Google I/O: Google Gemini, Project Astra Etc

Alphabet’s Google opened its 2024 developer conference Google I/O with a predictably heavy focus on artificial intelligence (AI).

To this end one of the major announcements centred around its flagship Google Gemini AI model (previously known as Bard), which now has a faster Flash version to compete with the new and faster GPT iteration (GPT-4o) that OpenAI announced a day earlier.

Google also revealed that Gemini is being rolled out to more Google services, including Search and Gmail, with Google Workspace.

AI focus

Search for example has been redesigned with AI Overviews, which can summarise the web in response to complex queries.

Meanwhile a new Ask Photos assistant can dig through a user’s photo archive for the answer to a question like “what’s my license plate?” or “show the progress of my daughter’s swimming lessons.”

With Gmail, Gemini can be used to summarise emails that are part of a longer email chains.

Gemini will also give the option for a smart reply function, allowing a user to deliver more tailored replies after analysing an email conversation.

Google is also updating Gemini on Android, allowing Gemini to be set as the default assist on an Android phone.

Gemini can already summarise or answer questions about a webpage or a screenshot, but the upcoming Project Astra prototype – an AI agent – has been designed for everyday life on either a smartphone or smart glasses.

For example one demo showed someone using Project Astra to help them solve a coding problem using a smartphone camera to analyse the code on a computer screen.

Another demo was a user asking the AI assistant where their glasses were left. The AI assistant answered after analysing the video and detecting the glasses that had been left on a desk next to a red apple.

Google also confirmed that more AI functionality will be included in the upcoming Android 15, although it didn’t discuss exact features here.

Google Veo

Another new AI model is called Google Veo, which utilises generative AI to create high quality, detailed 1080p videos from either text, image or video prompts.

In February 2024 Google had unveiled a new text-to-video AI model called Sora, which could by just using text instructions, “create videos of up to 60 seconds featuring highly detailed scenes, complex camera motion, and multiple characters with vibrant emotions.”

Besides Gemini Pro, Google is now also offering a lightweight version called Gemini 1.5 Flash, that will appeal to the developer community.

On the hardware side Google unveiled a new sixth generation GPUs called Trillium, deliver 4.7x performance increase over the previous generation.

Google offered a 10 minute video recap here.