Microsoft Build 2024 - Keynotes
![Microsoft Build 2024 - Keynotes](../assets/microsoft-build-2024.png)
Microsoft Build Opening
Satya Nadella
Microsoft's Microsoft Build conference opened with the keynote with a video talking about entering a new era with a new generation of AI where everyone can save time with a personal assistant, where anyone can access a personal tutor to learn new skills, where people can analyse with a personal coach. AI is unlocking creativity for us all, with expertise at your fingertips and you can build what matters - welcome to the age of AI transformation.
Satya Nadella welcomed everyone to Seattle and online and talked about developer conferences are the most fun when there are fundamental changes on the horizon. Satya Nadella remembers .NET and Azure being discussed and yet again we are at a moment like that, but the scope is so much deeper. Everything from the power draw of data centres to NPU processors and advancements are shaping every layer of the tech stack. There have been two real dreams, can computers understand us and in a world with so much information can computers help us reason, plan and act more effectively on all that information and we have had those dreams for the past seventy years. This is a real golden age of systems, so what is driving it, what are the scaling laws of machine learning, which is driving this revolution, Moore's law was every eighteen months but with these systems is every six months that have multi modal interfaces and recalls knowledge across our apps and devices along with performing complex tasks and reducing cognitive load.
Satya mentioned how developers can take these capabilities to change the world around us, from reasoning over farming data which showed the power of a model developed on the West Coast of the United States just a few months prior to improve the lives of farmers in India. It is great to see the democratisation force of this technology, it is great to see the impact of this. Satya gave a big thank you to those bringing these developments to the world and the progress that has been made since Microsoft Build 2023 including Microsoft Copilot, Copilot Stack and now introducing Copilot + PC.
Copilot + PC is exposing AI as a first-class namespace for Windows and introducing the Windows Copilot Runtime to provide applications and experiences with the Windows Copilot Library. This is a collection of ready to use local APIs to integrate AI capabilities to directly call capabilities and can also leverage Phi-Sillica on every Copilot + PC. You can take advantage of RAG and take advantage of vector stores and take advantage of local storage data for these models. AI frameworks and tool chains include PyTorch which is available in preview today, allowing developers to take advantage of AI features in their applications.
Satya mentioned they are introducing over fifty updates to the Copilot Stack and a apply the power of this technology to your applications. Microsoft has most complete AI infrastructure that meets the needs of AI and have over sixty regions and have expanded AI capacity and are making their best-in-class AI infrastructure available everywhere. They are on track to have all their data centres powered by renewable energy by 2025. End-to-end systems optimisation including datacentres using every megawatt of power efficiently. They have custom IO for high throughput including local storage and have the number one supercomputer for AI compute for training and are scaling the inference fleet of Azure. Microsoft offer best AI accelerators from nVidia and AMD with better performance and lower cost since it launched. nVidia and Microsoft have a full-stack partnership and will be offering their Blackwell Platform and working with them to optimise models such as GPT 4.o. They also have deep integration with the wider Microsoft Cloud including Microsoft Fabric with full access to Fabric data, also working on industry specific applications. Microsoft is delivering AMD technology with leading price and performance on GPT-4 inference plus Microsoft Maia hardware will be delivering many prompts along with Azure Cobalt powered VMs to power billions of conversations on Microsoft Teams and will be delivering functionality for customers such as MongoDB.
Satya mentioned with Azure AI they deliver the widest set of models with over 50,000 organisations using it today. Microsoft's most important partnership is with OpenAI with the latest model of GPT-4o which was trained on Azure and is the top performing model on a variety of modalities. Copilot will leverage GPT-4o which includes being able to share your screen or session and have Copilot help you such as helping craft a sword in Minecraft. Copilot is your personal assistant helping you with any task and helps you stay in your flow. Microsoft will bring all that innovation to Azure OpenAI too and have brought GPT-4o to Azure AI. Now you can have these groundbreaking apps that developers can build and any app or website can be turned into a full multi-modal full-duplex canvas. Satya thanks the OpenAI team and their responsible approach to developments in AI.
