Google Data Cloud & AI Summit 2023: Reveal opportunities to transform your business

Key Takeaways

[Music] thank you [Music] [Music] hello everyone welcome to data cloud and AI Summit my name is Jun Yang I lead our AI team here at Google Cloud we're at the start of a new era of AI innovation consumer grade generative AI has captured attentions of Millions with intelligent chatbot and lifelike digital avatars at the heart of this waiver Innovation it is about Foundation models so what is a foundation model I ask our Palm bundle for definition here is how it responded Foundation models are large AI models that can be adapted to a wide range of tasks after being trained a massive amount of unlabeled data pretty good couldn't have said better myself while AI models are not new this Foundation models have several important characteristics that represent a step function change from the previous generation of AI models Foundation models are multitask rather than single task One Foundation model can perform a wide range of tasks out of the box such as summarization q a classification Foundation models are generative in nature and capable of generating high quality text images speech code and much more and with zero and minimal training required Foundation model works well out of the box and can be adapted for targeted use cases with very little example data Foundation model have been years in the making and Google has been at the very Forefront of it let me share with you a few key moments starting with infrastructure as general purpose compute began to reach the very limits of physics deep learning requirements LED Google to create the TPU this specialized Hardware has since been made available to Google Cloud customers and allow teams to innovate with large models in 2017 Google introduced Transformer technology this spurt the beginning of foundation models and the generative AI innovation in 2018 Bert a groundbreaking large language model was introduced to the world later deepmind evolved the notion about reinforcement learning the ability for models to learn from Human feedback which further accelerated the generative AI innovation these are Foundation model breakthrough that reinforce the rapid democratization of AI and the creation of new category of generative AI application or what we call gen apps with Gen apps organization can pursue a whole set of new application and customer experiences just as a shift from web application to mobile application made it possible for more people to access information and services from anywhere in the world anytime gen apps is poised to enhance the way human interact with technology while Foundation models are powerful and capable to build enterprise-grade gen knobs Enterprises need more than just Foundation models at Google Cloud we're combining the power Foundation model with Enterprise search and conversation AI to enable Enterprises to be in the driver's seat to Define how they want generative AI to surface in their application for example every Enterprise has a large internal repository of data and will want to generate specific responses from it rather than just from the general information from the model Enterprise search complements Foundation model to provide up-to-date and targeted information from the internal data alongside with public data conversation AI enables Enterprises to steer interaction down productive path not just about information but also about transaction to determine when a response should be retrieved from your backend database like how much is my bill this month or when a transaction such as pay now should occur we will talk through a few use cases in just a little bit to bring generative AI to Enterprises we introduced two platforms recently first gen app builder is a brand new offering that brings together Foundation model with the power of search and conversation AI to enable Enterprises to develop new generative AI apps and vertex AI is our end-to-end unified ml development and deployment platform now with this biggest update yet with generative AI support our Ranger Foundation models are made available to developers and data scientists developers benefit from Easy API access and data scientists have a full Suite of tuning options for customizing these Foundation models now let's go ahead and take a look at the some of the generative AI use cases in Enterprise let me introduce Lisa O'Malley one of our leaders in product management who will take us through some amazing demos Lisa welcome hi June great to be here yeah Lisa what have you been hearing from our customers about their ideas with generative AI you know June our customers have so many interesting ideas with remarkable potential for business value and across all the conversations many common themes emerge to start when online interactions become more conversational and human-like it creates an opportunity for you to better connect with your customers employees and partners let's take a look at how generative AI can transform the customer experience [Music] and [Music] thank you foreign [Music] foreign [Music] even more exciting is that gen apps will be incredibly easy to build which means more people than ever will be able to create experiences like you just saw let me show you how by using our new gen app builder let's start by adding the company's URL and some business documents like these how-to guides and policies the data is stored in the company's private protected environment and Google will only use it with the company's permission using gen app builder you can either power an Enterprise search experience a conversational chat experience or both in this case let's choose chat and hit create the custom multi-turn conversation bot is ready it's that simple the bot comes with granular control over generated responses here for example you can choose what types of responses to allow or even set specific terms that you do not want to appear you can also disable generated responses completely at any time and the bot can still answer complex questions thanks to Google's search technology for now let's leave it on as you saw in the video the bot doesn't just provide information it can also actually complete transactions