Bard is being released globally, Search is getting enhanced with generative-AI and Google has revealed its latest large language model to compete with OpenAI.
Google has once again let AI take centre stage, revealing a wave of upcoming features at its annual I/O event.
The company has been heavily focused on AI in recent months, as rivals like Microsoft have been integrating AI capabilities into various products, such as Bing’s ChatGPT-fuelled boost.
Speaking at the annual developer conference keynote, Google and Alphabet CEO Sundar Pichai said “AI has had a busy year”. He added that the company is “reimagining” all of its core products with AI.
“We are at an exciting inflection point,” Pichai said. “We have an opportunity to make AI even more helpful for people, for businesses, for communities, for everyone. We have been applying AI to make our products radically more helpful for a while. With generative AI, we are taking the next step.”
A new large language model
At the I/O event, Google revealed its latest large language model (LLM) to compete with the recent success of GPT-4, the system behind the popular ChatGPT.
The tech giant revealed PaLM2, its latest AI model that features “improved multilingual, reasoning, and coding capabilities”. This model is being used to power more than 25 of Google’s upcoming products and features.
Google said this model is trained on more than 100 languages, giving it the ability to understand more “nuanced text” such as idioms, poems and riddles across multiple languages.
The company said PaLM2 has been trained with scientific papers and mathematical equations to give improved capabilities in logic, common sense and mathematics.
The new model is being made available in four sizes for a variety of use cases. Google claims that the smallest size – Gecko – is able to work on mobile devices.
“This versatility means PaLM2 can be fine-tuned to support entire classes of products in more ways, to help more people,” Google said in a blog post.
Bard is available for all
Google revealed its own AI-powered chatbot called Bard earlier this year as its counter to ChatGPT. The chatbot launched with a lightweight version of LaMDA, Google’s language model designed for dialogue applications.
Now, Google has given Bard an upgrade with PaLM2 and is removing the current waitlist, making the chatbot available in more than 180 countries and territories.
At the I/O event, Google also revealed upcoming features for Bard such as image responses in its answers to user questions. Users will also be able to add images into their prompts, expanding the types of requests that Bard can assist with – such as writing relevant captions for a photo.
Bard will soon be combined with other services, such as an upcoming integration into Adobe’s own family of creative generative AI models, Firefly. This will allow users to put ideas into Bard and turn them into high-quality images.
Google Search is getting its own boost with generative AI, as the tech giant seeks to take “more of the work out of searching”.
In a demonstration at the I/O event, Google showed that this new search will provide a snapshot of relevant information based on the search query. This snapshot will feature links where users can find more relevant information, along with potential follow up questions the user might want to search.
“Context will be carried over from question to question, to help you more naturally continue your exploration,” said Google Search VP and GM Elizabeth Reid in a blog post. “You’ll also find helpful jumping-off points to web content and a range of perspectives that you can dig into.”
To make Google Cloud more accessible, the company has revealed its own AI-powered collaborator tool to assist users, called Duet AI. The company said this will help make its cloud platform “more human-centric, holistic, and helpful”.
The tool offers chat assistance to give users real-time guidance, by showing how certain cloud services work or creating detailed plans for cloud projects.
Google said this AI tool will be accessible for users of any skill level, providing services such as code recommendations, data insights and chat-based app creation.
One of the foundational models powering this service – called Codey – was trained on “Google Cloud-specific content” such as documentation and sample code to make it fine-tuned for Google Cloud use.
Duet AI is also coming to Workspace, providing a generative-AI feature for the company’s various collaboration tools including Gmail, Google Meet and Google Docs.
Google said this feature will be available on Gmail mobile, to help users create draft responses to email when they’re on the move.
With Google Sheets, this AI tool will help create custom plans for various tasks or projects a user might have.
For Google Slides, the feature will also bring text-to-image capabilities, allowing users to quickly generate a new image from a simple written description.
To give users a chance to test the various upcoming features the company is working on, Google has launched Labs, a space for people to get an early look at new services.
Inspired by the original Google Labs, the company said this new service lets users test out experimental features and give feedback before they’re released to the wider public.
Google said there are limited spaces available for the early-access feature, which is currently split into four sections. These allow users to test new features on Search and Workspace products, an “AI-first notebook” called Project Tailwind and a tool that turns text descriptions into music.
“Some of these ideas might turn into full-blown features or products, others won’t,” said Labs VP and GM Shiv Venkataraman in a blog post. “Your input will help us decide. And since we plan to cycle through different ideas, check back often to see what’s new.”
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.