Google Updates Cloud TPU V5e Accelerator

At the Google Cloud Next ’23 held at San Francisco’s Moscone Center, Google predominantly continued its focus on artificial intelligence technologies as highlighted during Google I/O 2023. This emphasis seeks to fortify Google Cloud’s technological collaboration spectrum and integrate its technologies with numerous industry players to propel the advancement of AI applications.

Google unveiled the new Cloud TPU v5e accelerator, aiming to cater to the training and inference acceleration demands of small to medium-sized businesses in AI applications, offering superior computational cost-effectiveness. Moreover, Google is leveraging the NVIDIA H100 GPU to augment computational efficiency, introducing the A3 cloud virtual machine, equipping businesses to cater to the computational requirements of extensive natural language models.

Given the current multi-cloud application trend, Google announced the Cross-Cloud Network service, enabling Google Cloud customers to connect various applications across cloud environments while ensuring data privacy. Through the Google Distributed Cloud infrastructure, Google Cloud customers can execute various AI or data computation workflows using Vertex AI and AlloyDB Omni services.

Regarding updates to the Google Kubernetes Engine (GKE) managed container service, Google introduced the enterprise version of GKE, enabling businesses to effortlessly establish multiple clusters and scale AI and machine learning computations horizontally.

Google also upgraded its large-scale natural language model, PaLM, increasing input parameter length fourfold and supporting 38 languages. Over 100 language interfaces are currently in preview. The coding auto-generation tool, Codey, now boosts content generation accuracy by 25%, catering to major market languages. For the auto-generating image application service, Imagen, while bolstering its features, a Style Tuning function has been incorporated, allowing brands to swiftly produce 10 reference images for brand promotion and product design.

In terms of expanding Vertex AI functionalities, Google added support in its model database for Meta’s open-source model LLaMA 2, the AI startup Anthropic’s Claude 2, established by former OpenAI staff, and the UAE government-funded technology innovation institute’s open-source AI model, Falcon. This allows Google Cloud users easier access to these model resources and swiftly establish AI application services.

At the Google Cloud Next ’23 event, Google announced technological collaborations with AI tech companies like AI21 Labs in Tel Aviv, Bayer Laboratories, employee experience platform Culture Amp, FOX Sports, General Electric, General Motors, the Hospital Corporation of America, the largest pancake breakfast chain IHOP, AI startup Runway AI, Meditech, Morgan Stanley Capital, the world’s largest amusement park chain Six Flags, and online wedding planning platform The Knot.

Simultaneously, Google Cloud is integrating its AI technology with SAP and digital document signing service DocuSign. Deep collaborations have also been established with data marketing firm Acxiom, Bloomberg News, data analytics firm CoreLogic, consumer credit surveyor Equifax, consumer research and retail data monitoring agency NIQ (after their merger with GfK), and credit monitoring service TransUnion. This partnership endeavors to create and train a variety of AI technology application services through the Google Cloud platform.