for the user as well pre-built flows cover common interactions like check order status explain Bill or book an appointment and these can be added to the bot with a single click if you wanted to build your own transaction flow you can choose either a simple graph based interface to Define high-level business logic or you can choose to leverage prompt-based flow creation to specify logic using just natural language instructions in either option AI dynamically handles all of the potential paths the user might take making development dramatically faster and easier now I'm ready to test the bot awesome this looks great out of the box Integrations make it easy for you to launch your Bot to your website on popular messaging apps I can also connect with telephony partners I'm choosing website it's easy to get the widget deployment code and now you're ready to deploy wow Lisa that's such a great use case and just about every Enterprise could benefit from having a virtual assistant to deliver better experiences for their customers Partners employees it's great to see that developers can leverage Genet Builder to build gen apps with ease absolutely totally agreed June now let's talk about another use case many organizations have huge knowledge bases and want to leverage generative AI capabilities for fast effective q a with their data let's take a look at how we can make complex data more accessible and more useful I'm an analyst at simple Investments a fictional company today I need to do detailed research to assess the semiconductor Market in order to build an investment strategy in the past this might have taken me days but my team just launched a new research application powered by gen app builder and I'm excited to see how it can help me save time I know some basic facts about the industry for example that there has been a global semiconductor shortage so I start by asking in plain language which Industries have been most impacted now with one single prompt and interface I can see a variety of responses from both the internal and external data sources that my company has provided each entry has an AI generated summary to help me quickly understand what's important within it here I see a comprehensive analysis written by a well-respected in-house Analyst at symbol and I want to dig in further on the left you see automatically generated section headers known as facets which make browsing super simple my organization subscribes to some top Financial journals that will add important color to this topic so I go to the application settings to make sure these results are also included in my search here I can easily connect to a variety of data sources based on my organization's needs I'm also going to enable a synthesized summary to help contextualize findings now that I've included additional sources I want to ask a follow-up question on interest rates since that was the focus of the article I just read you can see here that the synthesized summary concisely explains the findings in this case it helps me understand that despite the supply shortages and Rising rates the demand for semiconductors is expected to remain strong long term which is two queries I'm already starting to get a perspective on the industry that I can share with my manager using my company's new Search application powered by gen app builder I was able to find relevant and up-to-date information with just a few simple clicks and prompts making my research faster and easier than ever now I can show you how easy it is to create this with Gen app builder let's get started by following the same steps for data ingestion that we just walked through for creating the chatbot URL and business files are pre-loaded here from there we select Enterprise search and hit create genop Builder creates a search engine in minutes it's that easy now we want to quickly preview the out of the box experience which we can do right here we can click on a recommended sample search the application provides for testing this works well we can customize it further so let's do that here you can create rules for tuning the search experience for example if you wanted to return more information about us-based companies you could boost that search parameter you also have granular control over the widget UI settings like the ability to enable or disable facets summarization autocomplete and more everything looks great and once again it's easy to get the widget deployment code and we're ready to deploy your AI practitioners can also tune and develop AIML models in vertex to incorporate their organization's own embeddings and Knowledge Graph data all this can be easily imported into the Enterprise search experience to summarize bringing together Foundation models conversational capabilities and search we can create really powerful gen app experiences that are Enterprise ready we just saw how you can use gen app builder to provide helpful factual up-to-date information go beyond text interactions to full multimodality combine structured flows with conversations and transact instead of just inform and you can do all of this in minutes Lisa thank you for the great demo and showcasing how generative AI can be at work for Enterprises thanks for having me June I really enjoyed it now I have one more use case to show you for this use case we will leverage generative AI for Content Creation in Enterprises not just any content creation but content Creation with your organization's brand voice personality by customizing Foundation models let's Jump Right In all right I'm going to go ahead build an application for my marketing team that can generate custom copies and images for symbol a fictional company first I'll create a custom large language model for healthy snacks one of symbols Brands this will serve as a backbone for Content generation in vertex AI there's a new feature model Garden where I can select from a variety of models including Foundation models built by Google open source models and even third-party models for this application I'm going to go ahead choose one Google's Foundation model Palm bison text next I'll tune my model to create a custom Foundation model for my business for extensive training I can select fine-tune