Satya also mentioned that Microsoft are bringing lots of other models and want to support the broadest set of models from every country including new models as a service and Microsoft also love open source and partnered with Hugging Face and are bringing more models into Azure AI. Microsoft are not just adding Large Language Models or LLM but also Small Language Models or SLMs including ones with language and vision capabilities with over four billion parameters and also ones with more parameters supporting multiple platforms with one AI model, customers are using them to summarise patient history in healthcare and make math tutoring more accessible, with Small Language Models the cost is a lot lower compared to Large Language Models.
Azure AI Studio provides an end-to-end studio including being able to evaluate models for performance and quality and is now generally available. You can build, train and refine models including built-in support for AI safety including detecting hallucinations and understand which prompts are triggering certain outputs and Content Safety will be adding more features such as prompt shields. Azure AI custom models will be coming that allow you to train a custom model for domain specific output and multi-modal and best in class including language proficiency that may be required.
Satya mentioned that your data needs to be in the best shape to be able to train models including Microsoft Intelligent Data Platform with over eleven thousand organisations using Microsoft Fabric and is integrated with unified compute and storage including the business model and it works with data anywhere such as AWS, GCP or your own Datacentre and real-time intelligence is being introduced to Microsoft Fabric. You can manage how your organisation consumes data with governance built in and can explore insights of petabytes of streaming data and can make efficient business decisions in real time and allow users to monitor the data they care about. Microsoft is making it easier for applications in interoperate with Microsoft Fabric with Microsoft Fabric Workload Development Kit. Microsoft is introducing the power of AI across the stack including adding AI RAG into Azure AI Search with built-in One Lake integration and Azure AI Studio to bring over your own AI models.
Microsoft is redefining development including GitHub Copilot with over fifty thousand companies using it to empower developers to access knowledge in their own native language and with Copilot Workspace makes staying in your flow even easier and go from idea to code to create a plan that can be executed to generate the code, but from every code from issue to spec to code you are in control and you can edit it, a fundamentally different way of building software. Microsoft are bridging third-party ecosystems with GitHub Copilot Extensions or with GitHub Copilot for Azure and get information about infrastructure and operations. Developers can build extensions for GitHub Copilot including your own in-house tools. Copilot Workspace looks at the intent of a specification of an issue and the whole process is iterative and can then build a plan on how to implement it and get Copilot's help to implement the code for this and you can see what changes are being proposed and can even get a live preview of the code changes and can submit changes as a pull-request to be evaluated.
Microsoft Copilot was built to allow you to tap into the worlds or your organisation's knowledge and is having broad enterprise impact by raising the ceiling and lowering the floor for accessibility. Microsoft Copilot connectors will allow Copilot to be grounded with data from a variety of sources in first- and third-party line of business data. Copilot is being expanded beyond a personal assistant with Team Copilot wherever you collaborate in Teams such as Loop or Planner and can be your meeting facilitator including making notes, tracking action items, addressing unresolved issues and make sure projects are running smoothly. With Copilot Studio anyone can build Copilots that have agent capabilities that can work on your behalf, or you can delegate authority to automate long running processes that work independently.
Satya mentioned that all around this stack is wrapping it with robust security, they are committed to a Secure Future Initiative with secure by design and default with secure operations including protecting engineering systems and protecting identities and secrets. The core of why we choose to be in this industry is to empower every person or organisation, power that everyone and that comes down to developers, it is not about celebrating tech for tech's sake but creating magical experiences that make a different and this new generation of AI is having an incredible impact to make a more accessible world.
Rajesh Jha - Empowering Productivity: The Future of AI in Microsoft 365
Rajesh talked about Copilot and how it is evolving. The journey Microsoft has been on to bring Copilot to the applications that hundreds of millions of users use every day such as Word, Excel, PowerPoint with dozens of businesses using Copilot. Microsoft has brought in more capabilities along with adding it to more applications. Microsoft has also brought the stand-alone Copilot app and what makes it powerful is the grounding it has in the web and the user's work context with the Microsoft Graph which represents their context. Copilot app is no different from any other Microsoft application and data is not used to train the models and in the app in the web tab you can get answers and in the work tab you can get information about communications, meetings and can make the experience personal. Learning to ask the right question is key and have a list of prompts in the prompt library. Rajesh even used Copilot to name the part of his keynote, Empowering Productivity: The Future of AI in Microsoft 365.
