AI Roundup 117: Google Killer

May 9, 2025.

May 09, 2025

Google Killer

The DOJ's antitrust case against Google continues to generate new revelations about Big Tech. For example, Apple's Eddy Cue shared that Google searches fell in Safari for the first time ever last month: "That has never happened in 22 years."

Why it matters:

Cue connected the numbers dip to the increased use of AI, which Apple is now considering adding to Safari (for what it's worth, Google stated that overall searches are still growing, but no mention of Safari).
Many people (including me) have scoffed at the idea of ChatGPT "killing Google" - but should Google stop being the default search engine of Safari and Firefox, or be forced to divest Chrome to a rival, that calculus could quickly change.
And what began as a case about Google's search monopoly has transformed into a debate about the future of AI - the government is arguing that Google should be prevented from dominating AI in the same way as search.

Elsewhere in the FAANG free-for-all:

Apple aims to begin mass production of a specialized chip with improved power efficiency for its smart glasses by late 2026 or 2027.
Microsoft announces support for Google's Agent2Agent specification in Azure AI Foundry and will join the A2A working group.
Google is beta launching AI Max, a suite of AI-powered enhancements for Search ad campaigns.
Apple is partnering with Anthropic to develop an AI-powered version of Xcode using Claude Sonnet for code development.
And Google plans to make Gemini available to children under 13 with Family Link parental controls, promising not to use their data for AI training.

Evolving structure

OpenAI has abandoned its plan to convert to a for-profit structure. Instead, it will keep its nonprofit division in control while transitioning its business wing to a public benefit corporation (PBC).

Between the lines:

This reversal comes after significant pushback from multiple stakeholders, including legal challenges from Elon Musk (who's still unhappy) and concerns raised by prominent third parties.
The plan still needs buy-in from Microsoft (apparently) and the California and Delaware attorneys general, who will have to sign off on the conversion. Previous reports suggested some funding could be in jeopardy if the conversion doesn’t happen this year.
Despite the structural decision, CEO Sam Altman maintains the company needs "trillions of dollars" to achieve its mission, highlighting the ongoing tension between OpenAI's charitable mission and its capital needs.

Elsewhere in OpenAI:

Instacart CEO Fidji Simo will join OpenAI as CEO of Applications to lead business and operations teams and oversee sales, marketing, and finance.
OpenAI has been meeting with the FDA to discuss cderGPT, a potential AI project focused on drug evaluation.
The company launches OpenAI for Countries to help nations develop AI infrastructure and localize its products.
And OpenAI reaches a deal to acquire AI coding tool Windsurf for approximately $3 billion.

Elsewhere in AI politicking:

Microsoft and other AI companies are urging US senators to streamline federal permits for AI energy use and increase access to government datasets.
The US Commerce Department plans to replace Biden's AI chip export restrictions with a simpler rule.
Nvidia CEO Jensen Huang warns that missing out on China's projected $50B AI market would be a "tremendous loss."
The US sees its position as the leading destination for AI talent weakening as immigration numbers reach equilibrium.
And Nvidia is developing modified AI chips for Chinese tech giants that comply with US export regulations.

The kids are not alright

For two years, educators have been concerned about AI's effect on students and homework. Now, rampant AI cheating has triggered an existential crisis for educators amid questions about the future value of college education.

The big picture:

Just two months after ChatGPT's launch, one survey found 90% of college students have used ChatGPT for homework. Today, students of all types use AI for everything from essay writing to coding assignments - and many don't see it as a problem if AI does some (or most) of their homework.
Professors and institutions are largely powerless to stop it - AI detectors are unreliable, academic policies are inconsistent, and students continue to evolve methods to avoid detection.
Some argue that coursework and pedagogy need to change, and that AI isn't creating a new problem - merely exposing deeper issues with the transactional nature of higher education and loopholes created by remote learning.
But the implications go beyond academic integrity - experts worry about diminished critical thinking skills, and whether we're creating a generation of students who are "illiterate" in both writing and problem-solving abilities.

Elsewhere in AI anxiety:

Reddit CEO announces plans to work with third-party services to verify user humanity after unauthorized AI persuasion experiments.
An investigation reveals how an "AI slop farmer" exploits social media platforms by targeting older women with AI-generated copies of real content.
The growing use of AI in job applications has created an arms race between employers and job seekers, further dehumanizing the hiring process.
Duolingo's AI-first strategy exemplifies a broader crisis in creative fields, marked by job cuts and declining freelance incomes.
A judge was emotionally moved by an AI-generated avatar of a man killed in a road rage incident.
And mental health experts warn that ChatGPT interactions may be exacerbating religious delusions and psychological issues in vulnerable individuals.