New AI announcements from Google and Baidu over the last few days could be sending a new signal about where the AI race is going.

Our Deirdre Bosa has more in today's tech check. Morning, Deirdre.

Hey, Sarah. So AI, it is getting cheaper, more accessible. Over the weekend, as you said, Google made a bunch of new Gemini features available to a wider group of users at no cost.

And in China, Baidu released two new versions of its Ernie model at a fraction of the cost

of competitor models from Open AI and Deepseek.

Now for Google in particular, helped to drive buzz and adoption to elements that it has been

chasing since this race kicked off in full, that Gemini 2.0 flash thinking experimental model,

it includes a major image generation upgrade, which led to some of these viral examples over the weekend.

One user added a very realistic looking girlfriend to his dinner where he was actually alone.

Another who said she was running late to work, asked Gemini to turn her selfie at home into a selfie on the subway with a thumbs up on her route to work.

Tried it out myself. Here's a photo of me holding a puppy at our one market set.

Sadly, this is not an actual tech check puppy you're seeing.

The downside of all of this though is that these features, they're still really hard to find for the average person,

hitting on a major complaint that has dogged Google for the last few years.

For all of those breakthroughs, all of its breakthroughs, its AI is often feels buried under layers of menus or experimental labs or separate apps,

rather than just seamlessly integrated into products that people are

already using. Now, even as it's vertical integration, that's chips, cloud business models,

app ecosystem, even as that gives it a cost advantage that's not available to competitors

like Open AI and Anthropic. On that front, some developments here too. A report from the information

says that Google is tapping

media tech for AI chips.

Or in other words, looking for more ways to scale its AI,

...