(Original title: Heavy! Google released the latest AI technology: the launch of a super virtual machine; the evolution of the new model collaboration tool family bucket, cloud, search, email, etc.)
On Wednesday (May 10), local time, at the 2023 Google I/O conference, Google demonstrated a series of the latest technological achievements in the field of artificial intelligence (AI).
Bringing generative AI to products like search engines
According to the Financial Associated Press, Pichai, CEO of Google’s parent company Alphabet, said: “We are reshaping all core products including search.”
According to Pichai, Google is integrating generative AI into its search engine and other products, such as Gmail, which can generate draft messages, and Google Photos, which can make significant changes to images.
The new search service, dubbed Search Generated Experience (SGE), will display AI-generated responses at the top of search results designed to answer user queries while summarizing key information and linking to website resources.
Cathy Edwards, vice president of Google, said that users in the United States can experience the new search service through the waiting list in the next few weeks. This is an experimental phase, and Google will monitor the quality, speed and cost of search results.
The new model PaLM2 drives Bard, the collaboration tool family bucket evolves, and the cloud, search, email, etc. are all in one place
This time, Google officially released the new general-purpose large language model PaLM2. PaLM2 is an upgraded version of the model that drives the AI robot Bard, which can generate a variety of text responses to users. Google says it can speak 100 languages and is good at math, software development, language translation reasoning and natural language generation.
Introducing the new large language model, Pichai said that PaLM2 has some advantages in specialized fields such as medicine. It can answer medical exam questions at an expert level.
Zoubin Ghahramani, vice president of DeepMind, Google’s AI research lab, called PaLM2 “better than our previous state-of-the-art language model.” PaLM2 uses Google’s custom AI chip and runs more efficiently than the original PaLM. PaLM2 can use more than 20 programming languages such as Fortran, and it can also use more than 100 spoken languages.
At the same time, Google announced that its products widely apply the functions of AI technology, from various collaboration tools such as documents, to various services such as email, search, and cloud, so that the AI functions supported by the new large model can be fully integrated in various products.
Workspace’s DuetAI AI search feature Converse
Google said that Google’s collaboration and productivity tool Workspace will soon allow users to use AI to generate complete documents and fill out spreadsheets. The AI service, called DuetAI, will be available to subscribers later this year.
DuetAI allows users to integrate AI services into documents, sheets and slides. For example, users can ask AI to help write resumes, organize corporate charts, or generate images for slideshows. It can help users write articles and sales promotional materials, and put them in Google Docs for everyone to share. Google Slides can generate AI images within the app.
Google has launched a generative AI feature called Converse for its search service. In the past, users who wanted to know the answer to a question would get multiple search results from Google and had to synthesize the results themselves. Converse can synthesize search results for users. Specifically, it provides a kind of “snapshot” that includes some generative AI content blocks of search result text or images, with relevant links below it.
Ads will be integrated into Converse search results, Google said. Like some of Workspace’s AI features, Converse will offer a “conversation mode” in which users can “ask” follow-up questions.
While Google didn’t say when Converse will be available or in which regions it will be available, the feature is part of the Search Generation Experience (SGE) effort. Labs users will be able to use it in the coming weeks, Google said, and can be accessed by clicking the Labs icon in the Google app or in the desktop version of Chrome.
Google launches A3 hyper virtual machine
Google announced the launch of A3VirtualMachines, paired with Nvidia H100, for training and serving state-of-the-art AI models.
It is reported that an A3VirtualMachines is powered by eight H100GPUs based on the Nvidia Hopper architecture, which is three times faster than its predecessor chip, the A100.
Roy Kim, director of product management at Google, and Chris Kleban, group product manager, said state-of-the-art AI and machine learning require massive amounts of supercomputing power, which needs to be provided by infrastructure tailored specifically for it.
Google releases new phone Pixel 7a, first foldable phone Pixel Fold tablet Pixel
In terms of hardware, Google released a new generation of Android phone Pixel7a, priced at $499, equipped with Google’s self-developed TensorG2 chip, running memory 8GB, more affordable than Pixel7 and Pixel7Pro.
Google also released its first foldable phone, the Pixel Fold, which costs $1,799 and is also equipped with a TensorG2 chip.
Google announced that from this Wednesday, the new version of the tablet Pixel will accept pre-orders, also equipped with the TensorG2 chip, with a starting price of $499, and two versions of 128GB and 256GB of memory are available.
Disclaimer: The Securities Times strives for truthful and accurate information, and the content mentioned in the article is for reference only and does not constitute substantive investment advice, so operate at your own risk
Download the “Securities Times” official APP, or follow the official WeChat public account, you can keep abreast of stock market trends, gain insights into policy information, and seize wealth opportunities.