Google is expected to unveil a suite of new AI products today that it hopes will supercharge its already-dominant search engine, aiming to beat the likes of OpenAI, Microsoft and Meta in a fierce competition for global AI dominance.

At its annual Google I/O conference, which typically sets the tone for Google’s strategic direction in the year ahead, it will provide updates to the Gemini AI model that could make its suite of consumer-facing products even smarter as it also builds out more advanced features to sell to business and creators for a subscription.

Companies like OpenAI, Anthropic, Meta and Google have continued to refine their large language models in recent years, rolling out rapid advancements in hope of keeping pace with one another.

Lately, the industry has also been racing to develop AI agents, or tools that automatically come up with plans and execute them upon users’ requests. The technology has become so accessible that some platforms now are even using AI agents to help everyday people create their own AI-powered tools.

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947158)

it's that simple

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947182)

i mean this is all talk, though. anyone can say that they can do this. the technology is not there yet. the tech for AI gfs is definitely already here though

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947519)

yeah ok openAI is winning this game

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947186)

“I like it with lots of oil.”

*stares directly into camera, does creepy gay jew smile*

“I get fucked by men. I’m Sam Altman.”

bright future ahead! really serious and important stuff

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947198)

By: Evan39 (Night Manager, Safeway Aurora Station | Equity Partner, Perkins Coie: UX Death Panel Subdivision)

May 20, 2025

I tuned into the Google I/O stream from my break room, expecting to be briefed on AI-enhanced tax optimization or perhaps a breakthrough in autonomous grocery shrink control. Instead, a man in a bucket hat was beatboxing into a spatial microphone while telling me he planned to "jam with the computer."

An adjacent Black woman announced the arrival of “Project Astra,” which I now understand is Google’s term for an AI girlfriend who listens without judgment, simulates boundary-respecting intimacy, and can help you reset your Google password.

I watched, mouth slightly agape, as executives promised a future where your fridge knows your insecurities and your email drafts are rewritten by something more self-actualized than you’ll ever be.

I am once again reminded that in the arms race for synthetic sentience, it is not OpenAI or Meta who lead—but the Mahchine™ itself.

And it already knows I microwave egg salad.

http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947158

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947256)

i really cannot imagine that this kind of presentation appeals to any actual individual human beings

this is, paradoxically(?), the kind of overall aesthetic and presentation that you get when you try to take an overall average of all human preferences - like an LLM does - rather than creating a presentation and aesthetic from scratch, with a particular inspiration, like a human does

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947557)

when Thielbucks

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947572)

these people suck man

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48947842)

The Deep Think model gets 50% correct on US math olympiad. It’s incredible how quickly these models are improving.

(http://www.autoadmit.com/thread.php?thread_id=5727746&forum_id=2#48948272)