The Google I/O 2025 keynote made me so depressed about our AI future

As I sat down to watch the opening keynote for Google I/O 2025 yesterday, I glanced out the window behind my computer. It was a cold, rainy, and overall gloomy day. A gray sky hid any trace of the sun. A frigid breeze and a constant drizzle squashed my hopes of an afternoon run or a long walk with my dog.

It was a perfect illustration of how Google’s two-hour keynote would make me feel throughout its entire runtime.

As expected, artificial intelligence was a big focus for Google. Between AI in search, content creation, XR, and more, we knew AI would be everywhere we looked. But having now seen the specifics of what Google’s been working on — and what its vision of an AI future looks like — I couldn’t be more disinterested about where the company is going and where it wants to take us.

Since Google has been on this AI journey, it’s been clear that it intends to create AI tools that make everyday and complex tasks easier to accomplish. That’s ultimately the goal of any AI company, and there’s nothing inherently wrong with that. But it’s the way Google is approaching this goal and the features it’s making to achieve it that have me feeling so bad about it.

One of the first examples we saw was an update on Gmail’s Smart Replies. AI can already generate email replies for you, but with the new Personalized Smart Replies, those AI-generated emails will contain more relevant information while mimicking your personal writing tone.

Instead of you having to perform the dreaded task of personally writing an email to a friend, family member, or work colleague, you can have Gmail’s AI tool do it for you, all while writing with your own style and mannerisms to trick the person on the other end into thinking it’s really you. Or, as Google CEO Sundar Pichai put it, you can use Personalized Smart Replies to “be a better friend.”

The human on the other end of that message may not like to know you couldn’t be bothered to email them yourself. But you know who doesn’t care? AI agents, baby. Not long after the Gmail demo, Pichai showed an example of using “Agent Mode” in Gemini to help you find an apartment. On paper, it sounds fairly interesting. Agent Mode can research apartments for you, create a spreadsheet of your findings, and even schedule tours for places you like.

But when you step back and really look at what’s going on, I don’t think this is actually solving any problems.

Gemini doesn’t do all of this for you magically. You have to type long, detailed paragraphs about what you want in an apartment. How many bedrooms and bathrooms you want. What location you want to live in. What your price range is. All of this needs to be written in a way Gemini can understand. Tell me how that’s any easier than clicking a few filters on Zillow or Apartments.com? And when you do finally get all of your apartment results from Gemini, are you going to trust that what it’s given you is 100% accurate?

Google frames the apartment scenario as this time-consuming, awful, horrible process. Why should you have to do that when the AI can? Because I trust myself. Because scrolling through Zillow listings is objectively easier than writing 500 words to Gemini to make sure it knows exactly what I’m looking for. Because I have fond memories of sitting up late at night with my partner, having a couple of beers together, and listening to music while making a spreadsheet of our apartment must-haves. But thank god I could take all that away and have Gemini’s Agent Mode handle everything for me.

Google’s I/O keynotes have never been perfect. Between wonky tech demos and features that take years to materialize, there have always been valid criticisms to raise. But in years past, those keynotes also gave us legitimately cool software and hardware products to look forward to.

Google I/O is where Google introduced Wear OS to the world. It’s where the company launched the first Pixel A-Series phone. Google I/O is where we met Google Assistant, the first Google Home speaker, and Google’s first big push into VR with Daydream. This used to be a show where Google regularly impressed us by showing the best and coolest things it could make. But now, it feels like the company has lost all of that.

Google is still obviously making things, but almost everything it’s putting out into the world feels bad. Right now, the most interesting things to Google are finding ways to let us communicate with loved ones less, have AI be in charge of everything from dress shopping to apartment hunting, and creating more powerful tools for computers and not humans to be artists.

What pushed me over the edge more than anything was the end of the keynote, where Sundar said how one day AI will be used to “find treatments for the world’s deadliest diseases.” How AI will build futuristic robots and create fully autonomous driving. It creates an illusion that dumbing us down to the point where we can’t write an email or look for a new place to live is a critical step toward saving lives. If you ask me, that verges on being outright evil.

I want to be excited about the next big thing in tech. I want to be dazzled by whatever crazy new invention Google has whipped up. But having AI impersonate me in an email isn’t going to do it. Removing human experiences certainly won’t do it. But I don’t think Google understands that, and pending a drastic leadership and focus shift, I’m not sure it ever will.