Image credit : Engadget
A large language model (LLM) for general use will be released, among other generative AI enhancements, at Google’s annual developer conference on Wednesday, CNBC has learned. Artificial intelligence will also be a major topic.
Internal Google I/O docs obtained by CNBC state that the company will present PaLM 2, its most latest and sophisticated LLM. More than 100 languages are supported by PaLM 2, which has been working under the internal codename “Unified Language Model.” Additionally, it has completed a variety of coding and maths assessments as well as analyses and tests for creative writing.
At the event, Google will make announcements on the theme of how AI is “helping people reach their full potential,” including “generative experiences” to Bard and Search, the documents show. Pichai will be presenting his company’s AI accomplishments to a live audience of developers.
The upgrades come as Google and Microsoft compete to include conversational AI technologies in their products as the AI arms race heats up. While Google has moved swiftly to try and integrate its Bard technology and its own LLM across several teams, Microsoft is utilizing its investment in OpenAI, the company that created ChatGPT, to support its Bing search engine.
In April 2022, Google initially revealed the PaLM language model. The business unveiled an API for PaLM and a series of AI enterprise solutions in March of this year, claiming that these will enable businesses to “generate text, images, code, videos, audio, and more from simple natural language prompts.”
Google said last month that its “Med-PaLM 2” medical LLM, which it calls a “expert doctor level” answerer of medical exam questions, is 85% correct.
According to the docs, Google also intends to share improvements to Bard and Search with “generative experiences,” including the use of Bard for coding, arithmetic, and “logic” as well as language expansions to Japanese and Korean.
The technology was formally introduced as a trial in March after the corporation worked on a number of more potent Bard models.
According to independent paperwork seen by CNBC, the company has working on a multi-modal version internally called “Multi-Bard,” which employs a larger data set and resolves challenging arithmetic and coding problems. The business has also tested “Big Bard” and “Giant Bard” variants.
Google also intends to talk more about its “Workspace AI collaborator,” including how it can create images for its Slides and Meet and Sheets apps. A select group of users would be given access to AI tools in Gmail and Google Docs as part of a test, the company announced in March. It also announced plans to add more generative AI features to its Meet, Sheets, and Slides programs.
Also read : For $29 you can learn how to code video games
CNBC saw one image that featured a Slides sidebar with a chat box where users could type text and choose to “create” an image from the words.
Use cases for Google Lens, an image recognition technology, are among the additional upgrades. The company will demonstrate improvements to “multi-search” for voice and camera after enabling users to inquire about what they are viewing in photographs last year.
Google will debut The Pixel Fold, a revolutionary foldable phone, outside the AI industry, as CNBC previously reported. The Pixel Fold, according to the manufacturer, will boast “the most durable hinge on a foldable” phone and will support phone trade-ins. Google intends to advertise the Pixel Fold as a pocket-sized, water-resistant device.