Rajesh mentioned that Microsoft want to do more, including expanding Copilot beyond a personal assistant with Team Copilot, it will act as a valuable team member. It integrates into a Team chat like any other participant and can help teams focus on having a productive conversation and anyone can add topics and even have reminders when to move onto the next topic and flag follow-up topics for everyone to see. As the conversation happens Copilot maintains an up-to-date overview for everyone to see and can provide suggestions for everyone to see based on documents in the context of Copilot to allow teams to be more productive and collaborative. Team Copilot will be a valuable team member to any team and will be available later this year.
Rajesh mentioned using Copilot to enable a reimagining of what you can do with the Copilot architecture, a Copilot can reason over data and is grounded in the web and your data in the Microsoft Graph and will understand the application context such as generating a slide show from a document and will surface capabilities based on the user context. With extensions you can further customise the Copilot experience, there are solutions that extend Microsoft Copilot such as Service Now to help users with their services and there is also Adobe to bring their capabilities such as Adobe Express to Copilot. Microsoft Copilot is already helping people save time and extend experiences in meaningful ways such as Lumen to enhance their business processes as they move from telecommunications to technology. Lumen uses Copilot to help them move faster and understand their customers better doing in minutes what took hours. Before Copilot people had to go to many systems but can now increase efficiency with outreach to customers from hours to fifteen minutes and this saves money and helps unlock growth.
Jeff Teper - Copilot Extensibility
Jeff mentioned that you can help your employees and grow your business and you expand your Copilot with your own extensions and reach hundreds and millions of users today and Microsoft is making it even easier to build Copilots from a few steps in Copilot Studio to full control in Visual Studio Code. Extensions run across all Copilots including Microsoft 365 apps, Microsoft Copilot reasons over user's prompt and it forwards that to the right extensions or you can @ mention to specifically leverage that functionality. This allows Copilot to have access to real time knowledge in your environment and allow users to focus on what they need to get done. You can build a Copilot quickly with a few steps in SharePoint that allow users access to information that they have permission to access. You not can only quickly create a Copilot in SharePoint you can share this with your team and there is a preview coming soon.
Jeff mentioned you can go into Copilot Studio to perform more advanced edits to a Copilot that you created, and you can make a Copilot do work for you to add more additional data sources such as websites or files along with thousands of Copilot connectors, you can also add Actions and tailor this along with inputs and outputs. You can see the conversation map for a Copilot seeing the knowledge source, flows and actions and then you can publish this Copilot for anyone to use.
You can create a Copilot Extension as a developer using Visual Studio Code along with the Teams AI Toolkit and can build an extension you can select a refined model or build your own tailored to what you need, then there is RAG integration where can use Azure AI Search or anything else and then can see any actions defined or registered along with being able to debug any Copilot extensions with breakpoints to allow end-to-end debugging. User experience can be defined as an Adaptive Card in JSON for the user experience for information returned from an Extension with full flexibility withing the Teams AI Library and Visual Studio Code, you can do this today targeting hundreds of millions of Teams Users and soon will be able to turn your extension into a custom Copilot. Microsoft have a unified marketplace to promote Copilot extensions for a vibrant third-party ecosystem. You can use Copilots to be more productive from something anyone can do in SharePoint, customise in Copilot Studio or develop with Visual Studio Code. Microsoft Teams will support syntax highlighting of source code along with co-editing of code and switch to compact mode to see more content along with keyboard shortcuts and can meet now to resolve any issues. Microsoft Teams also supports custom emoji to make the application more fun to use. Teams will be a great place for developers to work together and build this next generation AI.
