Content
summary Summary

Google opens access to internally developed AI models and launches PaLM and an image model. In addition, new partners such as Midjourney join the Google Cloud Program.

In 2017, Google researchers published the Transformer paper, an architecture that has since spawned numerous large AI models and is a central foundation for ChatGPT or Midjourney. Google itself has been reluctant to release its own models, but the successes of OpenAI and Microsoft have caused the company to reconsider. Now Google is releasing API access to the PaLM language model and generative AI for images in the cloud.

Google brings generative AI via API and to the Google Cloud.

Google's PaLM is one of the best language models, and the company is now offering access via API and Google Cloud. For now, it's an efficient model in terms of size and capabilities called Bison. Other variants are to follow, and access will also be extended via a waiting list for the time being.

The API access is meant to be an easy and secure way for developers who want to experiment with AI to build on the language model. In addition, Google offers a tool called MakerSuite, which it says can be used to quickly create prototypes. Other features such as prompt engineering capabilities, synthetic data generation, and customization of custom models are expected to follow.

Ad
Ad

For developers who want to build their own generative AI models and applications, Google provides access through the Google Cloud. PaLM, an as-yet-unknown image model, and other open source models are available there, and Google says third-party models will be available soon. The new models are also supported by Vertex AI, Google's solution for easily building and deploying AI models without programming skills.

Through Vertex AI, PaLM can also be specialized for specific contexts or fine-tuned with additional training data.

Midjourney, AI21Labs and Osmo now rely on Google Cloud.

Google's cloud business is also welcoming new AI companies: Midjourney, a generative AI for images startup, AI21Labs, a language modeling startup, and Osmo, an AI-powered olfaction startup. This brings more companies to Google as customers, which some might perceive as competitors - similarly Anthropic and Cohere also rely on Google's cloud.

Google also announced a number of other partnerships today with companies including Anyscale, Weights & Biases, Labelbox, and Replit. Leading Google partners including Accenture, BCG, Deloitte, McKinsey, Quantiphi, and TCS have also committed to expanding the AI-based consulting and implementation offerings available on Google Cloud, according to Google.

Today, Google also announced plans to bring generative AI to all Workspace applications, including Gmail, Docs, and Slides.

Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Recommendation
Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • Google launches PaLM API and cloud services for generative AI models. For now, there is an efficient PaLM variant and an image model.
  • The generative AI models will also be available in Google's Vertex AI.
  • Google Cloud also announced a number of new partnerships.
Max is managing editor at THE DECODER. As a trained philosopher, he deals with consciousness, AI, and the question of whether machines can really think or just pretend to.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.