Google unveiled an experimental way to search the internet that gives more conversational results, and said its artificial intelligence chatbot, Bard, is now available for much of the world to use online.

As part of a suite of AI announcements at its I/O conference earlier today (from 1am Singapore time), the company also introduced a new large language model, called PaLM 2, that developers can use to train tools like chatbots. Google said it has already woven the update into many of its marquee products, including Gmail and Bard.

“We’re at an exciting inflection point,” chief executive officer Sundar Pichai said. “We have an opportunity to make AI even more helpful for people, for businesses, for communities, for everyone.”

The efforts to reinvent major parts of its business with AI, which the company is unveiling at its conference in Mountain View, California, showcase how Alphabet Inc.’s Google is responding to competitive threats. Google is asserting its dominance in the market following moves by companies including Microsoft Corp., which is now using technology from the startup OpenAI in its own search product, Bing.

“Consumers now have more access to choice, so AI is Google’s market to lose,” said Zeno Mercer, research analyst at investment advisory firm ROBO Global.

Underpinning all the new product developments are Google’s large language models — massive AI systems that ingest enormous volumes of digital text from news articles, social media posts or other internet sources and use that material to train software that predicts and generates content on its own when given a prompt or query. Google has pioneered research into LLMs, which power the current crop of AI chatbots like OpenAI’s ChatGPT.

See also: <a class='gtm_see_also_1_inread' style='text-decoration: underline;' href='/digitaledge/news/amazon-release-q-ai-chatbot-corporate-customers?utm_source=website&utm_medium=article&utm_campaign=ONPAGESEO'>Amazon to release Q, an AI chatbot for corporate customers</a>

Any changes to Google search have the potential to upend users’ habit of typing keywords into a blank search bar, and then getting a list of relevant links to sources.

Google adds generative AI product for cloud clients

Google’s cloud service announced a variety of new ways that clients can use its generative AI tools.

See also: <a class='gtm_see_also_2_inread' style='text-decoration: underline;' href='/digitaledge/news/amazon-updates-homegrown-chips-even-it-grows-nvidia-ties?utm_source=website&utm_medium=article&utm_campaign=ONPAGESEO'>Amazon updates homegrown chips, even as it grows Nvidia ties</a>

The company unveiled a new service, Duet AI, which it called an “AI-powered collaborator.” Google said the product would help clients with a range of tasks, from providing input on users’ code to helping them create applications without writing code at all.

The company also said it would release three new AI models to its clients: Codey, for help with coding; Imagen, for image generation; and Chirp, which processes speech. In addition, Google said it would release tools that will make it easier for companies to tweak their models based on human feedback.

Other features in Duet AI let cloud customers and app developers use AI-powered prompts to write code, as well as build custom features into their Google Workspace apps, the company said. The cloud customers themselves don’t need to be able to code.

Google said that Duet AI for Google Cloud was available starting Wednesday “for limited users” through its AI trusted tester programme and that it plans to expand access soon.

Google adds way to test new generative AI features in search

Generative AI is coming to Google’s flagship product. The new search offering, which Google is calling “search generative experience,” or SGE for short, displays an AI-produced response at the top of results, which seeks to answer a user’s query while summarising key information and linking to website sources.

“Information quality has been the foundation, always, of our search experience,” Liz Reid, Google’s vice president of search, said in an interview ahead of the announcement. “And we really think, with generative AI, how do you bring both the strength of that but really live up to the promise? Users have deeply relied on us for information use cases and have deep trust in the mission of the company to do that.”

To stay ahead of the latest tech trends, <a class='gtm_see_also_3_inread' style='text-decoration: underline;' href='/section/digitaledge?utm_source=website&utm_medium=article&utm_campaign=ONPAGESEO'>click here for DigitalEdge Section</a>

If someone is using the new search features to figure out where best to go on a family vacation, for example, they could type “What’s better for a family with kids under three and a dog, Bryce Canyon or Arches?” into Google’s search bar. Google then makes a summary using AI.