this results in changes to the model's weights and is a great option for data scientists who needs to produce outputs with specialized results like legal or medical vocabulary but since I'm creating a simple marketing app for Content generation that's not necessary I'll just go ahead and prompt tune this model in a simple tuning interface I'll upload some of my company's data using documents that were written in the brand voice like press releases tweets and blog posts my uploaded data is stored in my company's environment so I know that's protected I know that Google will only use my data with my permission and for the tasks I specify after I've customized the model I can test it and ensure I'm getting the results I want without leaving vertex AI this allows me to iterate quickly I deployed my model to get a custom API endpoint that will make it easy in that marketing app the marketing application also needs to generate and edit images so that every social post has a picture and we'll also deploy a image model via an API vertex AIS manage endpoints makes it easy to build into my application with just the fuel line code just like any other Google Cloud API I don't need to worry about the complexities of provisioning storage Computer Resources or optimizing models for inferences it all just works with vertex ai's developer friendly tools you've seen how easy it is to customize and deploy Foundation models at scale now let's take a look at this application in action we build a campaign for simple healthy snacks highlighting the ingredients of its most popular granola bar generative AI makes simple tax prompts really powerful for example using a simple prompt I can describe the image I would like add a sentence or two about the desired content and click generate and here we go was the power of generative AI I now have custom social media blog post email campaign post that I can use across all of my marketing channels from there I can refine further I can edit my images I can make changes to the content I can make Tech cassettes directly I'm going to go ahead and select the pairing of images and the text that I like and done this shows you how generative AI support on vertex AI allows you to leverage Foundation models with developer friendly apis access through Google's latest generative AI models and the ability to tune deploy and manage models this is just a taste well Google Cloud generative AI capability can do for you and we are just at the very beginning of the era of gen naps let's recap the three categories of use cases we saw today first when we think about how to make online interaction more conversational more human-like and helpful the potential for improving customer experiences engagement and loyalty is immense second the ability to intuitively interact with complex data across sources means you can better enable your teams with business and industry insights third content generation gives people the ability to turn their creative ideas into targeted content rapidly this is all made possible by bringing together Foundation models conversation Ai and Enterprise search since our launch two weeks ago we have seen such incredible demand from our customers we're so excited to be working with these leading customers partners and more to help them realize this tremendous opportunity with generative AI when Google Cloud brings new AI advances to our products our commitments is twofold deliver transformative capabilities while ensuring the proper protection are in place for our customers their users and Society at Large our AI principle established in 2017 form a living Constitution that guides our approach to building advanced technology conducting research and drafting our policies our new generative AI offering are no exceptions factuality and freshness is met by providing the right sourcing and attribution while serving the most relevant information and we ensure Enterprise needs are met by providing choices and control including data isolation and privacy private data is kept private and is not used for foundation model training without your permission organization always maintain control over where the data is stored and how and if it's used our customers also benefit from an open ecosystem our partners are ready to support generative AI at every layer of the stack from the chip makers to the technology providers to the app Builders to the service providers together we're committed to make our Enterprise customers take advantage of generative AI today we're excited to highlight a deeper partnership with Nvidia to provide our customers with high performance AI infrastructure for generative Ai and other modern workloads like augmented reality video streaming Robotics and more together was Nvidia our goal is to help our customers run generative eye workloads with greater efficiency reduce cost and more choice now please welcome Ian Buck who will share more details on how Google cloud and Nvidia a helping organization innovate in this gen up era thanks June we're excited to be partnering with Google Cloud to help companies accelerate generative Ai and other modern AI workloads in a cost-effective energy efficient and sustainable way our partnership focuses on the latest AI models and accelerated Computing for AI inference training data analytics video processing and integration with open source software tools Google cloud is the first cloud provider to offer The L4 tensor core GPU which is purpose built for generative AI inference visualization and video applications we're also excited to be bringing the h100 tensor core GPU onto Google Cloud for large-scale training of the most challenging models Google Cloud's new generative AI capabilities with vertex AI combined with the power of Nvidia L4 and h100 gpus will make it easier for customers to build applications that leverage large complex models and we're excited to offer Nvidia AI Enterprise a software suite with application Frameworks and pre-trained models available inside the Google Cloud Marketplace this offering helps customers build AI applications faster and with Enterprise grade support directly from Nvidia we're also thrilled that Google Cloud now offers the GPU accelerated Apache