Just as Apple is dealing with a self-made AI nightmare after confirming its smarter AI-powered Siri is nothing but vaporware for the time being, Google is out with new Gemini upgrades that make its chatbot even more useful. The enhancements include personalization tricks that we actually expected Apple Intelligence and Siri to deliver before other companies got there.
Google announced the new features on Thursday via a pair of blog posts that highlight some of the Gemini apps that will be available. The updates will be free to some users and to everyone with a Gemini Advanced premium subscription.
Better Gemini Deep Research
After announcing a Deep Research feature for Gemini back in December and a 2.0 Flash Thinking Experimental AI reasoning model to compete against ChatGPT o1, Google is now ready for upgrades.
Sadly, 2.0 Flash Thinking Experimental did not get a better name, but maybe that’s in the cards for a future upgrade. Names aside, 2.0 Flash Thinking Experimental now supports file uploads. It’s also more efficient and faster than the previous version.
The 2.0 Flash Thinking Experimental reasoning AI also gets a larger token context window, assuming you’re on the Gemini Advanced plan. That’s a 1 million token context window that lets you analyze even larger amounts of data.
Google is also adding 2.0 Flash Thinking Experimental to the Deep Research feature, which should improve the quality of reports compared to the version launched in December.
Deep Research is now available to Gemini users for free, while Gemini Advanced users get expanded access. The feature is rolling out worldwide with support for over 45 languages.
Google Search Personalization
The Gemini app is getting another update some users will appreciate. It’s called Personalization, and it’s powered by the same 2.0 Flash Thinking Experimental model that badly needs another marketing name.
Personalization allows Gemini to connect to your Google apps and services to extract information and provide better assistance. It all starts with Google Search history access, which sounds a little troubling:
For example, you can ask Gemini for restaurant recommendations, and it will reference your recent food-related searches, or ask for travel advice, and Gemini will respond based on destinations you’ve previously searched. To enable this, select Personalization (experimental) from the model drop-down. You remain in control of your Search history and can always disconnect it from Gemini after enabling.
Google explained in a different post how Personalization will work, stressing that the user is in control of connecting their Gemini prompt to their Google Search history.
To take advantage of Personalization, you’ll have to enable Web & App Activity in your Google account. That means you must save your Google Search history for the AI to access. If your activity is off, you won’t be able to use it.
Gemini will only access your Search history when you select the Personalization model. Google says you can always disconnect Gemini from the Search history.
As someone with the Google Search history turned off, but also someone who doesn’t rely on Google Search for internet searches, the Personalization feature would be useless.
I’d also be interested in whether the Personalized Gemini chats are used in any way for advertising purposes. It’s easy to see how Google could gather more personal data from such a feature to serve better ads online. Google doesn’t specifically mention advertising in the blog post.
Also, as a company found to operate an online search monopoly, it probably should do it considering the obvious implication here: You have to get on Google Search and enable history to use Gemini Personalization.
Unsurprisingly, the Gemini Personalization feature won’t be available in the EU, where Google faces tighter regulation. But it’s rolling out to Gemini and Gemini Advanced users almost everywhere else, starting with the web version of Gemini.
More Gemini personalization
Google has also added its thinking model to more apps, including Calendar, Notes, Tasks, and Photos. Gemini will be able to execute complex tasks like this prompt that Google offers as an example:
Look up an easy cookie recipe on YouTube, add the ingredients to my shopping list and find me grocery stores that are still open nearby.
This is the kind of functionality I’d have expected from Siri in Apple Intelligence, but it won’t be ready for another year. I’ll also say this type of Gemini Personalization is much better than Google Search inclusion, as far as I’m concerned.
In the future, Gemini will also be able to look at your photos and surface information when asked about it. Gemini will be able to recreate itineraries from past trips or tell you when an ID is about to expire.
Finally, Google announced that Gems, or custom Gemini models you can create for specific tasks, are rolling out to everybody in the Gemini app for free.
