Google IO to feature AI updates, showing off PaLM 2 LLM

Artificial intelligence will be a major topic in Google CNBC has learned that its annual developer conference is on Wednesday, where the company plans to announce a number of generative AI updates, including the launch of a general-use language large model (LLM).

According to internal documents about Google I/O seen by CNBC, the company will unveil PaLM 2, its newest and most advanced LLM. PaLM 2 includes more than 100 languages ​​and operates under the internal codename “Unified Language Model”. A wide range of coding and math tests were also taken as well as creative writing and analysis tests.

At the event, Google will make announcements on the topic of how AI can “help people reach their full potential,” including “generative experiences” for Bard and Search, documents show. Pichai will speak to a live crowd of developers while showing off his company’s AI developments.

The updates come as competition in the AI ​​arm race heats up, with Google and Microsoft It is racing to integrate AI chat technology into its products. Microsoft is using its investment in ChatGPT creator OpenAI to boost its Bing search engine, while Google has quickly mobilized to try to integrate its Bard and LLM technology across different teams.

Google first announced the PaLM language model in April of 2022. In March of this year, it was released The company has launched an API for PaLM Along with a number of AI-powered enterprise tools, it says it will help companies “create text, images, icons, video, audio, and other simple natural language prompts.”

last month google He said This medical LLM called “Med-PaLM 2” can answer medical exam questions at an “expert physician level” and is accurate 85% of the time.

Google also plans to share developments in Bard and Search with “generative experiences,” including use of Bard in coding, mathematics, and “logic,” as well as expansions into Japanese and Korean languages, the documentation shows.

The company has been working on a series of more powerful Bard models, and officially launched the tool as a trial in March.

Internally, the company has been working on a multimedia version called “Multi-Bard,” which uses a larger data set and solves complex math and coding programs, according to separate documents reviewed by CNBC. The company also tested versions called “Big Bard” and “Giant Bard”.

Google also plans to expand on the Workspace AI Collaborator, including discussion of creating templates in Sheets and creating images in Slides and Meet products. In March, the company said it would give access to the AI ​​capabilities of Gmail and Google Docs to a small number of users as part of testing, with plans to bring additional AI features to the Meet and Sheets and Slides apps.

One image, seen by CNBC, showed a presentation sidebar with a chat box allowing the user to enter text with the option to “create” an image based on the words.

Additional updates include use cases for the Google Lens image recognition tool. The company will show advances in “multi-search” for camera and audio, after last year allowing users to ask questions about what they’re seeing in photos.

Outside of the field of artificial intelligence, Google will show off its new foldable phone, the Pixel Fold, as previously reported by CNBC. The company claims that the Pixel Fold will have “the most durable hinge in a foldable phone” and offer the option to replace the phone. Google plans to market the Pixel Fold as waterproof and pocket-sized.

A Google spokesperson did not immediately respond to a request for comment.

He watches: Google “on borrowed time”