spark on dataproc this uses the Nvidia Rapids Library to dramatically increase AI ML and ETL workloads without any code change the Strategic partnership with Nvidia and Google cloud is just the beginning of what the Enterprise world is going to activate with AI and we look forward to the ongoing collaboration to bring generative AI services and accelerated cloud computing to Google Cloud's customers [Music] thank you Ian and Nvidia for ongoing support and collaboration we can't wait to see what gen apps you're going to build with Google Cloud to sign up for Early Access please join The Trusted tester program and visit g.co Cloud slash trusted tester and stay tuned for the data Cloud keynote with my colleagues Garrett and Andy and stick around for more sessions on generative AI for the rest of the day thank you so much for joining us today bye-bye [Music] Welcome to our customers and partners and thank you for joining our third annual data Cloud Nai Summit I'm Garrett katzmeyer and in today's keynote I will be covering the latest Innovations in data analytics and Ai and I'm joined by Andy gutmans who will take us through the latest Innovations in our database services but let us start with what matters most how you our customers unlock your data Advantage with Google's data cloud Woolworths has seen huge improvements in their NPS by migrating from Hadoop to bigquery and improving their Insight time to insights from five hours down to just three seconds Lufthansa group subsidiaries Swiss International Airlines they saved seven million dollars and over 8 000 tons of CO2 in 2022 by optimizing the operations with data and AI and sound regular brand runs machine learning models against a catalog of 100 million songs with up to a 140 times performance Improvement we are so proud of how you our customers transform your businesses with our data cloud now looking forward 2023 is the year of both exciting Innovations and big challenges and there are three big priorities that we have heard from you first you need flexibility for predictable and unpredictable data needs with optimal Financial control you need to work with data across Enterprises in a privacy safe way and you need a faster time to value in all of your data projects so let us start with achieving optimal flexibility we worked hard to push the boundaries in bigquery serverless architecture and its storage technology to turn them into an advantage for you so today we are excited to announce bigquery editions with two groundbreaking Innovations to optimize for the best price performance in this industry with big query additions you can access different tiers to match your data needs standard edition it's a great choice for new customers a talk SQL workloads and data exploration Enterprise Edition offers Advanced functionality such as embedded machine learning in-memory caching full text search and advanced security features an Enterprise Plus Edition is for Mission Critical workloads with cross Regional High availability and to give you the options that you need for any scenario you can mix and match additions across the same data and to increase predictability and reduce costs editions offer a single or multi-year discount in addition to out pay as you go option the first groundbreaking innovation in additions is auto scaling very simply put customers can scale dynamically to take advantage of bigquery serverless infrastructure and scale up or scale down in a matter of seconds so instead of spinning up virtual machines that take time to start and stop and that overprovision capacity and a drive up costs bigquery intelligently optimizes performance by dynamically adding resources while your queries are running compute resources are added on a fine granular compute unit basis matching the demands of data workloads in real time it gives you exactly the capacity when you need it for the time that you needed and auto scaling is not only for spiky workloads it will deliver up to a 40 compute efficiency gain over our current fixed capacity offering customers like PayPal are using this Market leading capability to make sure they respond to their business needs dynamically the second big innovation is with bigquery storage model today we are announcing compressed storage with bigquery additions compressed storage helps manage costs and grow your data footprint all at the same time so you can scale your data without ever worrying about costs and this is the result of over decade of innovation and that includes proprietary columnar compression along with bigquery's automatic data sorting Plastering and compaction beam has gained a compression rate of more than 12 to 1 and they can store more data at a lower cost so they can solve the most complex security problems for their customers all of these Innovations provide customers with flexibility predictability and price performance advantages to remain agile in the most uncertain times the second opportunity is opening up your data and increasing the level of financial control security and data privacy all at the same time so let us start with bi how much have you struggled with inconsistent data are you okrs and kpis hard to compile this data inconsistent like from one meeting to another that's what we hear from Enterprises and that's why we are announcing the preview of looker modeler a new way to extend a single source of proof across your Enterprise and lucra's Innovative semantic layer and Metric layer will allow you to power bi tools such as power bi thoughtspot connected sheets Tableau and liquor studio all with trusted data and trusted metrics this is made possible by your new open SQL interface to the looker semantic model and any tool that works with SQL will be able to connect to Lookers modeling service local modeler Works across Cloud databases querying the freshest data possible and it enables a new level of financial control by allowing you to manage how tables are getting queried which improves your data privacy cost and security we have also heard that you want to share your data beyond your Enterprise boundaries with Partners suppliers manufacturers and Distributors and so we have turned to privacy safe data sharing and essential in today's