message constructors</p><pre>    public Message(String content, Role role) {<br>        this.role = role;<br>        this.content = new TextContent(content);<br>    }<br><br>    public Message(String content, String imageUrl, Role role) {<br>        this.role = role;<br>        this.content = new ImageContent(content, imageUrl);<br>    }</pre><p>The library represents a text-based prompt as a <em>TextContent</em> object, while an image prompt is an <em>ImageContent </em>object. A custom serializer emits the appropriate content field when converting the message to JSON.</p><p>When receiving the JSON response, a custom deserializer converts it to the appropriate content object. However, typically, the response only contains string content.</p><h4>Using the library for image use cases</h4><p>As the library uses constructors of the <em>Message</em> class to differentiate the type of content, it is pretty straightforward to build out a use case that requires understanding image details to decide on further actions.</p><p>An interesting use case could be in a customer support scenario where you could allow customers to share the image of a product with an issue and ask questions about it.</p><p>Maybe your customer ordered two <em>samosas</em> and got only one!! Just ask them to take a snap and send it to a support bot with the complaint. The bot should be able to validate the count using the completion API and then pass on the issue to a human agent to take action.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DRU9es131QDZz-1zfNns_w.jpeg" /></figure><p>You can implement more such use cases in Java using the library. A sample code is as below</p><pre>public void seekInput() {<br>        System.out.println(&quot;Do you want to send a message that includes image (y/n default:n)? (text/image)&quot;);<br>        String messageType = System.console().readLine().toLowerCase();<br>        <br>        String content;<br>        Message userMessage;<br>        <br>        if (messageType.equalsIgnoreCase(&quot;y&quot;)) {<br>            System.out.println(&quot;Enter the image URL:&quot;);<br>            String imageUrl = System.console().readLine();<br>            System.out.println(&quot;Enter your message text:&quot;);<br>            content = System.console().readLine();<br>            userMessage = new Message(content, imageUrl, Role.USER);<br>        } else {<br>            System.out.println(&quot;Enter your message:&quot;);<br>            content = System.console().readLine();<br>            userMessage = new Message(content, Role.USER);<br>        }<br>        <br>        messages.add(userMessage);<br>    }</pre><p>You can access the latest version of the library from the GitHub repo <a href="https://github.com/GenAI-People/openai-java">https://github.com/GenAI-People/openai-java</a></p><p>An accompanying project that demonstrates how to use the library is available at <a href="https://github.com/GenAI-People/java-ai-client">https://github.com/GenAI-People/java-ai-client</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=42d64a8d4a7b" width="1" height="1" alt=""><hr><p><a href="https://blog.gopenai.com/chatting-about-an-image-using-java-and-openai-42d64a8d4a7b">Chatting about an Image using Java and OpenAI</a> was originally published in <a href="https://blog.gopenai.com/">GoPenAI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building a Java Library for OpenAI APIs]]></title>
            <link>https://blog.gopenai.com/building-a-java-library-for-openai-apis-bca43cc2ed01?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/bca43cc2ed01</guid>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[java]]></category>
            <category><![CDATA[openai]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Fri, 20 Sep 2024 11:07:52 GMT</pubDate>
            <atom:updated>2025-01-30T12:46:23.080Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LLxq7oaQj8dmPW3wKZTMJA.jpeg" /></figure><p>Java is one of the most popular programming languages, and there are nearly 17 million Java developers worldwide, second only to the number of Javascript developers.</p><p>Java-based implementations dominate enterprise software development. The JVM has evolved to be one of the most sophisticated pieces of software, powering much of the business worldwide.</p><p>For some reason, though, Java has been losing out on being part of AI software development. Python has dominated this space for various reasons.</p><p>With the emergence of Generative AI and the shift towards large language models (LLM) and its API-driven approach, Java has a great chance of retaining its prominence in software stacks of large AI projects, especially in the enterprise world. This would require Java developers to start becoming familiar with the world of GenAI and acquiring proficiency in developing GenAI Applications.</p><p>Unfortunately, there isn&#39;t enough literature to help Java developers. Even OpenAI, the most popular LLM, lists only one Java-based library for working with their APIs. That <a href="https://github.com/TheoKanning/openai-java">library</a> is also archived as the APIs have changed, and it is no longer compliant with the latest APIs.