In a demonstration, Google showed how its systems pulled from sources around the web to describe how both Bryce Canyon and Arches National Park are family-friendly, though Bryce Canyon has hiking trails that allow dogs.

On a right-hand carousel beside this description, Google lists the websites where its AI drew information from — though, unlike Microsoft’s Bing, it doesn’t footnote sources. The company said that one situation it thinks generative AI can handle particularly well is assisting people who want to use Google to research what to buy online.

Google also announced a new filter called Perspectives, which aims to show users long- and short-form videos and other forum posts describing the experiences of people from around the web — the company’s bid to appeal to a younger user base and counter the massive popularity of ByteDance Ltd.’s TikTok.

Reid said the Perspectives filter could pull in videos from YouTube and feature social media influencers, though not every search query will trigger Perspectives. Google is launching Perspectives first for topics like shopping, travel and how-to advice, but the filter won’t be triggered for sensitive issues, including health, finance, or politically-charged topics.

Google announced a flurry of other AI-related features in experimental search coming soon, including ads. The company also said it would soon allow users to get information about AI-generated images with labels so users can tell when photos are created by AI services like Midjourney and Shutterstock, and a new tool called “About this image” will disclose information such as when Google first registered the image and where else it has appeared. The feature will debut in the US first.

Google is opening signups to its Search Labs features for US users today, but access to the generative AI search features will roll out to those who get off the waitlist over “the coming weeks.”

Google weaves AI technology into Google Docs, Sheets and Slides

Google has woven generative AI into its popular Workspace productivity suite, letting users tap into the technology as they work on files in Google Docs, Sheets and Slides.

The company is calling the new generative AI features Duet AI for Workspace.

“Workspace was built to allow you to collaborate in real-time with other people. Now you can collaborate in real-time with AI,” said Aparna Pappu, Google’s vice president for Workspace, on stage at the event.

In Gmail and Docs, Google said Duet AI could help users draft complete responses using simple prompts. For example, users can ask the AI system to generate a thank you letter for a job interview or request a glowing review for a team member. In Google Slides, Duet AI can generate images for marketing campaigns. And in Google Sheets, the company said people can use Duet AI to ease the burden of manual data entry, for instance, by creating custom plans for tasks, or automatically assigning labels to particular columns within a worksheet.

Google lifts waitlist for AI chatbot Bard

Google’s AI chatbot Bard, which is seen as the company’s answer to ChatGPT, is opening up to all who want to use it and adding new features.

Google is aiming for parity with its biggest competitor, Microsoft’s Bing chatbot, which dispensed with its waitlist earlier this month.

Sissie Hsiao, Google’s vice president of product for Bard, said that Bard is now available in Japanese and Korean and will soon extend to more languages. In order to let users more easily transfer over answers from Bard, Google on Wednesday said it would let users export responses from the chatbot to Gmail and Google Docs.

Soon, users will also be able to prompt the chatbot with images, the company said — another feature that Microsoft recently announced for its Bing search engine.

Bard will also soon connect to more apps from Google and third-party companies, the company said. Partnerships are coming soon with companies including Adobe, Kayak and Instacart. And Google said it would add features to make it easier for developers to work with Bard to write code and complete other tasks.

Google debuts language model PaLM 2 touting language, and math skills

Earlier today, Google introduced a new large language model known as PaLM 2 that it said had already been woven into many of its marquee products.

PaLM 2 was trained on massive quantities of text from across the web. The model was also trained on coding language and mathematical data, helping it fare better with logic and math problems, Google said. As some in the field of AI have noted how OpenAI’s ChatGPT struggles with word problems, the new model from Google is expected to be evaluated closely for its performance against OpenAI’s technology.

Google is releasing the model in various sizes, and the smallest can be used on mobile devices, the company said.

Google shows off latest AI tools: Updates from I/O event

Related News

Highlights

MSCI boss warns of climate risks

Related News

Why are we all freaking out about AI

Dell inks US$150 mil hardware deal with AI startup Imbue

How AWS empowers Asean organisations to innovate with generative AI