Unlock Editor's Digest free of charge
Roula Khalaf, editor of the FT, picks her favorite stories on this weekly newsletter.
Google has launched a sophisticated version of its Gemini artificial intelligence model, allowing it to take motion on behalf of users because the US tech giant struggles to bring AI-powered assistants to consumers.
The Silicon Valley giant on Wednesday also unveiled its vision of two “AI agents” based on the brand new model that may reply to real-time queries via text, video and audio. These have been tested by a small group of users within the US and UK over the past few months.
It comes at a time when tech giants like OpenAI, Meta and Apple want to launch AI-powered personal assistants that may solve and complete complex tasks for humans, as they appear to generate revenue from their powerful but costly models.
On Wednesday, Apple also released an update to its operating systems that marked its first major foray into generative AI. This includes giving iPhone users free access to OpenAI's ChatGPT and its most advanced models via Siri, Camera, and writing tools. The timing for this move was the launch of Apple Intelligence for the primary time in markets outside the US, including the UK.
Google wouldn't confirm when it will release the prototypes – generally known as Project Astra and Project Mariner – more broadly to consumers, but said it was a crucial step closer with these working versions.
“These are people in the actual world, in a controlled environment. . . We want to start out getting real-world feedback as early as possible,” said Praveen Srinivasan, technical director at DeepMind, who worked on the Astra project.
Astra could be accessed via either a phone or smart glasses, while Mariner can perform tasks in a user's Chrome browser, including adding groceries from a recipe to a web based shopping cart, filling out forms, or planning travel routes .
Google's presentation of its improvements shows how AI agents have change into the latest front within the battle between tech corporations.
In October, AI startup Anthropic introduced a tool that may take actions on behalf of users aimed toward the developer market. Now Google, Amazon, Meta and OpenAI are amongst those developing general-purpose agents that could be utilized by anyone. OpenAI recently said it expects AI agents to achieve the mainstream in 2025.
“Over the last 12 months, we've invested in developing more agent models, meaning they will understand more concerning the world around you, think several steps ahead, and take actions in your behalf under your supervision,” said Alphabet CEO Sundar Pichai.
In a live demo of Astra Assistant on a smartphone, the AI software answered a series of questions from the Financial Times about paintings the phone camera was pointed at, including accessing memories of works it had recently seen. Because of his ten-minute photographic memory, Astra was also capable of memorize pages in a recipe book after which reply to questions on ingredients and wine pairings.
A video demonstration of Astra showed a user wearing glasses that acted as a camera that they might activate by tapping the page. The product was paying homage to Google Glass, an ambitious but failed attempt at wearable technology that the tech giant announced in 2012 and shelved three years later.
Mariner is a Chrome-based browser add-on that may read web pages in addition to tap, click, and scroll in your behalf. Megha Goel, a product manager at Google, said that to maintain users protected, the corporate has currently blocked certain actions, corresponding to purchasing items online or accepting cookies on their behalf.