globalized hyper-connected world of business today we are announcing our plans for bigquery data clean rooms to help the world's largest Enterprises to share and match data in a privacy Safe Way without having to move or copy the underlying data with this new innovation you will be able to power your ads and marketing Networks in the future you will be able to use bigquery data clean rooms to upload your first party data and combine it with your ads campaign data to improve measurement all by preserving privacy protections for healthcare companies this might enable sharing genomic information across partner Networks Financial Services may be able to share highly sensitive data and risk data to detect fraud and support anti-money laundering with big query data clean rooms and our partners we will deliver a strong marketing ecosystem lytics is a customer data platform built on bigquery to help customers activate insights across marketing channels integrates with bigquery to provide privacy safe data orchestration and their own data clean room service life ramp on Google Cloud will enable privacy Centric data collaboration and identity resolution right within bigquery to drive more effective data Partnerships now let us talk about faster time to value for all of your data projects let us make migrations fast let us make the development of machine learning models easy and let us make data broadly available we know the journey to the cloud has been hard you have to do code rewrites lifting data rewriting ETL pipelines we saw this journey and we have heard your struggles and we made the journey to the cloud better for you we focused on delivering migration accelerators and services in June of 2022 that have helped hundreds of customers by automatically translating over 9 million statements of code to bigquery CNA Insurance moved a 20 year old data warehouse to the cloud to gain a global 360 view on all of their data with real-time data feeds into bigquery PayPal migrated more than half of their teradata system and during peak times they were processing 1 000 payments per second with bigquery they were able to scale up or down and manage peak times effectively now let's move from migration to Innovation with the latest machine learning and AI we are bringing Ai and machine learning to all of our users who are working on data it is time to make data work faster just imagine having a conversation with your Enterprise data by leveraging the work that June spoke about earlier we are embedding generative AI right into data and analytics workflows in the months ahead you can access preview features that will bring generative AI right into bigquery and that includes new Integrations with vertex AI to use Foundation models easily accessible from bigquery for all of your text and document data and Beyond text analysis imagine assistive data exploration imagine a no code experience to write SQL imagine an intelligent developer assistance and much more yes it is all coming to Big theories so watch this space the best thing is that you can start right now in fact build a machine learning usage in big query has grown 200 in the past year with hundreds of millions of models in test and in production and to bring machine learning to your data we are announcing new Innovations to bigquery machine learning for model inference this includes state-of-the-art pre-trained vertex AI models imported models from Frameworks such as pytorch are an option to host your own model right within vertex AI but we can't do it all alone together with our partners we are stronger and we are excited to announce that we are growing the data ecosystem to give you more choice and flexibility today over 900 Partners put their trust in Google's data cloud our partner Crux informatics will make over 1 000 new data sets available on analytics Hub with plans to increase it to over 2000 this year Starburst has partnered with us to enable customers to query data anywhere across cloud data lakes or on-premise sources data robot brings their notebook experience to build and machine learning models faster and speed up the time of experimentation neo4j integrates their graph database for advanced graph analysis and outlier detection and thoughtspot on Google Cloud provides AI search capabilities to help users search and retrieve insights at speed thank you to all of our partners for enabling an open Unified and intelligent data ecosystem with all of the innovations that we have delivered today we know that you can achieve a greater flexibility with control for predictable and unpredictable needs privacy safe data sharing and governed insights across your Enterprise and a fast time to transformational business value so now let's see more from Andy gutman's on our strategy for building intelligent data Rich applications [Music] Garrett today every organization is going through some form of digital transformation and at the heart of this always-on digitally connected world are mission critical data-driven applications powering these applications are operational databases that must be reliable available performant and secure at Google Cloud our mission is to provide a unified open and intelligent data Cloud that helps every organization transform their business our Google database engines support Open Standards are designed for the cloud and are proven to power some of the most demanding workloads on the planet for example spanner consistently processes over 2 billion requests per second with customers like Walmart and Kroger bigtable has more than 10 exabytes of data under management with customers such as Equifax and firestore apps power more than one billion monthly active end users using Firebase oath with customers like Forbes this is why customers put their trust in Google Cloud databases saber a travel technology company chose bigtable and spanner as the foundation for their Airline reservation system which serves more than one billion Travelers annually we believe that open data ecosystems are at the heart of digital transformation we're committed to open source and Open Standards and offer managed services that are fully compatible with the most popular open source engines such as Cloud SQL from