Pavan Davuluri - Copilot + PC
Pavan mentioned this is favourite time to connect with developers and said that Windows I the best platform for Microsoft 365 and Copilot. Microsoft vision for Copilot is meeting you right in your workflow with quick actions and natural language and making Copilot more contextual and useful. Copilot + PC is for the most intelligent PCs ever built with NPUs that can be faster and more efficient for AI workloads and will be available from June 2024. Copilot + PCs are redefining what you can do with a PC and there will be a Qualcomm Snapdragon Dev Kit for Windows for your everyday dev kit for AI. Once thing that will never change is Microsoft's commitment to openness. Windows is the most open platform for AI. Building a powerful AI platform takes reimagining the entire system and the Windows Copilor Runtime extends those capabilities to Windows. It includes Windows Copilot Library which includes APIs which are shipped with Windows and built on foundation of silicon on devices.
Windows Copilot Library includes experiences like Recall to bring back anything seen on a PC and looking ahead Xbox is looking at Vision for Copilot experiences to help gamers with AI powered gaming experiences. Windows Copilot Library are APIs and Models including models that capture context on the screen with Recall and is indexed into a vector store called Windows Semantic Index so you can extend your applications into Recall to enhance your application, which will soon also include context from Microsoft Graph. It will also be possible to use Vector Embeddings APIs to create on-device vector stores to enable natural language search in your applications for your users. APIs cover spectrum to low code and sophisticated and multi modal models including Phi-Silica which is shipped with every Copilot + PC.
Pavan also mentioned DirectML is a low-level machine learning framework in Windows to give below the metal access to GPU and NPUs. DirectML will be brought to the web with WebNN for performant AI experiences with Machine Learning to use NPU on PC from web applications. Windows is the most open platform for AI, as Windows transforms using the power of AI. Also introducing Windows experiences beyond PC and making Windows a first-class experience on Meta devices to extend applications into 3D space with Volumetric Apps. For decades Windows has been stage for the world's innovation and will unlock a new era of innovation.
Kevin Scott
Kevin has been fascinated his entire life by tools and how they can help people to make extraordinary things, they have been building tools and infrastructure to help developers to build things for the reasons they need to make them. There is a range of customers from various categories and scales and how they can use AI to optimise what they've done for a while or innovate in new areas. Microsoft have a partnership with Cognition who developed Devin to help with various code tasks such as re-platforming, which must be done but it is rare that engineers enjoy doing that.
Kevin talked about what is driving this progress, there is a fundamental change in the universe of technology where Moore's law helped drive innovation and web connectivity brought things together. There is incredible scaling of AI systems as you bring in more data and compute into training them. The super important part is completing the stack when have a technology that is expanding at an exponential rate. Copilot stack is optimising the stack so it is more capable and cheaper and extend options around what you need and have more choices to build the things that matter under the constraints you may be under. No company has deployed more generative AI applications that Microsoft, this pattern started with GitHub Copilot where you help assist a user with tasks such as Copilots in Edge, Windows and more. Copilot Stack allows them to be built in a way they are safe and secure and cost optimised. Part of what the Copilot Stack is enabling Microsoft to combine everything into one logical Copilot, you don't have to pay attention to what service you are in but get that context in your tasks to you when you need it. It is not just about filling in that tool kit but also riding the wave into the development of the AI platform, compute for training is increasing exponentially and haven't reached diminishing returns as yet, are optimising the current frontier and investing at an incredible rate to push the frontier forward and driving efficiency.
GPT-4o is a multi-modal model that understand text, speech and images and interact with users in real time and can interrupt the model to have fluid interactions with the systems and investment into the model to make sure it is safe by design. This efficiency point is real where have managed a 12x decrease in cost and 6x increase of speed between GPT-4 and GPT-4o, it is extraordinary how much progress has been made with optimisations in data centres and software work and optimisation. You can count on things getting more robust and cheaper over a reasonably period of time as if you aim for truly ambitious then those will become ubiquitous. Just need to pay attention over what has happened over the past twelve months with the capabilities of CPT-4o would not have been possible when GPT-4 was released.
