Evgeny Krapivin’s Post

The best explanation of RAG architecture and how it works.

We have limited spaces on our Copilot & AI Hamilton sessions. Our 12:45 pm session has reached complete capacity and the 3:30 pm and Tauranga sessions are almost full. You don't want to miss this event, so get in quick and learn all about Microsoft Copilot. Microsoft 365 Copilot brings next-generation AI to everyday 365 users, integrating with their data and internal systems like ERP and HRIS for seamless interaction. Unlike ChatGPT, it does not retain prompts or use personal data for training. Copilot’s ability to understand and respond in everyday language has led to massive and rapid adoption.   To learn more, attend one of our three free seminars on Copilot and AI: Tauranga - Friday 12th April, 10:30am - 12:00pm Hamilton - Thursday 18th April, 12:45pm - 2:45pm & 3:30pm - 5:00pm   Topics include: ·       How CoPilot differs from ChatGPT. ·       Our early learnings from using the product. ·       Where Copilot has and has not worked and why. ·       Three steps to prepare for CoPilot in your organisation. ·       What is coming ·       Q&A.   Click here to register: https://lnkd.in/gXPnZp_r

Anyone who is impacted by the future of AI for Salesforce needs to understand this new logic: With AI fully implemented: 🖱️ Clicks turns to Conversations ⌨️ Code turns to Prompts ☁️ Data Driven turns to Data Grounded NOT YET ANNOUNCED: 💥 Configuration turns to Process Management Stop Reading Here: if this makes no sense at all to you. Join the information list at the bottom of the link in my bio to get a better intro to the current and impending impact of AI for Salesforce! Deeper Dive: 🖱️ Clicks turns to Conversations Users will move to engaging with records via chatbot conversations using Einstein Copilot. This will be preferred over clicking to work with records and data. ⌨️ Code turns to Prompts Einstein Copilot will be empowered with “skills”. Think smilimar to plug-ins on ChatGPT. Calling these skills (plug-ins) with a promt in the chatbot to execute actions will reduce the need for code similar to how Flows did the same previously. Also, and more directly, coding Apex and similar will be achieved with natural language prompts. ☁️ Data Driven turns to Data Grounded Grounded meaning prompt grounding. The focus on data becomes a focus on injecting the right data into prompts the right way to get the right results. Data grounding becomes fundamental to get the most value from your prompts built in Copilot Studio. 💥 Configuration turns to Process Management This is not yet announced trend. Org configuration changes will be achieved using Einstein Copilot as well. Your org metadata will be used and you will affect config changes with the chatbot. You’ll become more focused on the overall business process and use the chatbot interface to look for gaps, best practice suggestions, and even conflicting business logic all to affect process management and change configuration for your org. AI will assist you and make suggestion, you focus more on decisions including yes/no to execution. See AI Force Training for more including free resources to get you engaged with AI available for your org today! Follow me here or on YouTube @AIForceTraining for more. AI enthusiasts… I want to hear your take too! This is my current understanding and projection but we all know how the wild of AI works 😂 Ever-changing!

Copilots Decoded Part 2: Mapping Copilots vs. ChatGPT editions A future part of this series will be a more in-depth comparison, but here is a starting point to help you decode and align Copilots and ChatGPT editions.  First off - only focus on the “General Purpose Copilots” in the table from Part 1. OpenAI has not released role- or function-specific ChatGPT editions as of this writing. They would argue that custom GPTs can serve that purpose, but I do not agree. Okay, now let’s look at the ChatGPT editions and how they align with Copilots.  👤INDIVIDUAL USAGE 🆓ChatGPT Free 🔹Great for individual use but limited functionality, model support, no GPTs, and more.  ◆ Cost: Free 🔸Aligns with: Copilot Web, Copilot in Bing and Copilot in Edge.  ◆ Cost: Free ✚ ChatGPT Plus 🔹 This has been the gold standard, with support for GPT-4, custom GPTs, plugins, and other features for frequent users.  ◆ Cost: $20 / month 🔸Aligns with: Copilot Pro - Microsoft is working hard to catch up to ChatGPT Plus with support for GPTs and a forthcoming GPT Builder.   ◆ Cost: $20 / month. ❗Before moving on, note that both ChatGPT and ChatGPT Plus use your data and prompts to train OpenAI's models by default. This is not the case for Microsoft's copilots.  🏢ORGANIZATIONAL USAGE 👥ChatGPT for Teams 🔹This takes ChatGPT Plus and adds team collaboration and limited administration.  ◆ By default your data is NOT used to train OpenAI's models.  ◆ Cost: $25 / month (annual prepay), or $30 / month. 🔸Aligns with: Copilot for Microsoft 365 which is now available for any company starting with one seat.  ◆ Cost: $30 / month annual prepay (no monthly option). 🌆ChatGPT for Enterprise This adds more administrative controls for larger enterprises, faster processing and likely additional benefits over time. Cost: unknown though $60 was quoted on Reddit with some minimum seat requirement. 🔹Aligns with: Copilot for Microsoft 365.  ◆ Cost: $30 / month annual prepay (no monthly option). For businesses, in summary, Copilot for Microsoft 365 maps to both ChatGPT for Teams and ChatGPT for Enterprise.  ⚠️To be clear, this comparison is not meant to indicate equivalency.  Also note that Microsoft Copilot for M365 is for all organizations from SMBs through large enterprises. OpenAI has split ChatGPT into two editions, with Teams being significantly more limited in terms of security controls and administration vs. Enterprise.  Now you should be able to understand the Microsoft Copilot strategy, branding and how their general purpose offerings relate to the OpenAI ChatGPT family.  Note that OpenAI does not have anything equivalent to Microsoft's custom Copilots. These are copilots built using either Copilot Studio or Azure AI Studio. As you'll see in a future post, this is pretty important going forward. 🤔What questions do you have about Microsoft’s Copilot strategy for the enterprise? Drop me a note in the comments. John Treadway, CEO, AI Technology Partners