MySQL and postgres and memory store for redis in an IDC survey Cloud SQL customers achieve an average three-year Roi of 246 percent thanks to the value and efficiencies this fully managed service delivers and therefore it's no surprise that more than 90 percent of Google Cloud's top 100 customers use cloud SQL one example Manhattan Associates whose technology connects 2 billion people to 20 billion consumer choices runs hundreds of cloud SQL instances by moving to a managed service they were able to reduce unplanned downtime by 83 percent while continuing to deliver exceptional availability on their SAS platform our commitment to open source and Open Standards even helps you break free from Legacy proprietary databases with expensive and restrictive licensing the Renault group needed to modernize their entire portfolio and migrated 70 in-house applications from Oracle to Cloud SQL the migration has now resulted in lower costs and three times Improvement in performance and just last year we introduced alloy DB our fully managed postgres compatible database built to help you cost effectively modernize from Legacy proprietary databases or scale existing postgres workloads in our performance tests allowedb is more than four times faster than standard postgres and two times faster than Amazon's comparable postgres compatible service for transactional workloads since its initial launch lodb is now generally available across 22 regions with added capabilities such as point in time recovery cross-region replication and Integrations with data stream and database migration service let me tell you about a couple of customers who've already realized many of the benefits of Ado EDB b4a a Brazilian beauty technology company migrated their databases from Microsoft SQL Server to loadb and so performance accelerate by 90 percent across their queries fluidify a web 3 platform providing auditable decentralized financial data migrated 2 billion web3 analytics insights from Amazon Aurora to loadb and as seen three times faster performance and 60 lower cost and bring their data feed directly to the bigquery community we want to help ensure you have the best experience when using partner Solutions without ODB So today we're excited to announce the launch of Google cloud ready for Alloy DB a program that recognizes partner solutions that have met stringent integration requirements without ODB you can now save time and money associated with product evaluations and focus more time on modernizing your databases state so many customers want to get off Legacy databases sometimes they can't move as fast as they want their workloads are restricted to on-premises Data Centers due to regulatory or data sovereignty requirements or they may be running their application at the edge such as a retail store customers tell us they want to eliminate unfriendly licensing relationships expensive fees and vendor lock-in faster than they're able to re-platform everything to the cloud and they asked us to find a path to support In-Place modernization without ODB as a stepping stone to full Cloud migration what customers want is the support of a tier one provider and the performance and manageability needed to accelerate their modernization initiatives so we decided to do something unusual for a cloud provider we decided to take some of the best of iwdb the best of our high-end Enterprise support capability and deliver a high performance downloadable version of battery DB that customers can leverage for In-Place database modernization as part of their Journey to the cloud today I'm thrilled to announce the technology preview of alloy DB Omni a Google supported high performance downloadable postgres database at a fraction of the cost of Legacy databases that's right alloy DB Omni meets customers wherever they are including on premises at the edge on developer laptops in any cloud in addition lodb Omni will also be supported in Google distributed Cloud hosted for customers who need to run it in isolated environments powered by the engine underlying the Ado ADB cloud service allo DB Omni is more than two times faster than standard postgres for transactional workloads and delivers up to a hundred times faster analytical queries lodb Omni will offer Integrations with vertex AI REI and ml platform to enable model inferencing directly within the database transaction developers will simply call vertex AI models using SQL to power their intelligent transactions are you ready to try out ODB Omni you can download the tech preview today and give it a spin we look forward to your feedback with ludb Omni customers can now modernize their legacy databases in place as part of their Journey to the cloud now we know database migrations can be daunting so we've invested in a comprehensive database migration program to give you confidence and ensure that your experience is fast easy and cost effective every migration Journey starts with an assessment different databases a proper assessment analyzes the workload characteristics complexity risk and cost associated with each database migration as part of the database migration program we're excited to announce Google Cloud's new database migration assessment tool which provides easy to understand reports that demonstrate the effort required to move to one of our postgres databases whether alloy DB or Cloud SQL as you can see these reports help identify which databases can be automatically migrated partially automated or need more significant refactoring deeper analysis shows details on database workload characteristics that are used to make recommendations for the Target database it even estimates a migration effort the cost and anticipated return on investment for the migration our tooling and methodologies are complemented with deep expertise by our partners our Professional Services and our migrations black belt team our database migration program brings all these great resources together along with special incentive funding to offset some of your migration costs if you are interested in accelerating your move to Google Cloud databases contact us today as you can see from what we've spoken about today building a unified open and intelligent data cloud is core