💰 AI Billionaires Are Promising They'll Make It Rain Money

Happy Tuesday! It’s another big money day in AI land. Over the past two days, both Elon Musk and Anthropic have announced plans to give away a portion of their wealth to the public. While good deeds from the wealthy are welcome, you can’t help but wonder if these big announcements are part of a strategy to improve AI’s image with the public, which is in the gutter.

Even if these philanthropic plans have nothing to do with public relations, they still don’t quite land given the questions over whether they’ll actually materialize. After all, Musk doesn’t have the best track record of doing what he says he’s going to do, and Anthropic still needs shareholder approval before it gives away any money it makes from a potential IPO.

→ Elon Musk says he’s giving the 72 million kids in the U.S. his SpaceX shares.
→ Jensen Huang is having a moment
→ AI agents beat Minecraft for the first time. It only cost them $1,600.

The news: If you looked online, AI companies have been making it rain money in America these past two days—at least in theory. On Monday, Elon Musk made waves when he said he would “give one of my SpaceX shares to every kid in America” in 2027. With 72 million minors, legally defined as those under the age of 18, in the US and SpaceX shares at a price of $146.03 each as of last Friday, that would bring Musk’s total giveaway to roughly $10.5 billion.

Musk’s announcement came one day after Anthropic said it was launching a public benefit fund, structured as an LLC, “dedicated expressly to the public good.” The LLC would initially hold 5% of Anthropic’s shares—a proposal that would have to be approved by shareholders—and would receive an additional 2% every time the company’s valuation increased by more than 50% over an established baseline. Anthropic is planning an IPO next month after the midterms that could value it at around $1.5 trillion.

The takeaway: Musk is not exactly known for keeping his promises or adhering to deadlines. Furthermore, he announced this token of generosity in a social media post at 4 a.m., which is peak mindless-scroll-while-on-the-toilet territory. While Anthropic has more credibility, the pledge would have been a lot more believable if it had been announced after everything was signed and sealed.

All this is to say: Take these two announcements with a grain of salt, and don’t forget the promised funds are but a fraction of the wealth in question.

The news: Nvidia CEO Jensen Huang delivered a head-scratching response when speaking to German Chancellor Friedrich Merz at the Invest in Germany summit on Monday. Here’s a snippet of what he said:

“And so, in the future, I believe that most people here in Germany will mostly work inside a digital twin of their company. They will build digital cars. Those cars will have digital brains. The manufacturing—the factories—will have digital, of course artificial intelligence, robots, building inside a robotic factory, building cars that will be robotic. And everything will be completely software-driven. Everything will be completely AI-driven.”

Gizmodo’s Mike Pearl makes a valiant attempt to piece together what exactly Huang meant with that gibberish, hypothesizing that Huang was possibly referring to the Omniverse, a feature in the Nvidia Agent Toolkit. According to Nvidia, the Omniverse simulates robot workplaces for testing purposes, allowing what happens in a digital twin of a factory to be deployed in a real factory.

The takeaway: Huang isn’t known for delivering these kinds of rambling, nonsensical responses. Instead, he’s recognized for being a stable and logical voice in Silicon Valley. As such, this was very weird. If it’s a one-off, it likely won’t affect Nvidia in the long-run, but the AI world is likely paying attention.

The news: Google DeepMind researchers have revealed yet another AI agent breakthrough: The agents managed to beat Minecraft and slay the game’s Ender Dragon all on their own. In a new paper, researchers describe how four Gemini Argon agents worked together over 27 hours and 14 minutes to beat the game, dying 605 times. The task required them to process billions of tokens and write 2,783 programs, costing a total of $1,040 in tokens at the standard price.

While getting AI agents to beat Minecraft may not sound like a big deal, the researchers say the assignment is similar to complex, real-world engineering tasks and enterprise work, which require “long-horizon coding and multi-agent coordination.” Interestingly, the instructions given to the agents did not mention forming teams, maintaining tools over time, or looking out for each other.

The takeaway: The fun experiment demonstrated that agents can work together like human teams. Like the recent study involving agents left stranded on an island, the Minecraft experiment showcases new ways to teach agents rather than just feeding them data.

Turkle is out with a new book, Artificial Intimacy: Who We Become When We Talk to Machines, which warns that AI chatbots are coming for our capacity for attachment.
Read more →

OpenAI is racing to raise money before Anthropic's IPO, seeking to attract investors who otherwise might commit to Anthropic and stave off comparisons between the two.

The US military is developing AI agents to carry out offensive missions and then self-destruct when their tasks are completed.

Former Google CEO Eric Schmidt was an early investor in Anthropic. He stands to make more than $10 billion after the company's IPO.

A US federal judge denies a request for bail made by the Saudi national accused of planting a recording device at OpenAI's offices.

People are vibe coding tools and asking AI agents to remove the ads, viewing trackers, and other bloatware from their smart TVs.

An MIT social scientist believes we should ban chatbots from using the first person. She also hates the term "human in the loop."

AI is making indie game developers fearful of sharing their early work, as the technology enables others to copy their games effortlessly.

Companies like Amazon, Yelp, and Delta Airlines are banning AI agents from their websites, calling into question the overall utility of agents in the process.

Elon Musk fans are creating AI videos showing possible uses for Tesla's flying car, such as saving people from random bridge collapses.

AI companies likely have more debt than the public is seeing, S&P Global estimates.