Google I/O 2023 live updates: PaLM 2 announced

Google’s online office suite, Workspace, will soon allow users to generate full documents and fill out spreadsheets with AI.

Google’s “Help Me Write” feature that can write essays and sales pitches will be included in Google Docs for all people, the company said on Wednesday. Another example the company gave was automatically writing job descriptions for online ads by prompting a “senior level job description for a textile designer.”

Google also demoed the software, based on PaLM 2, filling out an entire spreadsheet with three made-up dogs and notes for a fictional dog walking company. Google says you can take the structure and put in your own data and details.

Google Slides can generate AI images inside the app, Google said. The example on stage created images ofa slice of pizza being dipped into fondue.

The new features will be available later this year for subscribers to a service called Duet AI, the company said. — Kif Leswing

Protestors are flying a banner over the Shoreline Amphitheater reading “Protect Abortion Privacy,” according to CNBC reporters on the ground.

“Google committed to deleting location data about health care visits including abortion care, but they are STILL retaining this sensitive data,” the protestors say on their website. — Kif Leswing

Google said it has built a new general-purpose large language model called PaLM 2 that can generate a wide variety of human-like text in response to human input. CNBC reported on Monday that Google was planning to announce PaLM 2 during the I/O conference this week.

The model is skilled at math, software development, language translation reasoning and natural language generation. “It’s better than our previous state-of-the-art language models,” said Zoubin Ghahramani, vice president for Google DeepMind, told reporters during a media briefing earlier this week. Google announced the LaMDA model in 2021 and the more powerful PaLM model in 2022.

PaLM 2 uses Google’s custom artificial intelligence chips and is more efficient to operate than previous models, Ghahramani said. The new model is able to work with over 20 programming languages, including the longstanding Fortran, and over 100 spoken-word languages, Ghahramani said.

Around 20 Google products now use PaLM 2, he said, adding that a lightweight version can run on a mobile device.

Ghahramani declined to specify the exact sources it used to train PaLM or the number of parameters the model uses. OpenAI has taken a similar approach to disclosure with its GPT-4 model. —Jordan Novet

Many of Google’s top leaders are at I/O, including the company’s heads of HR and search, CNBC reporter Jennifer Elias spotted.

The new AI powered functionality will pull in air quality information, weather, and will visualize a chosen route for Google Maps users. CEO Sundar Pichai showcased a Manhattan bike route that visualized a bike ride alongside the West Side Highway, with AI-backed renderings of traffic, weather, and the route itself.

The demo suggested flyover functionality and that the feature would helpful information for travel superimposed over the user interface. Pichai said the functionality would roll out to over a dozen major cities in the coming months.

— Rohan Goswami

Google’s first big feature announcement could make it much easier to write emails — or it could fill your inbox with low-quality computer generated messages.

Google will integrate generative AI into Gmail to help expand emails quickly, called “Help Me Write.” Google’s example was filling out a letter to get a full refund from an airline. Google has settings to make the generated text longer or shorter, and it uses context from previous messages to help write the copy.

“Just type in the prompt of what you want, hit create, and a full draft appears,” Pichai said. — Kif Leswing

Google’s CEO has taken the stage and he’s signaling that they’ll be talking about artificial intelligence all day.

“As you may have heard, AI has had a very busy year,” Pichai said.

“With generative AI, we’re taking the next step,” he added.

Google is also likely to discuss updates to Android, which is the most used operating system in the world with 3 billion active devices. Android 14 is expected to launch sometime this summer and the company could spend some time talking about some of the new consumer-facing features that will be available.

Google’s also expected to unveil new hardware under its Pixel brand, like the Pixel Fold, Pixel 7a, and a Pixel tablet. Google’s Pixel phones only have about 2% of the U.S. market, according to Counterpoint Research. — Kif Leswing

Ahead of the start of Google I/O, attendees are getting a free concert from indie electronic artist Dan Deacon.

Deacon’s song, “When I’m done dying,” was made partially by Google’s generative AI tools, he said. The background was made by an image generator reacting to the song’s lyrics.

“I’m going to play some songs, and a lot of the content is going to be made using Phenaki, Bard, and MusicLM,” Deacon said. — Kif Leswing

This year, lots of Google’s announcements are expected to center around artificial intelligence. AI has become the hottest part of the technology industry, partially due to attention from OpenAI and rival Microsoft, which has integrated so-called “large language models” that can spit out fluent text and image generators that can spit out original pictures into its Bing search engine.

But Google was heavily involved in the creation of these technologies, in particular, the transformers architecture that ChatGPT and other rivals use. It has its own chatbot, Bard, which is currently limited only to a select number of beta testers. It also has one of the strongest AI departments, which was recently reorganized to become Google DeepMind. — Kif Leswing

We’re here at Google I/O 2023 seated and ready for the event to begin.

The keynote should last a couple of hours and announcements will likely run the gamut from software to hardware news. But, as we’ve already said, artificial intelligence is likely going to be the key focus of Google I/O 2023.

The company is planning to announce a number of generative AI updates, including launching a general-use large language model (LLM), CNBC reported on Monday. — Jennifer Elias