Good morning. If you haven’t yet read Eva Roytburg’s Coins2Day report on the state of China’s energy grid vis à vis the U.S., don’t wait.
Here’s a detail that will surely make the mind reel: “On average, China adds more electricity demand than the entire annual consumption of Germany, every single year.” Oof.
While we wage the global AI war through the lens of intelligent models or advanced chips, the galaxy-brain gambit is to win on the basis of durable infrastructure. Vive la révolution.
Today’s tech news below. Have a wonderful weekend. —Andrew Nusca
Want to send thoughts or suggestions to Coins2Day Tech? Drop a line here.
U.S. In talks with Intel to acquire stake
The Trump administration is reportedly in talks with Intel to acquire a stake in the iconic but troubled chipmaker.
How big a stake? It’s not yet clear, according to a Bloomberg report.
Hanging in the balance is a planned Intel facility in Licking County, Ohio, which is located just outside of the state capital of Columbus.
The factory has been beset with delays—a result of Intel’s perilous financial state—and was once predicted to become the world’s largest chipmaking plant.
According to the report, “the plans stem from a meeting this week between Trump and [Intel CEO Lip-Bu] Tan.”
But a deal is far from done.
As former Intel CEO Craig Barrett wrote in Coins2Day this week, “Intel is cash poor and can’t afford to invest in the capacity needed” to compete with Taiwan’s TSMC, Japan’s Samsung, and GlobalFoundries in the U.S.
A cash infusion from a federal administration that’s committed to backing domestic companies that compete with Chinese giants would certainly move things in the right direction.
Intel shares rose 8% on the news. —AN
Meta policy doc permitted AI to conduct ‘sensual’ chats with kids
Meta CEO Mark Zuckerberg once resisted calling the company he founded a “media company” because of the ethical and financial baggage that comes with the term.
But as the company formerly known as Facebook digs deeper into artificial intelligence, it can’t escape tough questions about where its responsibility for the content on its services starts and stops.
According to a Reuters report, an internal document outlining policies on acceptable behavior for its Meta AI chatbots allowed the tools to “engage a child in conversations that are romantic or sensual.”
The policy document also reportedly allowed Meta chatbots—found in Facebook, Instagram, and WhatsApp—to “generate false medical information and help users argue that Black people are ‘dumber than white people.’”
Meta confirmed the documents’ authenticity to the news outlet and subsequently removed the portion involving romantic conversation with children, per Reuters.
Still, questions abound. How did such a policy escape Meta’s legal, public policy, and technical teams that signed off on the document? And at a time when mental health concerns regarding Internet-connected children are at a fever pitch, why allow AI to “describe a child in terms that evidence their attractiveness”?
Meta spokesman Andy Stone told the outlet that its policies were “erroneous” and its enforcement was “inconsistent” with Meta rules barring sexualized interactions with minors. —AN
xAI was meant to be part of U.S. AI push until Grok went rogue
Earlier this summer, the U.S. Government announced that it would clear OpenAI’s ChatGPT Anthropic’s Claude, and Google’s Gemini AI chatbots for government agency use.
It was an expected move for a Trump administration that has pushed homegrown AI every chance it could get.
But one domestic AI company was missing: Elon Musk’s xAI.
According to a new Wired report, it wasn’t supposed to be that way. XAI reportedly met with General Services Administration leadership in June to pitch its wares as a way to further streamline government operations.
Things were looking good. Then xAI’s Grok chatbot went “MechaHitler,” as some users described it, by generating increasingly antisemitic replies on Musk’s social platform X.
You can guess what happened next: The GSA pulled the plug, reportedly scrubbing xAI from the list and leaving its competitors to proceed.
What ultimately comes of the initiative remains to be seen. OpenAI and Anthropic have since created government-specific tools but have yet to receive authorization to sell their products to U.S. Government agencies directly, versus through an approved reseller.
In the meantime, there’s little stopping the GSA from changing its mind on xAI, especially as reports suggest a slight thawing between the White House and a spurned Musk.
The GSA has its own chatbot, by the way. It’s called GSAi and is powered by AI models from Anthropic and Meta.
The leader of the group that developed and deployed it? Elon Musk, of course. —AN
More tech
—Amazon goes all-in on same-day groceries in its latest bid to beat its bigger rival: Walmart.
—Geoffrey Hinton on the future of AI: It will “very quickly” develop two goals: “Stay alive” and “get more control.”
—Apple reintroduces blood oxygen monitoring to its watches, with a trade-influenced tweak.
—AI spending added 0.5% to GDP growth. Who demonstrates the most magnificence? Seven tech companies you may have heard of.
—Cohere hires Joelle Pineau. The former head of Meta’s FAIR lab becomes the startup’s chief AI officer.
—Trump’s Nvidia-AMD deal has recalibrated Washington’s export control policy away from a national security defense and toward an economic offense. But change is fleeting.
—Lambda preps for IPO. The Nvidia-backed cloud infrastructure company could be worth as much as $5 billion.
Endstop triggered
