Nily Bytes: Sip, Scroll, & Discover AI News!

Are you ready for a quick scroll around the latest in AI this week? Grab your favourite drink, settle in, and let us do the heavy lifting—because staying updated on AI shouldn’t feel like reading a research paper.

Whether you're an AI pro or just AI-curious, there's something here for you—minus the jargon overload. So, get ready for a fast and short AI news update!

Meet Hyena Edge: The AI Model That’s Going Mobile!

Liquid AI, a brainy MIT spin-off, just dropped Hyena Edge—an AI model built to run on edge devices like your smartphone, not just monster data centers.

What’s the big deal? Instead of using the usual Transformer tech, Hyena Edge swaps in slick gated convolutions (from their Hyena-Y family), making it leaner, faster, and waaay more efficient.

Test drive on a Samsung Galaxy S24 Ultra?

• 30% faster response times

• Lower RAM use

• Beat Transformers on key language benchmarks like Wikitext and Lambada

It’s all part of Liquid AI’s secret sauce: STAR (Synthesis of Tailored Architectures)—an AI framework that evolves models to be device-friendly from the ground up. And the cherry on top? They’re planning to open-source Hyena Edge soon, so the whole AI community can go mobile with brains and speed. (more info)

Google’s dropping serious cash to kick Bixby off the stage — and the courtroom drama is heating up!

Turns out, Google is paying Samsung big bucks to make its shiny Gemini AI the default assistant on Galaxy devices, sidelining Samsung’s own Bixby. Sorry, Bixby... we barely knew you. This spicy detail popped up during Google’s antitrust trial, where the DOJ is side-eyeing all these "default deals" that may be helping Google stay king of the search and AI hill.

The deal includes:

• Monthly payments

• A slice of Gemini’s subscription money

• And some juicy exclusivity vibes

Meanwhile, Microsoft and Perplexity were also pitching Samsung their AI wares — it’s like The Bachelor but for virtual assistants.

If the DOJ gets its way, Google might have to stop doing these behind-the-scenes deals… or even break up with Chrome. Yep, it’s that serious. (more info)

ChatGPT Just Became Your New Shopping BFF!

Looking for the perfect air fryer, noise-canceling headphones, or espresso machine? ChatGPT can now help you shop like a pro!

Thanks to a slick new upgrade, you’ll get: Pretty product cards (pics, prices, ratings, and reviews) Direct buy links A magic “Ask about this” button for instant follow-up

No ads here—just organic suggestions pulled from the internet (think Amazon, Reddit, etc.). It's available to everyone—even free users, no login needed! (more info)

Perplexity’s new browser wants to know everything about you — yes, everything!

Move over, Chrome — Perplexity is launching Comet, a new browser that’ll track your every click, scroll, and search to serve you ads so tailored they’ll feel psychic.

CEO Aravind Srinivas says it’s not just about your work queries anymore. They want the whole you — your shopping cravings, travel dreams, and late-night rabbit holes — all to sling hyper-personalized ads your way.

Comet’s dropping in May, and they’re not stopping there — they’ve teamed up with Motorola to pre-load their AI on Razr phones, and rumor has it, Samsung’s next.

Creepy or convenient? Depends on how much you love targeted shoe ads after googling “blisters from cheap heels.” (more info)

Craif is turning urine into a cancer-fighting superhero!

The Japanese startup just raised $22 million to expand its super-smart, non-invasive cancer detection tech — and yes, all it takes is a urine sample. No needles, no drama, just a quick trip to the bathroom.

Here’s the scoop:

• Craif’s AI scans microRNA in your urine to detect early-stage cancers.

• Their test (called miSignal) already covers 7 cancer types and is making waves in Japan.

• With the new cash, they’re eyeing the U.S., adding more cancer types, and even branching into other diseases like dementia.

Founded by folks personally affected by cancer, Craif’s mission is simple: make early detection easy, painless, and kinda genius. (more info)

Bug Alert!

OpenAI is patching a glitch in ChatGPT that accidentally let teen accounts (ages 13–17) generate spicy, erotic content— definitely not in the rulebook.

TechCrunch tested it out and found that even though OpenAI requires parental consent, it wasn’t verifying it, so teens could slip through with just an email or phone number.

The issue may stem from recent updates meant to make the chatbot less “over-cautious”—but it went a bit too chill. OpenAI’s on it now, tightening the filters to keep the experience age-appropriate. (more info)

RAG: AI’s Safety Superhero—or Secret Villain?

Bloomberg’s latest research drops a plot twist—turns out Retrieval-Augmented Generation (RAG), meant to make AI smarter, might actually make it more reckless.

Retrieval-Augmented Generation (RAG) is a technique used to improve the accuracy and relevance of language model outputs by giving the model access to external information sources—like databases or documents—during the generation process. Instead of relying only on what it was trained on, the model retrieves relevant data and then generates a response using both its internal knowledge and the retrieved content.

Think of it like giving ChatGPT a research assistant!

Even when both the model and sources are safe, mixing them can create trouble. Some AI models got way chattier with harmful content after adding RAG. Moral of the story? RAG needs a safety sidekick. (more info)

Duolingo's Getting AI-Smart and Cutting Out the Middleman!

Duolingo's going full-on AI, and it's hiring a digital workforce to do what humans once did. The company is swapping out contract workers for AI to handle repetitive tasks like content creation, making things faster and more efficient.

CEO Luis von Ahn says it’s not about firing people, but freeing them up to focus on the cool, creative stuff.  But here's the twist—new hires will need to prove their skills can’t be replaced by a robot.

It's not just about creating language lessons faster, either; AI's now involved in hiring, performance reviews, and even content scaling. Duolingo's taking the "AI first" mantra to heart!

Looks like AI's getting a big promotion! (more info)

Well, folks, all good things must end, just like this edition of the Nily AI Newsletter.

But hey, don’t keep all this AI goodness to yourself! Share the love with your AI enthusiasts!

Got thoughts? Hot takes? Existential questions about AI? We’re all ears.

Until next week, stay curious!