Meta's new Snipping AI ✂️

Hey readers 👋🏻

AI is evolving faster than we imagined. Tech companies are scrambling to capture the market with new product launches every week, while gov’ts are struggling to form regulations. At this point, we as individuals need to adopt a critical approach and be curious, since it’ll soon change the way we work, think, and create.

On to the update:

• Meta’s new snipping AI.

• Google’s AI supercomputer.

Reading time is 4 minutes. Let’s go!

The Main Thing
You should know

Meta’s new Snipping AI

Meta released ‘Segment Anything’, a new AI model that can “cut out” objects from images.

The AI model is tiny, with only 4 million parameters, making it small enough to enable local image segmentation. Additionally, Meta released the entire AI as open-source, including the trained model, datasets and source code. You could try the demo here.

Why it matters: Segmentation, which involves identifying which image pixels belong to an object, is a core task in computer vision. Meta’s AI model can accurately understand individual items in an image, opening up a wide range of new applications -

• New way to shop - You could select an item from a photo on Instagram and shop for it online.

• Ads - Meta could show you ads for items you browse on its platform, for instance, on Marketplace.

• AR/VR - Power new applications with the ability to identify items in Augmented Reality glasses, in real time!

Bigger Picture: Meta has a strategic advantage in AI, built upon troves of data from Facebook and Instagram. In the last couple of months, Meta has been positioning itself as developer-friendly by releasing new open-source products for free, in contrast to OpenAI's commercial APIs.

This move could help Meta transition into a larger AI platform.

Tech Roundup
Stay Informed

Google: Amidst the new frenzy for training AI models, GPUs are probably the most valuable resource (aside from your data). Google built a new supercomputer for training AI models using 4,000 custom-designed chips called Tensor Processing Units (TPUs). In an impressive feat of engineering, the supercomputer is about 1.7 times faster and 1.9 times more power-efficient than the current top-of-the-line Nvidia A100 chips, which sell for $10,000 each!

Around the Internet
Quick Juicy bites

• 🚌 Self-Driving Bus: Scotland to start a new driverless bus service next month.

• 🛒 Walmart: will build EV charging stations for its shoppers - 90% of Americans live within 10 miles of a Walmart store.

• 🚁 Mars Helicopter: Ingenuity Mars Helicopter, originally designed for only 5 flights, near its 50th flight on Mars!

• 🏞️ Text-To-Image: Microsoft’s Image creator tool is now available for general use - try it here.

• 🔌 FBI: Warns against ‘Juice Jacking’, aka malware via public phone chargers.
  

Misc