What Google I/O Shanghai Reveals About Cloud and Generative AI

Schedule and Venue

The conference spanned two days, with the first covering Mobile and Cloud topics and the second focusing on AI and Web. In addition to keynote speeches, workshops allowed hands‑on product experiences; the author mainly followed Cloud, AI, and Web sessions.

Cloud & AI

Google's Position

Google Cloud is a Gartner 2023 Leaders quadrant player alongside AWS and Microsoft, ranking fourth in market share in 2022 with a 40% annual growth rate, close behind Alibaba Cloud.

Google AI has a deep legacy, from Word2Vec to Transformers, and early involvement in large language models.

Based on Google Cloud, fully leverage generative AI to empower enterprises and developers.

Foundation Models

Vertex AI offers foundation models such as text‑bison for classification and summarization, embedding models, and code‑bison for code generation, with some models supporting up to 32k input tokens and 8192 output tokens.

Vertex AI Platform

Built on these models, Vertex AI provides a machine‑learning platform for deploying, training, and debugging large models, along with a Model Garden marketplace for publishing and consuming models.

Gen App Builder

Gen App Builder enables interactive, dialogue‑driven app generation (demo video linked).

Developer‑Focused Duet AI

Duet AI offers code completion, generation, and Q&A capabilities similar to code‑GPT, integrated via IDE plugins. A demo showed large‑model‑assisted Kubernetes operations.

Workshop: Vertex AI Generative AI

The workshop let participants try Vertex AI hands‑on, similar to OpenAI Playground, with full feature support.

Structured prompting with multi‑sample inputs yields precise sentiment classification (positive/negative/neutral) and limits hallucinations.

Exported prompts can be run as code; a Google Colab notebook generated a Tetris game in under five seconds on a non‑GPU cloud instance.

Summary of Cloud & AI Session

Google Cloud X Generative AI targets enterprises, offering AI infrastructure for customization, debugging, and deployment, along with data security commitments. For entrepreneurs seeking large‑model capabilities, Google’s ecosystem is highly attractive.

Web

The Web track covered new DevTools, PassKeys, and Privacy Sandbox, with a focus on improving user experience to drive business growth.

Enhancing Web UX for Business Growth

High conversion rates on marketing pages benefit directly from performance improvements.

Using SSR and View Transition API brings Web experiences closer to native apps, reducing checkout time and abandonment.

Web Vitals dashboards and A/B testing enable data‑driven optimization.

Workshop: MediaPipe

MediaPipe is an open‑source on‑device ML library for vision, text, and audio, with an online demo available.

Overall Summary

While the event didn’t exceed expectations, the Cloud & AI sessions were impressive, showcasing stable large‑model outputs and developer‑friendly tools. Combined with vector databases, the offerings are friendly for application‑level experimentation, even more so than internal solutions at Ant Group.

Duet AI’s assistance for development, operations, and data exploration provided valuable insights.