AI: Recent Trends and Applications, Emerging Communication and Computing Steam Power 1784 Electricity 1870 Information Technology 1969 Artificial Intelligence Today AI
developers could AI address? of developers spend 30 minutes/day searching for solutions. encounter a knowledge silo at least once a week. spend more than an hour searching for solutions each day. Dev Efficiency
primary application or service you work on, how important is the role of Artificial Intelligence (AI) in contributing to each of the following tasks today?” *Interval provided for each graph represents 89% credibility interval. Provided to show inherent uncertainty in our estimates. Analyzing data Writing code blocks or data functions Analyzing security Learning new skills Optimising code Analyzing logs Motoring logs Identifying bugs Writing tests Writing documentation Scaling running services Responding to incidents Managing my coding environment Recovering from incidents Organizing user feedback Making decisions Collaborating with my teammates Managing projects Solving file path issues % of respondents 0% 20% 40% 60% 80% 100% Did NOT select “Not at all important” Selected “Extremely Important”
System Starts writing code through IDE Gen AI suggests and autocompletes code from custom libraries Code Search for trouble- shooting Translate legacy code segments from an old language to a newer one Gen AI generates unit tests Gen AI debug issues Commits the code to a repo Github auto PR summary support Code efficiency review with Gen AI Analyzes the suggestions and accepts Pushes code to dev/staging Bug occurs Dev log analysis and debugging recommendations Analyzes and fixes prod bugs Release notes summary Pushes code to fixbug
owner at a healthcare company. She wants to build a personalized patient healthcare experience. Normally she takes weeks to do research, executive pitching and gathering of UX teams. She is ready to use Gen AI to help accelerate the product definition phase using Duet AI to do initial market analysis, generate her product pitch and one page strategy.
Teams accelerate: ➔ Market trend analysis ➔ Customer persona generated Ideation ➔ User Story Creation and Acceptance Criteria ➔ Strategic proposal decks Features subject to change
a healthcare company. He wants to use his design skills to create personalized patient healthcare experiences that improve the quality of care and patient outcomes. Normally he takes weeks to do user research and interviews, build mockups, code user prototypes and along with Jen write user stories.
engineer. He uses Visual Studio Code in his daily work and is new to Google Cloud. He wants to build new product features on GCP and is ready to use Duet AI in Cloud Code to write better code, debug errors and create a reliable and scalable website.
while writing code in the IDE As a developer, I want to write high quality code faster and stay in a flow state longer so that I can be productive translating my idea into code, and not the code itself. Goal: Deeply integrate AI Assistance for developer personas to improve productivity. Assisted development Chat interface for developers to get guidance by describing problems in technical terms As a developer, I want to solve bigger problems, worry less about implementation details, and spend more time on the business logic.
works as intended. He asks Duet AI to suggest an acceptance test based on the gherkin syntax of the acceptance criteria developed by the product team. Magic moment Duet AI generates the acceptance tests based on natural language from the user story Pairing: Suggest an acceptance test
make sure the code works as intended. He asks Duet AI to suggest a unit test plan for the code. Magic moment Duet AI suggest the test plan and elaborate on the edge cases Chat: Suggest a test plan
asks Duet AI to suggest a library to display a QR Code. Duet AI also includes the proper attribution and licensing requirements. Magic moment Duet AI accelerates research time by suggesting and attributing the correct code. Pairing: Reduce research time
ready to expand to new countries. Sam is tasked with translating the website into multiple languages to serve non-English speaking customers. Completion: calling GCP services | Google Translate Magic moment Based on his comments, Duet AI creates a Go function to translate text from English to Hindi. This allows him to write code with functions he's not familiar with.
a code base a dev is not familiar with, they can I ask Duet AI to explain the code so to quickly and effectively understand how to start testing it. Chat: Explain code
made aware that the suggested code falls under a permissive license, which requires attribution, and provides the needed information. This knowledge helps the organization stay compliant with the license terms Providing code license attribution
be able to quickly retrieve generic, relevant “how to” information about cloud infrastructure, configuration, and best practices. Optimization As an operator, I would like to be able to get generic recommendations on how to optimize resources for cost, performance, security, reliability, manageability and sustainability. Goal: Deeply integrate AI Assistance for cloud admin & operator personas to improve productivity. Assisted operations
their organization. Since Cloud SQL spend has been high, they navigate to SQL to investigate further. Ean opens Duet and notices some sample prompts related to the current page. Get help wherever you are
review the necessity of idle instances seems like a good starting point. Ean clicks the “Copy to Cloud Shell” button to run the relevant command to identify idle instances. With the list of idle instances, Ean and team can review and decide on next steps. Transition to cloud shell Magic moment Duet generates Cloud Shell commands that address the customer’s initial query and with one click those commands are sent to Cloud Shell and executed by the customer.
gets an alert showing issues with the application. She goes to logs explorer to find the source of the issue, but she’s having trouble understanding a potentially relevant log entry since she is new to this on-call rotation. Contextualized log explanations
Analyst. He is a seasoned visualization developer, but his SQL skills lean towards the novice side and he is new in BigQuery. His main responsibilities include using SQL to produce insights for his operations, marketing and finance teams.