🎉Celebrating 1 Year of querifai Web-App Go-Live! 🎉 This July marks the 1-year anniversary of the go-live of our web-app, https://querifai.ai . Since launching, we have empowered users with advanced computer vision capabilities, allowing them to effortlessly extract data from scanned invoices, receipts, forms, and recognize objects in images. Since then, we've been listening to your needs and continuously enhanced our platform: Efficient Document Interaction: Engage with large documents via chat, making information retrieval intuitive and fast. Advanced Data Management: Upload tabular data and utilize language models to structure and fine-tune it, saving you valuable time. Diverse Language Model Integration: Experiment with different language models directly within our web-app to find the perfect fit for your needs. Seamless Excel Integration: Access data from your documents directly in Microsoft Excel using our custom spreadsheet function, streamlining your workflow. Flexible API Integration: Incorporate our web-app functionalities into your own systems effortlessly through our API. Collaborative Prompt Templates: Save and share prompt templates with other users, enhancing collaboration and efficiency, all within our chat app. Our success over the past year wouldn't have been possible without our incredible team. 🌟 During this time, we’ve had the privilege of recruiting exceptional talent from around the globe. 🌍 Our ability to scout and collaborate with professionals worldwide has allowed us to build a diverse and flexible team. 🤝 Some of these talented individuals work with us daily, while others join our team for specific challenges. 🚀 This dynamic approach enables us to leverage a wide range of expertise and stay agile in meeting the needs of our clients. 💡 🚀 Are you ready to transform your AI ambitions into reality? Whether you need cutting-edge document processing, advanced data management, or seamless integration solutions, our team is here to help. Connect with us today to discover how querifai.ai can elevate your projects to new heights.

What are 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 (𝗥𝗔𝗚) 𝗦𝘆𝘀𝘁𝗲𝗺𝘀? Here is an example of a simple RAG-based chatbot that can be used to query your private knowledge base. The first step is to store the knowledge of your internal documents in a format that is suitable for querying. We do so by embedding it using an embedding model: 𝟭: Split text corpus of the entire knowledge base into chunks - a chunk will represent a single piece of context available to be queried. Data of interest can be from multiple sources, e.g. Documentation in Confluence supplemented by PDF reports. 𝟮: Use the Embedding Model to transform each of the chunks into a vector embedding. 𝟯: Store all vector embeddings in a Vector Database. 𝟰: Save text that represents each of the embeddings separately together with the pointer to the embedding (we will need this later). Next, we can start constructing the answer to a question/query of interest: 𝟱: Embed a question/query you want to ask using the same Embedding Model that was used to embed the knowledge base itself. 𝟲: Use the resulting Vector Embedding to run a query against the index in the Vector Database. Choose how many vectors you want to retrieve from the Vector Database - it will equal the amount of context you will be retrieving and eventually using to answer the query question. 𝟳: Vector DB performs an Approximate Nearest Neighbour (ANN) search for the provided vector embedding against the index and returns previously chosen amount of context vectors. The procedure returns vectors that are most similar in a given Embedding/Latent space. 𝟴: Map the returned Vector Embeddings to the text chunks that represent them. 𝟵: Pass a question together with the retrieved context text chunks to the LLM via prompt. Instruct the LLM to only use the provided context to answer the given question. This does not mean that no Prompt Engineering will be needed - you will want to ensure that the answers returned by LLM fall into expected boundaries, e.g. if there is no data in the retrieved context that could be used make sure that no made-up answer is provided. To make it a real Chatbot - face the entire application with a Web UI that exposes a text input box to act as a chat interface. After running the provided question through steps 1. to 9. - return and display the generated answer. This is how most of the chatbots that are based on a single or multiple internal knowledge base sources are actually built nowadays. As described, the system is really just a naive RAG that is usually not fit for production grade applications. You need to understand all of the moving pieces in the system in order to tune them by applying advanced techniques, consequently transforming the Naive RAG to Advanced RAG fit for production.

