Trump plans an 'AI Force'

Trump's AI Force, a US-China AI hotline, and more.

Trump plans an 'AI Force'

Hi there, this is your daily β˜•οΈ Techpresso.


In today's Techpresso:

πŸ‡ΊπŸ‡Έ Trump plans an 'AI Force'

πŸ”΄ US proposes AI 'red phone' with China

πŸ›’ Amazon blocks Meta's Muse AI from shopping

πŸ’» Google launches Googlebook laptop

πŸ€– Humanoid robot hype outpaces actual sales

🎲 Polymarket CEO pushed staff to commit fraud

Plus: 🎁 10 other news you might like, 🧰 6 tools, and πŸ“š 5 papers.

πŸ‡ΊπŸ‡Έ Trump plans an 'AI Force' LINK

  • President Trump said Saturday he'll create an "AI Force," a new military branch modeled on the Space Force he founded during his first term, and will soon appoint an AI "czar" to run it.
  • Announced in a Truth Social post, the plan would likely centralize the military's scattered AI programs under one roof, much as Space Force pulled together satellite and missile-warning operations from the Air Force, Army, and Navy.
  • The czar job has sat empty since venture capitalist David Sacks stepped down in March, and the news follows Pentagon AI failures, including a false AI-generated report claiming a Chinese ship carried nuclear parts.
  • πŸ”΄ US proposes AI 'red phone' with China LINK

  • The US has floated an AI "red phone" with China, a direct channel where both governments would alert each other about major national security incidents tied to artificial intelligence.
  • Treasury Secretary Scott Bessent said he pitched the notification setup to Chinese Vice Premier He Lifeng in New York yesterday, calling for more transparency between the world's two top AI powers ahead of Trump's summit with Xi Jinping.
  • China did not publicly back the plan, with state agency Xinhua calling the talks "constructive" but skipping the mechanism, while chip export controls stayed off the table and a separate deal covered "non-sensitive" trade goods.
  • πŸ›’ Amazon blocks Meta's Muse AI from shopping LINK

  • Amazon has shut Meta's new Muse AI agent out of shopping on Amazon.com for customers, saying it never agreed to let the agent buy items and asked Meta to remove the site voluntarily first.
  • Amazon says Muse browses without identifying itself and appears to capture and store customer login details, which it calls privacy and security risks; Meta counters that Muse never sees passwords or payment methods, keeping them in secure storage.
  • Shoppers now hit a popup saying an unauthorized AI agent violates Amazon's Conditions of Use, a contract-based argument that survives after courts ruled in Perplexity's favor on anti-hacking claims, with Amazon declining to say if it will sue Meta.
  • πŸ’» Google launches Googlebook laptop LINK

  • Google has launched Googlebook, a laptop platform blending Android and ChromeOS, with five models from Acer, Asus, Dell, HP, and Lenovo starting at $899, available to preorder now and reaching stores on October 4.
  • The new Googlebook OS ties into Android phones so you can open phone apps and reach your documents, downloads, and photos from the desktop, and it leans heavily on Gemini AI features like Magic Pointer and voice tool Rambler.
  • Each Googlebook carries an RGB Glow Bar, at least 16GB of memory, Intel Core Ultra or Snapdragon X Elite chips, and a decade of promised updates, while Chromebooks keep their support through 2034 for now.
  • πŸ€– Humanoid robot hype outpaces actual sales LINK

  • Only about 7,000 humanoid robots were sold worldwide last year for industrial and professional work, according to the International Federation of Robotics, a far smaller figure than the boom that forecasts have promised.
  • The count is tiny next to the roughly 542,000 regular industrial robots installed in 2024 and another 199,000 service robots sold that year, and many humanoids were bought by research groups rather than doing real work.
  • Carmakers, seen as early adopters, are testing only single or double-digit numbers of robots per plant, though Bank of America expects 90,000 humanoid shipments this year and 1.2 million by 2030.
  • 🎲 Polymarket CEO pushed staff to commit fraud LINK

  • Polymarket CEO Shayne Coplan told his compliance team to keep expanding and simply pay a fine if regulators ever found out, after staff flagged a $10 million attempted fraud scheme, The Wall Street Journal reported yesterday.
  • A company handling debit-card payments for Polymarket's U.S. betting platform was rejecting 80% of transactions as fraudulent, far above the roughly 1% industry norm, and the Commodity Futures Trading Commission is now investigating the firm.
  • Polymarket faces several legal threats, including a New York City probe of its ads and inquiries from more than a dozen states questioning whether it and rivals Kalshi and Coinbase run unlicensed gambling platforms.
  • Other news you might like

    • Elon Musk’s latest Boring Company pitch involves a Hyperloop between Austin and San AntonioLINK
    • Apple home hub device built around Siri AI may be almost readyLINK
    • Inside Petal: Building the World’s First Petabit-Class Transoceanic Subsea CableLINK
    • Apple Mac mini (M6) review: The best compact desktop is no longer a great dealLINK
    • iFixit tears down iPhone 18 Pro, explores variable aperture, finds concerning repair issueLINK
    • SoftBank is reportedly raising about $11bn in bonds to cover its next OpenAI paymentLINK
    • Apple Fitness+ team hit by layoffs, as β€˜major changes’ reportedly loomLINK
    • Microsoft patent describes in-game advertising triggered by boss fights, loading screens, and cutscenesLINK
    • CXMT begins mass production of fifth-generation DRAM platform in ChinaLINK
    • Google fined €403m by DPC over GDPR violationsLINK

    🧰 Trending tools

    Google Flow for iOS & Android: brings Google Search to mobile devices, letting you find webpages, images, and videos while managing cookie and privacy preferences.LINK

    Sell to State: structured global government procurement data covering tenders, suppliers and agencies, helping you research what governments buy and find sales opportunities worldwide.LINK

    Flicka: browser-based screen recorder with cursor-following auto-zoom, timeline editing, screenshot annotation, and MP4, WebM or GIF export, all processed locally.LINK

    AppGrowthKit: an AI screenshot maker for App Store and Google Play listings, adding device frames, layouts, and copy for one-click, store-ready exports.LINK

    SecAIQ Watch: free, open-source local dashboard that tracks which AI tools run on your computer, what they access, connect to, and how many tokens they consume.LINK

    Refoid: automates App Store refund responses for iOS and macOS developers, tracking each request and Apple's decision with retries and email alerts.LINK

    πŸ“š Trending papers & reports

    Robot failure prediction flags when a vision-guided robot will botch a task under unfamiliar conditions, catching roughly 85% of failures partway through by watching both the odd input and how the job is going.LINK

    Robot manipulation training teaches robots to focus only on the objects and motions relevant to each task step, improving accuracy on precise, multi-step jobs without slowing the robot down at run time.LINK

    Brain-signal conversion generates the harder-to-capture blood-flow brain readings from easy-to-record electrical ones in a single fast step, boosting motor-imagery accuracy by up to ~4 points and enabling real-time hybrid brain-computer interfaces.LINK

    Solubility prediction forecasts how well a molecule dissolves across different solvents and temperatures, beating existing tools on two benchmarks and cutting the costly lab testing that drug and chemical developers rely on.LINK

    Latent-space rendering lets designers apply classic 3D graphics controls like lighting, camera angles, and geometry directly inside image-generation models, giving precise, physically accurate control over AI-generated visuals that were previously hard to steer.LINK


    See you tomorrow for a new dose of β˜•οΈ Techpresso!

    More from the archive