</p><p>I have started developing a library supporting the latest APIs to address the situation. I will document my progress in a series of blogs, and I hope I will be able to add support for all the APIs soon enough. I hope the library will help Java developers start developing their GenAI ideas faster.</p><h4>Open AI APIs</h4><p>OpenAI has some of the most advanced APIs for working with LLM and is well-documented. It tends to evolve pretty fast and become obsolete in the process, but that is probably to be expected from a rapidly evolving technology.</p><p>The APIs can be referenced on the OpenAI website at <a href="https://platform.openai.com/docs/api-reference/introduction">https://platform.openai.com/docs/api-reference/introduction</a></p><p>The various APIs that OpenAI supports are</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/966/1*6mUP0d4T5e0OT1eDVDHaQA.png" /><figcaption>Various APIs supported by OpenAI</figcaption></figure><h4>Java Library</h4><p>The Java library will support all of the above APIs.</p><p>I have added support for the Chat Completion API. This API responds to the completion of a prompt and can help you achieve what ChatGPT does in its simplest form.</p><p>To start using the library, clone the repo at <a href="https://github.com/GenAI-People/openai-java">https://github.com/GenAI-People/openai-java</a>.</p><p>The project currently supports the Maven packaging system. I have not yet added this to any central repository, so install it in your local Maven repository, and you should be ready.</p><h4>Java APIs</h4><p>Completion API support is provided through the class Chat.java in the package com.genaipeople. open. This class&#39;s object can be created by passing in the API key that you can obtain from the OpenAI <a href="https://platform.openai.com/settings/profile?tab=api-keys">website</a>.</p><p>The completion API is asynchronous, and you need to handle the ComplableFuture object it returns.</p><p>An example of usage is below</p><ul><li>Create a chat instance</li></ul><blockquote>private final Chat chat = new Chat(“api-key”);</blockquote><ul><li>Create Message and ChatRequest</li></ul><blockquote>Message userMessage = new Message(content, Role.USER);<br>messages.add(userMessage);<br>ChatRequest request = new ChatRequest(messages, “apiKey”);<br>request.setModel(“gpt-4o”);</blockquote><ul><li>Call the <em>complete </em>method and pass the request</li></ul><blockquote>return chat.complete(request).thenAccept((response) -&gt; {<br> Message aiMessage = response.getChoices().get(0).getMessage();<br> messages.add(aiMessage);<br>}).exceptionally((exception) -&gt; {<br> System.err.println(“Error sending message: “ + exception.getMessage());<br> return null;<br>});</blockquote><ul><li>Handle the response asynchronously and deal with the choices</li></ul><h4>Sample Client Application</h4><p>I have added unit test cases to the project to try out the APIs. I have also created a sample Java application where I will use the various APIs and implement use cases that demonstrate the usage. This sample application is available in the repository <a href="https://github.com/GenAI-People/java-ai-client">https://github.com/GenAI-People/java-ai-client</a></p><h3>What next?</h3><p>I will gradually add support for the rest of the APIs and share their details and usage as I progress. So stay tuned</p><p>Image Chat : <a href="https://medium.com/gopenai/chatting-about-an-image-using-java-and-openai-42d64a8d4a7b">https://medium.com/gopenai/chatting-about-an-image-using-java-and-openai-42d64a8d4a7b</a></p><p>Function Calling : <a href="https://medium.com/@jerrykurian/how-to-use-function-calling-with-openai-and-java-5af27ba66a41">https://medium.com/@jerrykurian/how-to-use-function-calling-with-openai-and-java-5af27ba66a41</a></p><p>Streaming Response: <a href="https://medium.com/@jerrykurian/streaming-responses-from-openai-using-java-1a05ce2b15d7">https://medium.com/@jerrykurian/streaming-responses-from-openai-using-java-1a05ce2b15d7</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bca43cc2ed01" width="1" height="1" alt=""><hr><p><a href="https://blog.gopenai.com/building-a-java-library-for-openai-apis-bca43cc2ed01">Building a Java Library for OpenAI APIs</a> was originally published in <a href="https://blog.gopenai.com/">GoPenAI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Will code generation tools replace programmers?]]></title>
            <link>https://medium.com/@jerrykurian/will-code-generation-tools-replace-programmers-71d97bd7b544?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/71d97bd7b544</guid>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[co-pilot]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Mon, 15 Apr 2024 09:14:38 GMT</pubDate>
