Google I/O 2023: As Far As The AI Can See
As expected, AI was center stage at Google I/O 2023 and much of it was generative AI (GenAI). Google has always been powered by AI and now it continues to advance and expand the capabilities as well as productize them. What was power under the hood is now in the hands of all – consumers, developers, and enterprises – in new and different ways. As the GenAI race accelerates, Google has added more fuel to its engine. Let’s unpack!
A Raft Of AI In Productivity Apps
GenAI made an appearance in every act at Google I/O, fueling productivity, creativity, and even some self-indulgence. The general user of Google products will see GenAI capabilities embedded and ubiquitous in all the widely used applications in Workspace – through Duet AI, Docs will have a “sidekick” to help you write better, Sheets will auto-generate templates using natural language prompts, and Slides will provide better images to enhance storytelling. And you can even let “someone” write your email for you. There’s magic at your fingertips to edit photos, personalize your device with AI-generated images, and plan your trip with immersive experience on Maps. GenAI will also provide synthesized information about your Google Search and allow you to ask follow-up questions (in addition to usual search results).
While these capabilities may seem incremental and familiar, Google has provided a strong response to the market by productizing GenAI capabilities in the Google consumer ecosystem. So, if you’re a Google user, what it has done will certainly make your life easier. But, if you’re not already there, Google hasn’t given you a reason to move yet.
AI Powers Developers And Enterprises
Enterprises need to focus on the launch of PaLM 2 and Vertex AI. Google’s Pathways Language Model (PaLM) works behind the scenes to power and integrate AI across its product lineup. Last week, Google also announced the release of PaLM 2 in preview. PaLM 2 comes in a few sizes, including one to fit on device. Enterprises get the power to fine-tune Google models with private domain data and to build GenAI models that are ostensibly more responsive and computationally/fiscally efficient. Google mentioned Sec-PaLM and announced Med-PaLM 2, trained on security and medical domain data respectively. Google adds GenAI to many APIs and provides a simpler way to use large language models hosted/deployed on Vertex AI. PaLM 2 will power three new models in Vertex: Codey, a code completion and code generation model; Chirp, a universal speech AI; and Imagen, a text-based image generator. Vertex also powers Bard, Google’s AI chatbot. In addition, there’s new Nvidia hardware to support AI and other workloads on Google Cloud. Integration with external/internal tools will allow companies to accelerate their go to market.
Last week, Google reminded the world of its AI-driven past and put GenAI center stage for the future. Now that the AI genie is out of the bottle, it’s also working on techniques to identify and prevent misuse of AI technology: watermarks, metadata, and a standard to let web creators identify and prevent use of their content (all coming “soon” or “in a few months”).
A few recent missteps made it seem as if Google’s public-facing AI technology was lagging. With the integrations shown at Google I/O, it’s clear it has caught up to the pack.
Read Andrew Cornwall’s blog about non-AI announcements at the Google I/O.