Google's latest AI features initially will be rolled out to Gemini Pro and Ultra subscribers in the United States before coming to a wider, global audience. — AP
Alphabet Inc’s Google said its Gemini artificial intelligence assistant can now proactively tap into users’ data across Gmail, Search, Photos and YouTube, an attempt to make its consumer-facing AI product more personalised.
Using this extensive digital paper trail, the new feature – called Personal Intelligence – is designed to "make Gemini uniquely helpful,” Josh Woodward, a vice president overseeing the Gemini app, Google Labs and AI Studio, wrote in a blog post published Wednesday.
"This marks our next step toward making Gemini more personal, proactive and powerful,” he added. Personal Intelligence is initially launching in the US in beta.
To preempt any privacy concerns, Google has made Personal Intelligence an opt-in experience, meaning users must choose to turn it on – and they can choose which apps are connected to Gemini. For example, you could set it up to ignore your search history but factor in your email, photo library and YouTube viewing history.
This sheer volume of personal data lends Google’s chatbot a distinct advantage over other AI companies, few of which have as much information about their users. OpenAI has attempted to personalize ChatGPT – for instance, by adding the ability for it to retain specific information from chats.
Gemini could already retrieve information from various Google services when asked to do so, but the difference with Personal Intelligence is that the assistant can automatically tap into your chosen apps whenever it might lead to a better response.
In one example provided by the company, a user tells Gemini he needs to replace his car tires and asks for the best options – without specifying anything about his vehicle in the prompt. But based on the wealth of data it’s been granted through Personal Intelligence, Gemini already understands what he needs.
Personalization won’t be used for all Gemini responses, and users have the ability to regenerate an answer without it if they prefer, Google said. Guardrails are in place for sensitive topics, it added. But Woodward notes that errors and misunderstandings are inevitable."We’ve tested this beta version of Personal Intelligence extensively to minimize mistakes, but we haven't eliminated them,” he wrote, citing examples like over-personalization, where the model makes connections between unrelated topics. "Gemini may also struggle with timing or nuance, particularly regarding relationship changes, like divorces, or your various interests,” he added.
Google is encouraging users to provide feedback using a thumbs down button, and they can also correct Gemini directly in the conversation.
The beta rollout of Personal Intelligence will begin Jan. 14 for Google AI Pro and AI Ultra subscribers in the US and should reach all of those users over the next week. The company is working to expand the feature to other countries and Gemini’s free tier, it said.
Earlier this week, Google and Apple Inc. jointly confirmed that Gemini will power the iPhone maker’s upcoming artificial intelligence features, including a revamped Siri assistant. – Bloomberg