            <atom:updated>2024-04-15T10:02:50.601Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*otKerK-2OTQ7IHrooiRAxQ.jpeg" /></figure><p>I know that everyone in the IT industry has been talking about code generation tools and how they will impact software engineering jobs, going forward. The demo showcased by Devin team has raised curiosity even further.<br>I have been experimenting with code generation with the sole aim of making sure all my coding happens through code generation. Having spent a couple of months on it, here I am, sharing my experience. The good, and bad, and what I expect from it in the future.</p><h4>What is GitHub Copilot?</h4><p>GitHub Co-pilot (<a href="https://github.com/features/copilot">https://github.com/features/copilot</a>) is the code generation assistant that Microsoft launched. It seamlessly integrates with your IDE and acts as a peer programmer. A co-pilot by definition is someone who takes instructions and gets the job done. The main pilot (or in this case you as a coder) is the one who is in charge and would be ultimately held responsible for anything that goes to production !!</p><h4>What is Good</h4><p>Once you have gone past the initial learning curve, you will realize that the co-pilot has the potential to make you extremely productive. The learning curve isn’t so much concerning the features of the tool, but it is more about overcoming your instinct to code.</p><p>In the beginning, it was quite hard for me to not code myself. But then, once I made up my mind that I would not code but generate, I started seeing the benefits.</p><p>The greatest productivity boost I saw was in the UI code creation. In my project I used the Python-based Jinja framework and the kind of UI code it generated was of really high quality. I could just imagine a component and instruct and more often than not the co-pilot got it right within 2–3 attempts. I can safely say that what could have taken me an hour got done in 5 minutes, that&#39;s a massive saving !!</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ffast.wistia.net%2Fembed%2Fiframe%2Fftb4mrl5jl&amp;display_name=Wistia%2C+Inc.&amp;url=https%3A%2F%2Fgenaipeople.wistia.com%2Fmedias%2Fftb4mrl5jl&amp;image=https%3A%2F%2Fembed-ssl.wistia.com%2Fdeliveries%2Fe8952fdf0ef71b39778ed55b6fec36c1e37d1058.jpg%3Fimage_crop_resized%3D960x540&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=wistia" width="960" height="540" frameborder="0" scrolling="no"><a href="https://medium.com/media/9ab6829a80446bf25077a0ff32478c98/href">https://medium.com/media/9ab6829a80446bf25077a0ff32478c98/href</a></iframe><p>As I am getting more into the code generation mindset, I am truly starting to enjoy it.</p><h4>What is not good</h4><p>I was quite disappointed with GitHub Co-Pilot’s ability to keep things in context. There were many cases where I generated a piece of code in one file and wanted the co-pilot to generate related code in some other file. But I had to give it context all over again to get it done. I was expecting the co-pilot to understand as the files were all open. But then it required a bit of cajoling to get things done.</p><p>Overall, I was happy with the quality of the code generated, but the quality of unit test cases generated left a lot to be desired. I found it reluctant to generate deeper levels of test cases. This could again be related to its inability to keep the overall context in mind. Test case generation is one of the key use cases that everyone expects the co-pilot to help with, but it&#39;s not up to the mark yet.</p><h4>What do I expect in the future?</h4><p>I certainly expect the co-pilot to be much better in understanding the context and its ability to generate test cases has to improve. But I think both of this has to do with the fact that go-pilot is too focused on coding. It needs to align with the fact that code generation is all about prompting and prompting is all about context.<br>I expect the code generation tools to better integrate with documentation tools. In the case of the GitHub co-pilot, there is no reason why the co-pilot can’t be integrated with the GitHub project management and documentation tools to understand the context. The code generated would be so much better if the co-pilot kept the overall requirements in mind.</p><h4>What should you be doing?</h4><p>As the productivity benefits of GitHub Copilot become more widely recognized CEOs will start asking their engineering teams to start using GitHub Copilot, leading to a demand for folks who can work with Copilot or other such code generation tools. <br>The key skill going forward would be understanding customer needs, documenting them clearly, and then when it comes to coding, using the code generation tools to create things with amazing productivity. I am glad that the people I am <a href="https://learning.genaipeople.com/">mentoring</a> are realizing this and working on building products in the most productive manner using the code generation tools.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=71d97bd7b544" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Does a prompt engineer need to code?]]></title>
