Feed: Engadget is a web magazine with obsessive daily coverage of everything new in gadgets and consumer electronics
Feed Link: https://www.engadget.com/
---

Title: Google's most powerful AI tools aren't for us

Date: Thu, 22 May 2025 13:46:57 +0000
Link: https://www.engadget.com/ai/googles-most-powe...

At I/O 2025, nothing Google showed off felt new. Instead, we got a retread of
the company's familiar obsession with its own AI prowess. For the better part
of two hours, Google spent playing up products like AI Mode, generative AI
apps like Jules and Flow, and a bewildering new $250 per month AI Ultra plan.

During Tuesday's keynote, I thought a lot about my first visit to Mountain
View in 2018. I/O 2018 was different. Between Digital Wellbeing for Android,
an entirely redesigned Maps app and even Duplex, Google felt like a company
that had its pulse on what people wanted from technology. In fact, later that
same year, my co-worker Cherlynn Low penned a story titled How Google won
software in 2018. "Companies don't often make features that are truly
helpful, but in 2018, Google proved its software can change your life," she
wrote at the time, referencing the Pixel 3's Call Screening and "magical"
Night Sight features.

What announcement from Google I/O 2025 comes even close to Night Sight,
Google Photos, or, if you're being more generous to the company, Call
Screening or Duplex? The only one that comes to my mind is the fact that
Google is bringing live language translation to Google Meet. That's a feature
that many will find useful, and Google spent all of approximately a minute
talking about it.

I'm sure there are people who are excited to use Jules to vibe code or Veo 3
to generate video clips, but are either of those products truly
transformational? Some "AI filmmakers" may argue otherwise, but when's the
last time you thought your life would be dramatically better if you could
only get a computer to make you a silly, 30-second clip.

By contrast, consider the impact Night Sight has had. With one feature,
Google revolutionized phones by showing that software, with the help of AI,
could overcome the physical limits of minuscule camera hardware. More
importantly, Night Sight was a response to a real problem people had in the
real world. It spurred companies like Samsung and Apple to catch up, and now
any smartphone worth buying has serious low light capabilities. Night Sight
changed the industry, for the better.

The fact you have to pay $250 per month to use Veo 3 and Google's other
frontier models as much as you want should tell everything you need to know
about who the company thinks these tools are for: they're not for you and I.
I/O is primarily an event for developers, but the past several I/O
conferences have felt like Google flexing its AI muscles rather than using
those muscles to do something useful. In the past, the company had a knack
for contextualizing what it was showing off in a way that would resonate with
the broader public.

By 2018, machine learning was already at the forefront of nearly everything
Google was doing, and, more so than any other big tech company at the time,
Google was on the bleeding edge of that revolution. And yet the difference
between now and then was that in 2018 it felt like much of Google's AI might
was directed in the service of tools and features that would actually be
useful to people. Since then, for Google, AI has gone from a means to an end
to an end in and of itself, and we're all the worse for it.

Even less dubious features like AI Mode offer questionable usefulness. Google
debuted the chatbot earlier this year, and has since then has been making it
available to more and more people. The problem with AI Mode is that it's
designed to solve a problem of the company's own making. We all know the
quality of Google Search results has declined dramatically over the last few
years. Rather than fixing what's broken and making its system harder to game
by SEO farms, Google tells us AI Mode represents the future of its search
engine.

The thing is, a chat bot is not a replacement for a proper search engine. I
frequently use ChatGPT Search to research things I'm interested in. However,
as great as it is to get a detailed and articulate response to a question,
ChatGPT can and will often get things wrong. We're all familiar with the
errors AI Overviews produced when Google first started rolling out the
feature. AI Overviews might not be in the news anymore, but they're still
prone to producing embarrassing mistakes. Just take a look at the screenshot
my co-worker Kris Holt sent to me recently.

Kris Holt for Engadget

I don't think it's an accident I/O 2025 ended with a showcase of Android XR,
a platform that sees the company revisiting a failed concept. Let's also not
forget that Android, an operating system billions of people interact with
every day, was relegated to a pre-taped livestream the week before. Right
now, Google feels like it's a company eager to repeat the mistakes of Google
Glass. Rather than trying to meet people where they need it, Google is
creating products few are actually asking for. I don't know about you, but
that doesn't make me excited for the company's future.

This article originally appeared on Engadget at
https://www.engadget.com/ai/googles-most-powe...
134657007.html?src=rss

---
VRSS v2.1.180528