Google’s Gemini assistant is a fantastic and frustrating glimpse of the AI future

I don’t know how to say this, but sometimes the emotional labor of opening another app on my phone and typing in some text is just too much.

I need to gather details about an Airbnb reservation from two different confirmation emails and send them to my friends. Or I want to figure out when to leave this coffee shop to get home by a certain time via bus. These aren’t hard things to do, but they require enough tapping around different apps or tabbing between screens that I start to think, you know what? I don’t really need to send that email yet. I’ll just wing it and hope for the best with the bus schedule.

These are the jobs I would like AI to take from me. AI, including Google’s new Gemini assistant, isn’t quite up to it yet. But Gemini feels like a preview of what that AI future could look like — provided you’re well entrenched in Google services.

Gemini is Google’s AI chatbot, formerly known as Bard. It’s an app you download from the Google Play Store, but it really it’s a piece of the Google app that’s probably already on your phone if it runs Android. Once it’s up and running, you can replace the standard Google Assistant with Gemini and invoke it in all the same ways you would the old Assistant. But instead of just setting timers and telling you the weather, it can do all the stuff Bard did — answer complex questions, make suggestions, and read your email, if you let it.

That last part is important. Gemini isn’t nearly as good of a conversationalist as ChatGPT, but its ability to hook into Gmail, Google Maps, and Google Docs is what makes it really interesting.

I asked it to summarize the details of that Airbnb reservation, and it did — grabbing info from two different emails and putting it together in a neat little bullet point list. Then I asked it to draft an email to my friends with all of the details. Most of the time when I ask AI to write an email or text, the results are too embarrassing to actually send to anyone. To my great surprise, this one was fine.

It really doesn’t sound like much, but it’s the first time I’ve been really impressed with AI as a tool to help me get things done. Maybe I lack imagination, but I get bored with ChatGPT pretty quickly; there’s only so many times I can brainstorm business plans for a retro arcade or ideas for vacations. What I actually want is help with the pile of digital crap I’m constantly wading through just to live my life.

Gemini isn’t that thing yet, unfortunately. I think it’s more useful than the regular Assistant by a mile, and there’s a lot I like about it now. There’s the little G icon at the bottom of every answer it gives you, which allows you to search Google and fact-check Gemini’s work. Since AI has a tendency to make stuff up occasionally, that’s kind of important.

I also appreciate that when Gemini comes up with something for me, like a recipe or a packing list, I have somewhere to put it. Gemini can export answers directly to Google Docs or Gmail. When I get the same kinds of things from ChatGPT, they just feel like they’re floating around in space until I copy and paste them somewhere. They’re saved to your history, but you know what I mean.

Gemini isn’t great at picking up on context, though. I told it I was planning to ride my bike to the next neighborhood over and asked it to suggest some things to do once I got there. It spit out a whole wall of text with suggestions that included, I kid you not, scuba diving, seeing live theater, and gambling at the casino. Technically, you can do all of those things in Burien, Washington, but not on a whim in the middle of the day. They’re also not places I would just casually roll up to on my bike.

In situations like this, Gemini feels like a little “let me Google Maps that for you” machine. ChatGPT’s voice chat, on the other hand, suggested I visit a couple of parks or a coffee shop. It also asked me if I had anything else in mind for my visit, and when I said I might want to buy a book, it suggested a specific bookstore. So thoughtful! However, later on, Gemini’s Google Maps integration proved to be more useful: I asked for a bus route home, and Gemini gave me the right route, while ChatGPT would have had me waiting for a bus that only arrives every 30 minutes.

Talking to Gemini feels like talking to a page of Google Search results. If you say, “Hey Google” and ask it a question, it’ll speak the response to you. Otherwise, you’re just reading text. And it’s often a lot of text; this assistant could use an editor. I’m also surprised that Gemini can’t access my calendar, but there’s currently no extension for it like there is for Gmail and Docs. If I want to add something to my calendar, I have to switch back to the regular Assistant. At that point, I’ll just make a damn calendar event myself.

A decade ago, smart assistants like Siri and Alexa were pitched as the next big way of interacting with our devices, but across the industry we’ve seen their progress stall out. Right now, Gemini is an entirely optional assistant. But it’s also easy to see how it might eventually replace Google Assistant as the default, especially since Google has been scaling back Assistant features in recent history. Maybe we just reached the limits of what non-AI voice assistants can do reliably.

For me, swapping Gemini for the usual Google Assistant feels like a low-stakes bet. I trained myself long ago not to use voice assistants because they never seem to be able to do the thing I want them to. Gemini can still set my timers and tell me if it’s going to rain, so why not bet on something a little smarter to help me with my day to day? Hopefully it learns a few lessons from ChatGPT along the way. In the meantime, I have to get to my scuba diving lesson.