            <link>https://medium.com/@jerrykurian/does-a-prompt-engineer-need-to-code-f61a5c13f9c1?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/f61a5c13f9c1</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <category><![CDATA[gpt]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[coding]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Tue, 27 Jun 2023 06:39:20 GMT</pubDate>
            <atom:updated>2023-06-27T06:39:20.198Z</atom:updated>
            <content:encoded><![CDATA[<p>Prompt engineering is the hottest buzzword around and everyone is curious about what it is and how to become a prompt engineer.</p><p>Now the term engineer in it also brings out various questions about what skillsets are required to become a prompt engineer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0Tl3yqXapFLvUtXB-V-Q9Q.jpeg" /></figure><h4>What does prompt engineering involve?</h4><p>I have been using the various LLM tools and playing around with them and from what I understand, prompt engineering involves the following</p><ol><li>Ability to visualize the end goal. This of course means you need to have clarity on what you would like to achieve from the task</li><li>Clarity about the details of your goal. LLMs know a lot and can look at your goal from various perspectives. It&#39;s important to direct it and ensure it does what you really want and not hallucinate</li><li>Knowledge of the model that you are working with. Its important to know the constraints and capabilities of the model to communicate with it effectively. Prompting is all about effective communication after all.</li></ol><p>Going by the above, the most important skill seems to be ability to understand and articulate what you want, within the context of model capabilities. None of that really require an ability to code.</p><p>If you have used chatGPT and got something useful out of it, then congratulations, you are on your way to be a prompt engineer !!</p><h4>Is that all there is to prompting??</h4><p>There are certain tasks that might not be as simple as asking GPT to suggest some health tips (and believing what it says !!). In certain cases you might need to clearly direct the model to produce the desired result. The direction will need to be set through various examples that might involve lots of sample data.</p><p>Refer to my earlier <a href="https://medium.com/p/7b275b472ed2">post</a> where I tried predicting the results of the final match of the soccer world cup. This required me to pass in sample data of past matches in the tournament and predictions vs actuals so that the model understood what I wanted it to do. I also needed to create a custom model. While this might be a simple problem there are going to be much more complex variants of this problem which will require even more complex prompt engineering. In such cases it will require automated mode of collecting, creating and formatting of sample data. This is where coding skills will come into play.</p><p>A lot of heavy lifting will need to be done in the tools that will help create complex prompts. The response generated by the prompts will need automated quality checks to test it against a wide range of inputs.</p><p>While everyone is getting all excited about the news of huge salaries being offered to prompt engineers, I think it will mostly be for people who can help solve complex problems, often times requiring some amount of coding skills.</p><p>Even if not coding directly, then such roles will require the prompt engineer to help guide an automation team.</p><h4>In Conclusion</h4><p>As more and more software tools start adding a prompt based interface, communicating clearly with technology will become as important as communicating with other humans.</p><p>Work on your listening skills to understand requirements clearly and then bring out your empathy and articulation skills to clearly communicate what you need to get done.</p><p>Happy Prompting !!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f61a5c13f9c1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[FIFA World cup final prediction by GPT]]></title>
            <link>https://blog.devgenius.io/fifa-world-cup-final-prediction-by-gpt-7b275b472ed2?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/7b275b472ed2</guid>
            <category><![CDATA[predictions]]></category>
            <category><![CDATA[messi]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[world-cup]]></category>
            <category><![CDATA[gpt-3]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Sun, 18 Dec 2022 04:45:23 GMT</pubDate>
            <atom:updated>2022-12-25T22:25:33.159Z</atom:updated>