Kevin mentioned what is possible with smaller models and have worked on a series of models called Phi, this is an efficient frontier by trading size versus quality, it is cheaper to do inference and are more amenable for running on devices. If you want extreme capabilities, then frontier model is what you need but quality you are able to achieve with these smaller models is pretty high. With a version of Phi-3 optimised to run on a mobile phone can respond to a prompt just like GPT could do over a year ago so these models could be appropriate if you have a set of constraints you are trying to optimise towards. Kahn Academy has been using AI to help improve the learning capabilities of their platform, need to think about how an AI can align to a standard and make sure it works as expected to help with tutoring, but there is so much to do, and it is a very exciting time. If you give people a tool they will run with it but you can also have parents and teachers drive that learning experience, Kahn Academy will deliver state of the art teacher tools for free to every teacher within the United States. Kevin also mentioned another impactful area where these AI tools will have an impact in healthcare and how AI can help improve the outcomes for patients.
Kevin talked about building applications on top of this incredible platform that is emerging right now is to focus on things that have made transition from impossible to merely difficult, in the case of technology platforms is the most reasonable place to aim, if it is too expensive and fragile now it will be cheap and robust in the future - focus on those phase transitions. While developers have been building things over the year and Microsoft has been making forward progress on the frontier along with working on supercomputer infrastructure to push that frontier forward and there is a relationship between exponential computer and the power and capabilities of the platform. In 2020 they built a supercomputer to train GPT-3 which is the scale of shark and then the next system is scale of an Orca in 2024 that trained GPT 4 and the next size is a whale in comparison and that is working on the next set of capabilities to do the next round of amazing things with it.
Sam Altman
Sam Altman is one of the busiest people on the planet and spoke with Kevin, there has been an extraordinary change and the thing that has surprised Sam most is that developers are the most important and the speed and talent over not a very long time and seeing what is happening with GPT-4 and GPT-40 has surprised them to see how quickly people have been adopting them and has been very cool to see the uses they hadn't even thought of. There is a version of AI building things at scale, but OpenAI has taken those same things and made them available to developers, they thought about how this can be a good API and how this can be a layer to build into an application and make it very easy to integrate it into any product or service. OpenAI is making good use of that whale sized supercomputer, the most important thing is the models will get smarter, between GPT-3 and GPT-4 it got smarter and useful plus the underlying general capability improved. The incredible things that developers will make with future models and with GPT-4o allowed them to half costs and double speed and can use voice mode but overall intelligence that will be coming is one of the most important things.
Sam's advice on how to spend developers time, is this is probably the most exciting time to build a startup since the mobile boom or even the internet but the big opportunities come at the platform shift times and we haven't had one of these for a while, don't wait for the next thing, this is a special time and a few years where a great few things are going to start getting going. When mobile revolution people would say they were a mobile app or company but a few years later no one would say that - AI alone is a new enabler but it doesn't break the rules of a business you need to think of what you want to do with it still. OpenAI and Microsoft have been doing an extraordinary amount of work to safely deploy AI and built up a volume of experience, models are more useful at making AI systems safer. When OpenAI made the technology, they worked on building in a level of safety and robustness which took a lot of work and research and was impressive but had to look into what it took to deploy it for safety systems and policy was a huge amount of work, but it was necessary and Sam is super proud of the work and how fast it took, when models become more powerful as we move towards AGI the level of research will increase.
Next generation AI for developers with the Microsoft Cloud
Scott Guthrie
Scott mentioned that AI is changing the world and developers are at the centre of it, every app will be reinvented with AI and new apps will be developed that were not possible before. Microsoft will introduce a lot of capabilities that they will be shipping during the conference. Visual Studio is the world's most loved developer tool along with GitHub, the home of open source. GitHub is the AI powered platform with GitHub Copilot being the first ever copilot developed by Microsoft and recently released GitHub Copilot Workspace to build and test code using natural language but doesn't take over ownership in control as it allows developers to stay in control. Copilot chat can add things using the standard of the code you are writing or allow new developers to get started more easily. Developers don't just write code but also decide what to change which can start with an issue on GitHub and Copilot Workspace can use details of issue and knowledge from codebase to produce a desired end state in a spec which can be manually updated and then can produce a plan to show the changes that would be needed and you can keep iterating the plan to guide GitHub Workspace to the output you want, then you can submit a pull request for review. AI can also help keep code secure by analysing code and show vulnerabilities that can be fixed with one click where needed. Over fifty thousand organisations use GitHub Copilot including some that over half of their developers use it or check in thousands of lines of code that is written using GitHub Copilot.