to our strategy enabling intelligence throughout your organization to all our customers thank you we're so grateful that you're all on this journey with us tune in to all our sessions at the data cloud and AI Summit for more details on the announcements you've heard today and dive into our breakout and partner sessions and access Hands-On content on the website and finally join our data Cloud live event series happening in a city near you thank you [Music] [Music] hello and welcome to the Gen AI app builder session at data cloud and AI Summit my name is Kalyan pomothy and I'm a product manager in the Google Cloud AI team joining me today I have my colleague Benazir Foote machine learning specialist customer engineer at Google Cloud thank you for joining me thanks Kalyan always a pleasure to work with you likewise all right let's Dive In large language models promise an incredible opportunity for Industries to innovate and create new business value the enable companies to reach larger audiences and enable knowledge workers to access complicated information faster here we'll be discussing how generative AI will enable Enterprises to engage with their internal and external customers and how developers can build generative applications like never before chat search and generative AI are all powerful tools in the Enterprise yeah they can be challenging to implement and create wonderful customer experiences for example users now expect to be able to describe an idea imperfectly but technology being able to anticipate their needs and provide personalized service consumers are moving from a simple query and response to a more human-like conversational experience by using words images and sounds as we do in the real world and with these changes consumers still expect to have a safe trusted and credible online experience these new ways of interacting with technology present many challenges in the Enterprise Legacy systems still provide rigid experiences and limited customer understanding they're hard to scale complex to maintain and don't provide a high quality user experience overall it could be difficult to deploy Technologies to achieve business impact in a cost-effective fair and ethical way now chat search and generative AI independently add a lot of value in the Enterprise for example conversational AI is wonderful when you can have human-like interactions to solve problems and find information Enterprise search AI is useful when you can scale a wide range of siled data to retrieve accurate and credible information AI is quickly becoming the path to creating new content and experiences to Delight your customers but when you bring all these capabilities together it enables Enterprise developers to build even more delightful user experiences we recently announced gen AI app builder platform which brings together the power of conversation search and generative AI to the Enterprise let's walk through how you can use this new platform to create transformative applications powered with your data within minutes take it away Ben is here thank you Kalyan hi I'm Benazir Fateh an ml specialist customer engineer at Google Cloud to get started with generative AI app builder you can start with a domain name point to a storage location or upload documents you can choose to create a chat bot a search experience or both at the SDS to commonly work together so let's choose our adventure and start with a chat experience let's walk through the user experience of a chat bot built by the fictitious retailer symbol bikes and discuss how it was built a consumer on the simple bikes website asks about investing in a bike that would be good for triathlons and for Community Computing in the New York City the bot takes into account both use cases for the bike and provides an answer the sources of the answer is cited and the consumer is navigated to the road bike section of the site where the suggested bikes are displayed when the consumer asks about handlebar materials and the answer is not found on the site the bot pulls from publicly available websites or general knowledge of the AI model to answer later in the conversation the consumer asks how one of the road bikes compares to their current bike the bot allows them to upload a picture and is quickly able to recognize the model of the bike based on that image next a comparison table is dynamically generated presenting the the answer to the question in the best format based on this user and their question after comparing bikes the consumer is still not quite ready to complete the purchase the bot creates a summary of the conversation for the user to take away and reconsider this summary can also be used when the consumer comes back to the site in the future when the consumer asks for suggestions on a plan for a day of biking that includes ocean views and is within two hours of New York City the bot creates a personalized itinerary to help Inspire the consumer let's step out of the user experience and talk about the configuration options that were used to build this the developer of this experience enabled generative responses to be sourced from the symbol bikes websites other publicly available websites and general knowledge from the AI model in this example both summarize answers from the site's content and newly generated answers were enabled another option would have been to ground answers only in the content from the symbol bikes website and documents but an enable generative responses like summarization in addition there are other detail settings like restricting terms and topics that are also available to Developers back to the user experience when the consumer returns to the site they are greeted in context of their last visit by being asked if they have any other question about the road bike that they were considering before in a single response the consumer says they want to buy the bike and ask if local delivery is offered and states the delivery time that works best for them the bot can understand all of this and it responds with delivery options that meet the user's requested time frame as you can see we have moved Beyond just providing answers to a user's questions and now we are helping them complete transactions