            <content:encoded><![CDATA[<p>The final of the world cup is here and I am waiting along with the rest of the world to finally see Messi lift that world cup. Go, Messi !!</p><p>While the anticipation and the unpredictable nature of sports are what makes them so exciting, as humans we love nothing better than predicting the future. I was wondering if GPT could help me with this and use its smarts to predict the results of the final.</p><h3>Is GPT-3 following the world cup?</h3><p>GPT-3 is a language model trained using data from the internet till 2021, so it would probably not even know that a world cup is underway. When I asked it “what are the latest soccer results that you know about”, it listed results from some matches held in 2021. Clearly, it&#39;s not following the world cup.</p><h3>Informing the model</h3><p>For GPT-3 to predict, it will need to be told about what’s going on and also a way to predict results. So following is what I decided to do</p><ol><li>Collect data from the internet that summarise the matches</li><li>Share the predictions for each match and the actual results</li><li>Fine-tune a model with this data</li><li>Hope that the model learns enough to now predict future games</li></ol><h3>Collecting Data</h3><p>For number 1, I had to identify a credible source of data and I chanced upon <a href="https://www.bbc.com/sport/football/63673446">predictions</a> that <a href="https://en.wikipedia.org/wiki/Chris_Sutton">Chris Sutton</a> has been making on BBC. He seems credible enough even if his predictions haven’t been all that accurate. But I wanted GPT to know that prediction is a hard thing and maybe learn from the mistakes that he made. So this data source seemed fine to me.</p><h3>Fine Tuning the model</h3><p>Next was how to fine-tune the model and train it with all the data. For this, I decided to train the <em>davinci</em> model with the following set of prompts and completions</p><blockquote>{“prompt”:”Qatar have to hope that home advantage pays off for them, but I very much doubt it. It’s going to be very difficult for them to get out of this group — in fact, I think they will finish bottom. Ecuador are quite a defensive-minded team so there won’t be many goals in this opening game, but they should still find a way of winning it.\\nTeams: Qatar vs Ecuador\\nPrediction: Qatar:0 Ecuador:1\\nActual Result:\n\n###\n\n”,”completion”:” Qatar:0 Ecuador:2\n”}</blockquote><p>Follow this <a href="https://drive.google.com/file/d/1iPX_j9mHMZ9TOzmdguFLqJt2BTNPOm45/view?usp=share_link">link</a> to get the entire file</p><p>The aim is to feed Chris’ commentary, predictions, and actual results for every match till the semi-finals. Then use the semi-finals to test if the model works. I would then use the same model to predict the 3rd place finish and eventually the final.</p><h3>Testing the model</h3><p>Once the model was trained it was time to test things out. Using the predictions that Chris made for the Semi-Finals, third-place finish, and the final, I ran a few tests.</p><h4>Semi-Finals</h4><p>The model seems to predict the results correctly almost every time. Even if its score keeps changing and of course does not match what really happened. But this is quite promising. I kept the API parameters pretty simple with the temperature set to 0.1</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-Vj3SQ_s24qVU1ZvzICGIw.png" /><figcaption>Predictions for the semi-finals</figcaption></figure><h4>Third-Place finish</h4><p>For the third-place match the model seemed a little uncertain and kept giving different results, but more often than not would show Croatia as the winner. Chris predicted a Moroccan victory. Below is the screenshot from a prediction that I ran moments before the match started</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7STOd_vR6poJKW2s16TKOA.png" /><figcaption>Prediction for the third-placed match</figcaption></figure><h3>The final</h3><p>So what does the model think about the big final? This time I will fine-tune the model with the actual results from the semi-final and the third-placed match so that it has all the info it needs before making the big prediction.</p><p>Chris predicted a French triumph with a 3–1 margin. The model on the other hand</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dxMJAEH18aY6KPCP1sJHqw.png" /><figcaption>Predictions for the final</figcaption></figure><p>AI clearly lacks any emotion and loyalty toward Messi. But I, like millions of other Messi fans, will hope that it will be proven wrong.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7b275b472ed2" width="1" height="1" alt=""><hr><p><a href="https://blog.devgenius.io/fifa-world-cup-final-prediction-by-gpt-7b275b472ed2">FIFA World cup final prediction by GPT</a> was originally published in <a href="https://blog.devgenius.io/">Dev Genius</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[NEP — An opportunity for schools to be re-imagined]]></title>
            <link>https://medium.com/@jerrykurian/nep-an-opportunity-for-schools-to-be-re-imagined-85a71a901201?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/85a71a901201</guid>
            <category><![CDATA[nep]]></category>
            <category><![CDATA[education]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Sat, 17 Dec 2022 04:39:01 GMT</pubDate>
            <atom:updated>2022-12-17T05:48:51.725Z</atom:updated>