Scott mentioned that you can use Azure Application Services in any application in any language including Azure Kubernetes Services, Azure Container Apps, Azure App Service, Azure Functions and Logic Apps securely and at scale. With Azure Application Services you can focus on building great apps knowing they will deliver on performance and scale that you need. Microsoft are introducing an AKS Automatic offering which automates everything from cluster management to security to ensure highest performance of applications so have access to self-service platform in minutes. .NET Aspire is now generally available and makes cloud-native application development easier in a repeatable, secure and reliable way. Combination of GitHub and Azure is optimised for the era of AI and making the process more efficient and reliable with GitHub Copilot for Azure which allows developers to ask questions about Azure to help experienced and new developers quickly along with troubleshooting issues from Visual Studio and Visual Studio Code.
Charles Lamanna
Charles talked about introducing Copilot Studio last year to extend Microsoft Copilot and build your own Copilots to better engage your customers and these can be published wherever you want such as Teams or Microsoft 365 apps and your own applications. New agent capabilities allow you to automate tasks asynchronously and where Copilots can be issued a defined task, and you can define meta-prompts for a Copilot along with the knowledge sources for information. You can also have connectors such as APIs that can be used within actions that can be used within Copilot Studio, you can control workflow with topics for conversations that runs a very specific workflow for what the user needs. You can test workflows for an incredibly robust workflow, but you also want to have agent capabilities where you can register triggers for your Copilot where data can be a trigger for your Copilot, and you can also understand what is driving the behaviour of a Copilot from within Copilot Studio. Over thirty-thousand businesses have been using Copilot Studio to create their own Copilots including reducing customer support time from fifteen minutes to thirty seconds.
Eric Boyd
AI solutions are transforming enterprise automation and building intelligence into products. Microsoft also runs own Copilots on Azure AI on the same platform as everyone else. No matter your use case Azure AI has you covered, within Azure AI Studio will find services like Azure AI Studios for pre-trained, turnkey solutions for intelligent applications, Azure Machine Learning for full-lifecycle tools for designing and maintaining AI models and Responsible AI Tooling to build and manage applications that are trustworthy by design. All of this is built on Azure's suer computing capabilities. Azure OpenAI service offers enterprise-grade security and compliance, nearly limitless throughput with the best SLA in the market. GPT-4o is available in Azure OpenAI Service supporting text, audio and vision as the most advanced multi-modal model. Azure OpenAI Batch Service optimises spend with scheduled use for batch jobs with 24-hour job completion with a capacity pool with more affordable prices. Many frontier and open-source models available including Hugging Face, OpenAI, Meta and more plus models-as-a-service are available.
Azure AI Search, where RAG support is needed, this offers revolution retrieval and allows customers to scale with performance and cost savings. OpenAI is harnessing the power of Azure AI Search and Azure CosmosDB - OpenAI have RAG-powered services at scale and to support this they need the best technology available. Azure AI Search is more than just a vector search engine it scales as needed and is enterprise grade for security and support scale needs and both OpenAI and Microsoft believe in safety. Azure AI Studio is now generally available.
Seth Juarez
Azure AI Studio allows you to develop capabilities and deliver them in a safe, reliable and observable way. It is your one-stop-shop for all things AI, there are a lot of models available that you can discover do a lot of amazing things to delight your customers. You can then craft a prompt but can swap values out later with database calls but can test how it work. There is a brand-new format to take a prompt from Azure AI Studio and take this into your Integrated Development Environment such as Visual Studio Code and you can plug some values in so you can use sample values for testing, but you can re-use this asset in your code or make any changes as needed in a safe repeatable and observable way.