like in this case making a purchase and booking a delivery appointment let's look at how a developer can enable the bot to complete tasks like placing orders and scheduling appointments generative AI app builder provides step-by-step conversation orchestration with three ways to add these type of task flows to your Bot one pre-built flows cover common tasks like authentication checking and order status and more you can drag and drop them onto a canvas and complete a basic form to enable them second you can describe what needs to be done and what information to collect using natural language and third you can visually map out business logic and include the pre-built and custom tasks you created the graph is simple as the AI handles guiding the user conversation now that we've covered generative AI for chat experiences Kalyan will show you how to put the Enterprise chat search experience into action that's right next in our adventure we'll showcase an Enterprise search experience we want to enable developers to imagine and build generative search experiences for many industry applications here is an example built for an external customer facing foodie app here the user wants to search for recipes in the app they say that they want to cook something vegan with broccoli and carrots the generative search platform understands their intent in natural language and shows the most relevant search results based on that understanding now the user can also ask follow-up questions in this example we see the user providing a location on a map the search platform understands multimodal inputs so in this case the text and the location of the map and generates personalized recommendations all these experiences can be built out of the box with the Enterprise search platform here's another example this time for an internal Search application to support an analysts doing Financial research let's say they want to search symbols data about the challenges faced by semiconductor companies due to the rising interest rates and inflation like before they can enter their question in natural language and the search platform understands their intent and searches over symbols data sources to provide relevant search results these results are grounded with Source citations and attributions the user can also see synthesized summaries generated from data across multiple sources within symbols data finally the platform provides personalized recommendations relevant to the user's search all of this out of the box without much setup so what use cases will this platform enable you to build we see opportunities across a variety of internal external use cases addressing the needs of customers staff and machines you can build search experiences for external users in your websites or internal Enterprise or technical applications the Enterprise Search application is designed to foundationally support building generative AI applications for any information retrieval use cases the Gen AI app builder platform for Enterprise search will serve application developers and AI practitioners for application developers who want an easy setup experience we offer out-of-the-box search and chat features for search developers who want to personalize and build multimodal capabilities we offer configurations and more advanced features and for AI practitioners who want to build on top of the search platform we have a build your own search mode with Advanced model training tuning and embeddings just like with conversational AI all of the features and capabilities will be easily configurable and controllable we can manage these configurations for features such as fastening and autocomplete summarization and conversational experiences and while you change these settings you can see a preview of these configurations to the right and of course Enterprise search will integrate with the vertex AI platform where an AI practitioner can leverage their institution's domain knowledge to build their own search models embeddings and knowledge graphs now these customizations can then be exported to the Gen AI app builder to power the Enterprise search engines generative AI app builder is where Google's latest foundational models Enterprise search and conversation AI capabilities all come together to help you easily build credible controllable and customizable next-gen experiences I want to thank you for joining me today Benazir And discussing the Gen AI app builder and thank you to all our viewers for joining for the Gen AI app builder session for the data cloud and AI Summit until next time [Music] okay [Music] hi and welcome to build customize and deploy Foundation models in vertex AI session my name is Keelan McDonald and I'm a product manager in Cloud Ai and I'm Warren Barclay I thought we'd start with like a brief history of AI just talk a little bit about some of the background that exists there and so large language models have really kicked off this new era of Technology Innovation and this whole new world for developers and AI practitioners so let's just kind of go back a few years and see where we were and now where we are in the early days you had these small models that were specifically made for specific tasks and so those specific tasks were not very elastic you know it's things like I'm gonna train something for time series prediction or things like that and so you spent a lot of time making sure that you're optimizing these models for one thing moving on into like 2015 to 2020 what we saw was that this invention of Transformer models Transformer models did a couple things one is it really made it much more easy to customize these models they're much more elastic and you could use them for different domains and so it was relatively straightforward on how to do this but we still weren't kind of there yet until we got to really 2022 where we started to see compute get more efficient and effective and so you know things like tpus which we had had for a while really allow inference to happen at scale and with kind of the cost Effectiveness that needed to happen you know today we have models that are from 1 billion to like hundreds of billions of parameters and they don't necessarily need to be trained per t