YOUR AI BRIEFING
Hey, it’s Louis.
Welcome to the first issue of “Cyborg,” where you’ll learn everything you need to know about AI—what’s happening, how people are using AI, and what it means for you and the future of humanity—explained in plain English. It’s been a busy few days for AI, so let’s get cracking.
WHAT’S HAPPENING IN AI
While America restricts its most powerful AI, open alternatives are closing the gap fast
Anthropic’s Fable and Mythos AI models, which “exceed the capabilities of any model they’ve ever made generally available,” remain unavailable worldwide after the US government restricted access to foreign nationals over security concerns. In the meantime, the Chinese company Z.ai’s new model GLM-5.2 is crushing coding benchmarks, and revealing that the gap between open-weight models (which people can download, adapt, and run independently) and closed frontier models (like ChatGPT and Claude) is falling away. In other words, today’s best open model may already match what the leading closed models could do just a few months ago.
WHAT THIS MEANS FOR YOU & HUMANITY
Powerful AI is becoming difficult to contain
Image by Colleen from Pixabay
Fable 5’s sudden disappearance shows us that governments can still restrict an AI company like Anthropic, but the rise of open-weight models like GLM-5.2 suggest that they may struggle to restrict the capability itself, at least soon.
For you, this could mean (much) cheaper tools, more choice between models, and less dependence on big US companies. But it also means that powerful AI will become much harder to regulate, and therefore potentially easier for anybody to abuse.
So, what do you think: Are increasingly powerful open-weight AI models a cause for concern, or healthy competition that puts AI into the hands of ordinary people? Hit reply and tell me why; I’d love to hear from you.
QUICK-FIRE NEWS
The other AI stories shaping this week, in under a minute
Midjourney is moving into healthcare. The AI company best known for generating images from text descriptions is developing a 60-second, full-body ultrasound scanner. Midjourney plans to open a “research spa” in San Francisco by the end of 2027, and have 50,000+ scanners worldwide by 2031—but the technology remains experimental, and increased capabilities will require regulatory approval; so we can only wait to see whether their goals are too ambitious, and how they’ll use AI.
Facebook is becoming an AI search engine. Meta is rolling out an “AI Mode” to answer questions using public posts from Facebook Groups, Reels and its other platforms. They’re also using AI to help you edit photos and videos, like creating “smooth, stylized montages that are ready to share.”
AI’s impact on jobs may begin at the bottom. New Stanford research found modest employment differences overall, but noticeably weaker growth among young people (aged 22-25) working in AI-exposed occupations. It looks like AI might narrow the first rung of the career ladder before it begins replacing established jobs outright.
CHECK THIS OUT
Your new dressing room… is an algorithm?!
Photo by Becca McHaffie on Unsplash
Upload a photo of yourself and an image of an item of clothing, and Kolors Virtual Try-On will generate a picture of you wearing it. It’s far from perfect, but a fascinating glimpse at how AI could remove some of the guesswork from shopping online.
Try it yourself for free here.
Until next week,
Louis from Cyborg