            <content:encoded><![CDATA[<h3>NEP — An opportunity for schools to re-imagine themselves</h3><p>Of late I am seeing more awareness among parents about the NEP and it also seems like schools are getting ready to adopt NEP as more and more state governments start mandating it.</p><p>The focus and a general understanding seem to be that this is primarily a structural change in the grades with the current monolithic system broken down into smaller parts. Another issue of interest is the use of the mother tongue as a medium of instruction.</p><p>While the above two points are changes that are easier to understand, I hope the larger objective of moving away from the current system that reduces education to merely intellectual exercise aimed at securing a job is noticed. The aim of the new policy has to be the holistic development of children. The breaking down of grades should not be seen as a mere grouping of grades but as different phases in the growth of a child that needs to be handled uniquely.</p><p>The stated goals of NEP as per the website are as follows. Note point number 4.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t_6WRfFFfuz94SHYhVG1Aw.png" /></figure><p>Unfortunately, I am already seeing advertisements from schools that claim to be NEP-ready. This seems pretty shallow to me. For example, I see schools imagining the first phase of education as children sitting in classes and being made to sing and dance and follow the teacher’s instructions. This sadly is not much different from what happens now.</p><p>Schools need to re-imagine themselves completely if the actual objective has to be met. The holistic growth of a child can not happen in a closed space without any scope for individual expressions and explorations. Teachers have to re-imagine themselves as co-learners and explorers and develop respect for every child’s ability to see things from a unique perspective. Such things can’t happen overnight and the government also has to plan the rollout in a manner that the real aim is achieved and schools do not end up “standardizing” things with the aim of scalability and profits.</p><p>If schools need a template for this then there is nothing better than Mahatma Gandhi’s Nayi Talim. As usual, the great man was way ahead of his time. While all of his ideas are not yet there in the NEP, it&#39;s inevitable that the whole world will adopt it someday. If you wish to learn more about Nayi Talim then do check out <a href="https://www.youtube.com/watch?v=LM8Lyz1usx8">https://www.youtube.com/watch?v=LM8Lyz1usx8</a></p><p>Overall, NEP is a good first step, but it will require schools to get themselves aligned with the needs of children and not merely look at it as a revised curriculum to be sold to parents.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qpGTXrQKY8XkNjuhiIe7EQ.png" /></figure><p>Note: The above image was generated using DALL-E. The AI can’t imagine something beyond the buildings and lawns. I hope humans can do a better job of it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=85a71a901201" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Opposite to courage is conformity]]></title>
            <link>https://medium.com/@jerrykurian/opposite-to-courage-is-conformity-9e4e5f541af?source=rss-fda4240beb4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/9e4e5f541af</guid>
            <category><![CDATA[beyond-limits]]></category>
            <category><![CDATA[cubicle-cultures]]></category>
            <category><![CDATA[strangest-secret]]></category>
            <dc:creator><![CDATA[Jerry Kurian]]></dc:creator>
            <pubDate>Wed, 26 Jun 2019 17:10:31 GMT</pubDate>
            <atom:updated>2019-06-26T17:10:31.187Z</atom:updated>
            <content:encoded><![CDATA[<p>I have been following my mentor’s advice and listening to <a href="https://www.youtube.com/watch?v=EFhkdzj-x80">“the strangest secret”</a> everyday.</p><p>The audio covers many interesting points. Thought of posting my learning</p><p>In the audio, Earl Nightingale says that the opposite to courage in today’s world is not cowardice but conformity</p><p>He is alluding is to the fact that we have all bought into a lifestyle prescription and following it blindly, mostly unhappily. How can prosperity be achieved through that?</p><p>Key to happiness and riches is to go beyond the routine, expand horizon, meet different set of peoples and breaking the imaginary chains that bind us.</p><p>In today’s world, with the opportunities that internet provides, there are multiple avenues available for achieving this. All that is needed is an intent to go beyond the obvious.</p><p>Start bringing in the change by connecting with people online, especially those that are in different profession than yours. Share your thoughts and engage with people. Stop being confined to a cubicle.</p><p>Connect with the world, even if virtually. Let new ideas come into your life, that’s the path to discovering the world beyond what you see every day.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9e4e5f541af" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>