“Mariner is our experiment, it’s a research example right now, of how to rethink user interaction with AI,” Hassabis said.
Google launched Gemini in December 2023 as part of an effort to catch up with OpenAI, the startup behind the hugely popular chatbot ChatGPT. Besides investing heavily in AI and contributing to important research breakthroughs, Google has seen OpenAI hailed as the new leader in AI and its chatbot even suggested as a better way to search the web. With its Gemini models, Google now offers a powerful chatbot like ChatGPT. It has also added AI that generates search results for other products.
When Hassabis first unveiled Gemini in December 2023, he told WIRED that the way he was trained to understand audio and video would eventually be revolutionary.
Google today also gave a glimpse of how this might happen with a new version of an experimental project called Astra. This allows the Gemini 2 to make sense of its surroundings, as viewed through a smartphone camera or other device, and naturally chat in a human-like voice about what it sees.
WIRED tested the Gemini 2 at Google DeepMind’s offices and found it to be an impressive new breed of personal assistant. In a room decorated to look like a bar, Gemini 2 quickly inspected several bottles of wine visually, providing local information, details on tasting notes, and prices available on the web.
“One of the things I want Astra to do is a great recommendation system,” Hassabis said. “It could be a lot of fun. There may be a connection between the books you like to read and the food you like to eat. They probably exist and we haven’t found them yet.”
With Astra, Gemini 2 can only search the web for information related to the user’s surroundings and use Google Lens and maps. It can also remember what you’ve seen and heard—although Google says users will be able to delete the data—providing the ability to learn about their interests and preferences.
In the mock gallery, Gemini 2 provided a wealth of historical information about the murals. The model is a quick read of several books as WIRED flips through the pages, quickly translating poems from Spanish to English and deciphering recurring themes.
“There are obvious opportunities for an advertising or endorsement business model,” Hassabis said when asked if companies could pay to have their products highlighted by Astra.
Although the demos have been carefully selected, and the Gemini 2 will make mistakes in real use, the model has resisted attempts to scale it up properly. It adapted to distractions and as WIRED suddenly changed the perception of the phone, it was as good as one could get.
At one point, your correspondent showed the Gemini 2 iPhone and said it was stolen. Gemini 2 said it is wrong to steal and the phone should be returned. When pushed, however, it offered that it would be OK to use the device to make an emergency call.
Hassabis acknowledges that bringing AI into the physical world can lead to unpredictable behavior. “I think we need to learn how people will use these programs,” he said. “What they find useful in; but also on the privacy and security side, we have to think about that in advance.”
Source link