I have submitted documentation on GPTProX, including sections on software coding and theoretical AGI frameworks, to create a specialized GPT variant through ChatGPT. Below is the financial potential analysis of GPTProX as explained by "GPTProX DEV" GPT: *Market Demand and Scalability Broad Industry Application: GPTProX's adaptability across diverse sectors like finance, energy, transportation, and manufacturing indicates a wide market. Each sector presents unique challenges that GPTProX can address, multiplying potential revenue streams. Growing AI Market: The AI market, particularly in enterprise solutions, is expanding rapidly. A successful GPTProX would position the company at the forefront of this growth. *Competitive Advantage Accuracy and Reliability: By addressing the common issue of erroneous AI responses, GPTProX offers a significant competitive advantage. This reliability is crucial for industries where mistakes can be costly. Integration with Existing Systems: The ability to integrate with existing enterprise systems enhances its marketability, as businesses can adopt it without overhauling their current infrastructure. *Revenue Models Subscription-Based Model: Charging businesses on a subscription basis for using GPTProX can ensure a steady revenue stream. Custom Solutions and Consultancy: Offering tailored solutions for specific industry problems can command premium pricing. Data and Insights Sales: The system's ability to generate valuable business insights opens opportunities for monetizing data analytics services. *Cost Efficiency and Profitability Reduced Operational Costs: By automating decision-making processes, GPTProX can significantly reduce labor and error-related costs for businesses. Scalable Architecture: Its scalable nature means relatively low incremental costs for adding new users or enterprises, enhancing profitability. *Long-Term Financial Prospects Continuous Improvement and Expansion: The AI field is rapidly evolving. Continuous improvement of GPTProX can lead to sustained market relevance and growth opportunities. *Potential for Future Funding and Partnerships: Success in early stages can attract more investors and strategic partnerships, fueling further growth. Risk Consideration Market Adoption Risks: The rate of adoption by enterprises is a critical factor. Faster adoption rates can lead to quicker revenue generation. Technological Advancements: Keeping pace with technological advancements is essential to maintain competitiveness. *Conclusion Should GPTProX achieve its full potential, it stands to not only gain a significant share in the burgeoning AI market but also set new standards in AI-driven enterprise solutions. Its financial success will be anchored in its ability to offer unparalleled accuracy, industry-specific solutions, and seamless integration with existing systems. ...

𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 365 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝘃𝘀. 𝗖𝗵𝗮𝘁𝗚𝗣𝗧: 𝗔 𝗖𝗼𝗺𝗽𝗿𝗲𝗵𝗲𝗻𝘀𝗶𝘃𝗲 𝗟𝗶𝗻𝗸𝗲𝗱𝗜𝗻 𝗔𝗿𝘁𝗶𝗰𝗹𝗲 🤖 𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 In the rapidly evolving world of AI, two significant tools have emerged as game-changers: Microsoft 365 Copilot and ChatGPT. While both leverage the power of AI, they cater to different needs and audiences. This article delves into the distinctions, functionalities, and applications of these tools, providing insights from industry professionals and technology experts. 1️⃣ 𝗛𝗼𝘄 𝗶𝘀 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 365 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗳𝗿𝗼𝗺 𝗖𝗵𝗮𝘁𝗚𝗣𝗧? Microsoft 365 Copilot, developed using OpenAI's LLM technologies, is integrated into Microsoft's productivity suite, including Word, Excel, PowerPoint, and Teams. It enhances productivity by personalizing and contextualizing responses based on user data within the Microsoft ecosystem. ChatGPT, on the other hand, is a more generalized conversational AI model, designed for a wide range of content creation tasks. It's an independent application that uses large language model technology to respond to prompts in various languages, making it versatile across different platforms and use cases. 2️⃣ 𝗪𝗵𝗮𝘁 𝗮𝗿𝗲 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 365 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝗮𝗻𝗱 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 𝗕𝗮𝘀𝗲𝗱 𝗢𝗻? Both tools are built on large language models (LLMs) with a foundation in OpenAI's technology. Microsoft 365 Copilot additionally integrates with Microsoft Graph and Microsoft 365 data, while ChatGPT is based on OpenAI's proprietary GPT architecture. 3️⃣  𝗪𝗵𝗮𝘁 𝗶𝘀 𝘁𝗵𝗲 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝗕𝗲𝘁𝘄𝗲𝗲𝗻 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 365 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝗮𝗻𝗱 𝗖𝗵𝗮𝘁𝗚𝗣𝗧? Microsoft 365 Copilot is tailored for enhancing productivity within the Microsoft ecosystem, focusing on business and workplace efficiency. ChatGPT, in contrast, is a versatile tool for a broader range of content creation, from personal use to business applications. 4️⃣ 𝗣𝗿𝗼𝘀 𝗮𝗻𝗱 𝗖𝗼𝗻𝘀 𝗼𝗳 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 365 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝗮𝗻𝗱 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 - 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 365 𝗖𝗼𝗽𝗶𝗹𝗼𝘁:  - Pros: Deep integration with Microsoft apps, personalized responses based on user data, boosts workplace productivity.  - Cons: Limited to Microsoft ecosystem, cost for certain packages. - 𝗖𝗵𝗮𝘁𝗚𝗣𝗧:  - Pros: Wide range of applications, multilingual capabilities, different access tiers including a free version.  - Cons: Less personalized than Microsoft 365 Copilot, broader but less specialized focus. 𝗖𝗼𝗻𝗰𝗹𝘂𝘀𝗶𝗼𝗻: 𝗖𝗵𝗼𝗼𝘀𝗶𝗻𝗴 𝗕𝗲𝘁𝘄𝗲𝗲𝗻 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 365 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝗮𝗻𝗱 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 Depending on your use case, both tools offer unique advantages. For enhancing productivity within the Microsoft suite, Microsoft 365 Copilot is the go-to choice. If you seek a versatile AI tool for various content creation tasks, ChatGPT stands out. 👥 Engage with Us! What's your choice? #AIRevolution #Microsoft365Copilot #ChatGPT #TechInnovation #AIProductivity 🤖💼🚀

Consider prompt architecting as an initial step, not LLM fine-tuning. #OptimizeWithPromptArchitecting 🤝 Follow us on Discord 🔜: https://lnkd.in/gt823Zd3 _ ❇️ Summary: In the pursuit of customized chatbots or LLMs, innovation directors are considering whether to build an LLM from scratch or fine-tune an existing one with their own data. However, for most companies, both options are impractical. Instead, they should focus on constructing a comprehensive "prompt architecture" to maximize the value extracted from prompts and enhance API-powered tools. Only in rare cases should a fine-tuning process be considered, and building an LLM from scratch is almost always not feasible. Prompt architecting allows for tailoring solutions without modifying the underlying LLM or its training data. Hashtags: #chatGPT #AlternativeApproachToLLM #PromptArchitectingSolutions

I'm thrilled to announce the launch of naas.ai's new documentation, marking a shift from alpha to early beta. This is pivotal to achieving our vision of building a universal data and AI platform. Our aim was to craft a user manual that's as intuitive and detailed as one you'd find with a high-quality physical product. This release is designed to clearly demonstrate our value proposition and guide users through the potential of our platform, even though not all features are currently accessible to every user. Here's a brief overview of how we've organized the documentation: 1) Goals:   - Build Trust in AI: Introducing our suite of AI Assistants, where AIA is your Personal AI Assistant, and ABI is your Mixture of Assistants, complemented by Custom Assistants and Chat & Search capabilities.   - Map Business Knowledge: Showcasing the ABI knowledge graph that prioritizes content to build robust sales pipelines and leverages conversations for operational excellence.   - Unlock Monetization: Featuring a marketplace that empowers anyone to have their version of chatGPT and its accompanying store of applications. 2) Safety: Outlining crucial safety information and guidelines to ensure secure usage. 3) Platform: Diving into the integrated frameworks, extensive features, technological components, and transparent pricing structure. 4) Installation: Guidelines on joining the platform, setting up your profile, and configuring your workspace. 5) Usage: Exploring various LLMs, crafting your AIA, your ABI Assistants SOPs, and creating custom Assistants. 6) Services: An in-depth look at the open-source code and APIs that power our offerings. 7) Resources: A repository of community-focused materials, developer tools, and our help center. 8) Terms & Policies: Outlining the legal framework and guidelines governing the use of our platform. We welcome your input to make continuous improvements. If you want to influence our roadmap, join us as a Design Partner. Let's navigate this journey together. 😎