Sarah Bird
Sarah mentioned that Microsoft infuses safety and security at every level and safety system looks for problematic inputs such as prompt injection and the model also has built in safety mechanisms and the output is monitored for copyrighted material or harmful content and can monitor it including for ongoing attacks, but there are also tools to manage and customise the safety experience with Azure AI Safety. Direct attacks such as prompts and indirect attacks in the grounding data can be prevented with safety settings. Can also see metrics for applications including risk and safety so can test this with adversarial datasets to stress test applications including testing jailbreak attack to allow developers to evaluate and mitigate risks. As AI becomes part of every application so need to monitor applications so can build secure and stay secure with HiddenLayer Model Scanner, Microsoft Purview and Microsoft Defender for Cloud to make AI systems that are trusted, safe and secure.
Arun Ulag
Data is the fuel that powers AI with Microsoft Intelligent Data Platform supporting Azure SQL DB, Azure CosmosDB, MySQL, Azure Databricks or Microsoft Purview. It is an exciting time to be in data, with Microsoft Intelligent Data Platform have a comprehensive list of tools and strong capabilities in each area and Microsoft do everything to make sure it can be used seamlessly as managed offerings on Azure. Data is both structured and unstructured and traffic patterns, companies have helped improved speed of their processes with the Microsoft Intelligent Data Platform. Copilot capabilities in Azure Databases for AI-enhanced management and performance now available in preview include self-help for managing or operating Azure SQL Databases, convert natural language to Azure SQL Database T-SQL queries and Chat with Azure Databases. Vector search for Azure Cosmos DB allows low latency and no additional costs and for PostgreSQL allows for AI-infused applications and seamlessly pass data to Azure AI and Azure OpenAI services. Microsoft Fabric is the data platform for the era of AI and gives you everything you need to go from raw data to BI value for users in a single SAAS platform, with unified compute and storage with unified governance and business model and there is also real-time intelligence coming which helps with accelerated decisions to unlock and act on real time data. Microsoft Fabric Workflow Development kit allows you acquire new customers with your own Fabric workflow. OneLake makes data lakes easier to use and is an open data format and provides industry level performance and that data can be worked with anywhere without having to duplicate it within Azure based on an open lake house architecture. Microsoft Fabric also supports integration with Snowflake allows them to work together seamlessly and Iceberg is another open format that can be supported by Fabric.
Infrastructure
Scott Guthrie mentioned investment in Azure which is the world's AI supercomputer and is being used to train the large foundational models with over sixty regions in the world and with all this power comes responsibility and next year every single data centre will be powered by renewable energy. When you use Azure you get price, performance and capability and Microsoft provides the most complete set of AI accelerators from nVidia and AMD and can dynamically choose which one is used by Microsoft Copilot and your Copilot apps. Developers and organisations are seizing the opportunity of AI with Microsoft. Microsoft is working with integrating work from industry software companies to harness the opportunity of AI. Microsoft Azure marketplace helps grow your business with billions of dollars are transacted in the past year. Safety and security are fundamental and have to be integrated into every part of platform and enable developers to build applications securely.
Julia Liuson & John Lambert
Julia and John talked about protecting and identities and secrets and protecting engineering systems, for security people incidents are how you define your job but need developers to fall into the pit of success to be able to secure what they are doing with secure by design, secure by default and secure operations. John talked about some of the trends and evolving landscape, we live in the most complex threat landscape in history and threat actors are doing more password attacks but Microsoft's goal is to make them really bad at it, stuff that developers make allows them to target multiple people so to hack one to hack any, hackers often think of connections between systems and target developers to access demo and test systems and build an attack graph.
Julia and John mentioned Microsoft are actively working on secure by design and get developers to follow the paved path of those to make sure they are doing things securely. Threat actors also perform supply chain attacks like SolarWinds which shows importance of protecting engineering systems. So, need to understand engineering systems and get rid of unneeded apps and systems whether it is managed or not, modern or not if it has access to resources is that is what they are after so best way is sometimes to get rid of these old systems. Get rid of anything you don't need from engineering systems and need to protect non-production systems with the same rigour as production, and systems have clear lifecycle governance. Developer projects themselves have been targeted such as XZ utilities which was detected by a slowdown in the project by a Microsoft engineer, but also shows the behaviour of a developer who has a security-first mindset. Security is a team sport, defending against threat actors is not only a challenge for Microsoft but also everyone else so can raise security posture and help customer improve their application code and processes to be more secure and help make the world a safer place.