AI provides suggestions of contextual SQL statement completion and contextual SQL infilling for functions and keywords. He sees the default suggestions and cycles through options for additional suggestions. He accepts a suggestion that fits his needs and continues to iterate. Magic moment Duet AI for BigQuery saves him a lot of time otherwise spent on looking up correct syntax.
can also use Duet AI to generate SQL from natural language statements. Lacking knowledge about newer datasets, he decides to use natural language comment syntax and to generate a candidate query. Magic moment Duet AI for BigQuery saves him a lot of time otherwise spent on digesting schema and writing initial query candidates.
not know the format and results of the query, specifically he needs to filter out nulls and round decimals. Using an additional natural language comment he can iterate via new statement generation. Magic moment Duet AI for BigQuery saves him time otherwise spent on looking up syntax and also helps him learn syntax.
development tasks, but is now faced with inheriting code from some other Analysts. He is unfamiliar with the schemas and business concepts in these queries. Using Duet AI Boaz can select the query and get a natural language summarization of the analysis concepts. As he continues to ramp up his SQL skills, but at times struggles with understanding more complicated query syntax. Using Duet AI Boaz can select the query and get a natural language summarization of the SQL syntax concepts. Magic moment Duet AI for BigQuery saves him time by providing both baseline context for queries and prescriptive syntax summarization.
Healthcare AI Vertex AI Conversation AI Enterprise Search Foundation Models Generative AI App Builder End-to-End ML Platform Generative AI Studio | Generative AI APIs | Model Garden Google Cloud Infrastructure GenAI GenAI GenAI GenAI Business users Developers AI Practitioners Google Cloud AI Portfolio To support the needs of AI centric enterprise development
to cost-effectively run AI workloads at scale Support diverse ML workloads with varying hardware needs. Flexibility to use CPUs, GPUs, or TPUs. Strong partnership with Intel, AMD, & NVIDIA. OSS Software Ecosystem Make the most of ML with an open software ecosystem. Easily use open source tools, frameworks, & APIs. Run workloads via VMs, containers, or managed services. Speed & Performance Deliver results faster. Set up ML environments quickly, automate orchestration, manage large clusters, and set up low latency applications. Workbench Training Serving State of the Art AI Manage complexity and innovate. Unlock more value from infrastructure with state of the art AI. Reduction Server (70% increased throughput) Optimized TensorFlow Runtime Matching Engine (few millisecond ANN) Vizier Hyperparameter Tuning Kubernetes Engine Compute Engine BigQuery Dataflow Spark Vertex AI Bigtable
AI/ML Platform on GKE GKE gVNIC Fast Socket Local SSD GCS Fuse TimeSharing Multi-Instance Autoscaling | Placement | Provisioning Kueue: Kubernetes-native Job queuing NCCL Build | Train | Deploy Team 3 Team 4 Team 1 Team 2 Team 5 Workflow and Data Processing Custom Frameworks Compute GPU TPU Storage Network
• Code support for more languages, More holistic assistance ◦ bespoke Terraform architectures from industry best practices • Assistance to help debug specific issues, refactor code, and generate comments ◦ iterative Debugging with Context • Migrate/Upgrade/Translate code languages and framework ◦ migrate code from COBOL to Java • Streamlining tedious or complicated Git processes ◦ Git branch auto summary support ◦ Git auto PR summary support ◦ Git merge auto conflict support ◦ Git large multi-branch sync
- #2 • Codebase documentation and navigation ◦ creating/updating natural language descriptions for all source code files in the project. ◦ Q&A based on the generated documentation. ◦ Identify all places in the codebase related to the given bug or feature request • Full AI assistance across the cloud platform • Low-code / no-code capabilities 10X (100X?)
-#4 • DevOps ◦ Alert and Summarize incidents based on signals from multiple sources ◦ Predictive failure alerting ◦ Allow for analysis and chat of prod logs ◦ Fully automated rollbacks on release failure, no prior planning ◦ Correlate incidents with lines of code recently changed ◦ Suggest changes to code to alleviate incident ◦ Allow SRE to search across bugs/code/external sources to help troubleshoot ◦ Generate postmortem report based on incident/code/troubleshooting data
of high quality documentation* Continuous integration 2.4x* Continuous delivery 2.7x* Trunk-based development 12.8x* Loosely coupled architecture 1.2x* SRE practices 1.4x* AI incorporation 1.5x* Impact of technical capability with high documentation quality Impact of technical capability with low documentation quality Documentation quality helps technical capabilities be helpful * Calculated by:
Your Terms Your Google Cloud Perimeter Duet Services Google Cloud Central Hosting Large Base Model (Frozen) Your Data Your Inference Vertex AI Adapter Layers Your Security Your Content Your Developers Developer IDE Cloud Console CMEK DRZ AxT VPC-SC Assistive Development Assistive No-code Assistive Data Exploration Assistive Operations Assistive Security
craigsdickson
kaminashi
moznion
recruitengineers
akitoshiga
ymotongpoo
bitkey
namiki64
kentapapa
kentarofujii
takahashim
codmoninc
ezaki
akihiro_kokubo
mlcsv
destraynor
skoyamalab
brettharned
aleyda
inesmontani
bethany
holman
baasie
davetheseo
