iPhone's Next-Gen AI Features

Today:

Apple research reveals some dazzling AI tech could be headed to your iPhone

Apple's diving into AI big time, with research pointing to some fancy tech possibly hitting iPhones soon. They're working on two major things: first, a way to get hefty AI language models running smoothly on iPhones and iPads. Think of it as making Siri smarter and more helpful, but without eating up phone memory. Second, they've got this cool method, called HUGS, that can whip up lifelike avatars from a short video. It's fast and doesn't need much to work with, making it perfect for creating personalized avatars for social media, games, and AR stuff. 

This means iPhones might get way smarter, with Siri getting a major upgrade and the ability to handle more complex AI tasks. Plus, these avatars could really change the game in how we interact digitally. Apple's not making a big fuss about it yet, but this research shows they're seriously upping their AI game.

READ THE ARTICLE ON ZDNET

Red Cell Partners Raised A $91 Million Fund To Bring More AI To Healthcare And Defense

Red Cell Partners, a venture capital firm, has raised a hefty $91.2 million to boost AI in healthcare and defense. They're big on building startups from scratch, offering nuts-and-bolts support to top-notch engineers and thinkers. 

This new fund, part of their over $200 million kitty, will mainly back in-house pre-seed ideas, with a keen eye on healthcare and defense tech, especially AI. They're shooting for rapid, revenue-generating launches in under two years. With heavyweights like former Defense Secretary Mark Esper and ex-Federal Reserve Vice-Chairman Roger Ferguson on board, they're all about breaking new ground and tackling big challenges in health and national security.

READ THE ARTICLE ON FORBES

Rite Aid’s A.I. Facial Recognition Wrongly Tagged People of Color as Shoplifters

Rite Aid got slammed by the FTC for misusing facial recognition tech, mainly tagging folks of color and women as potential thieves. This tech, meant to spot shoplifters, was calling out the wrong people, mostly Black, Asian, and Latino folks, making them feel embarrassed and discriminated against. The FTC's making Rite Aid stop using this tech for five years. 

This whole mess highlights how tech can inherit human biases, causing real-world problems. Rite Aid's now in hot water for this, even though they stopped using the tech before the FTC stepped in. This case is a big wake-up call about the risks of AI surveillance and how it can unfairly target certain groups.

READ THE ARTICLE ON THE NEW YORK TIMES

New ultra-high speed processor to advance AI, driverless vehicles and more

Scientists just whipped up this super-fast signal processor that can handle 400,000 video images at the same time. We're talking about a speed that's over 10,000 times faster than your average electronic processors – a whopping 17 trillion bits per second! This thing is a total game-changer for driverless cars, making them safer and smarter. It's not just cars, though; it could even help us spot planets like Earth in outer space.

The brains behind this is Professor David Moss and his team. They're using something called an optical microcomb, which is way better than traditional electronic stuff. It's not just about speed; it's about doing things smarter and more efficiently.

This processor's going to make a huge splash in AI and robotics, letting machines think and act in real-time, like in the real world. And in space? We're talking next-level discoveries, processing massive amounts of space data super fast. It's not just about tech bragging rights – it's about reshaping transportation, healthcare, space exploration, and AI. It's the future, and it's happening now.

READ THE ARTICLE ON TECH XPLORE

AI Health Coaches Are Coming Soon to a Device Near You

AI health coaches are the next big thing, ready to tell us how to stay healthy using all the data from our fitness trackers. Remember when counting steps was odd? Not anymore—over 40% of U.S. homes have a wearable gadget. These devices have been helpful, like making folks walk more and lose a couple of pounds. But the real magic is in AI, which can do way more than just count your steps—it's about understanding your health in your own world.

Researchers are pushing AI into the spotlight. They're teaching it to think, sort of, so it can give personalized health tips. For example, it might tell you what your heart rate really means for you. AI health coaches could be smarter than human ones, giving real-deep insights into your health. There's some early success with simple chatbot coaches improving sleep and exercise, but there's more work to do.

READ THE ARTICLE ON TIME

AI consciousness: scientists say we urgently need answers

Scientists are scratching their heads about whether AI, like the brains behind ChatGPT, could actually start thinking and feeling for real. A group called the Association for Mathematical Consciousness Science is telling the big shots at the UN that we're flying blind here. They're pushing for more cash to study if and how AI could become self-aware. 

Tech firms, including the ones who made ChatGPT, are gunning to create AI that can do a bunch of smart stuff like us humans. But, we're not even sure how to tell if a machine is conscious. And, there's not much dough being thrown at figuring this out.

READ THE ARTICLE ON NATURE

Running Mixtral 8x7 locally with LlamaIndex

Mistral AI, a European AI giant, dropped a hot new model, Mixtral 8x7b. It's a combo of eight expert models, each with 7 billion parameters, and it's giving GPT-3.5 and Llama2 70b a run for their money. For open source fans, Mixtral's a dream. Here's the lowdown on running it with LlamaIndex on your own turf.

First, grab Ollama – it's an easy-to-use, free, open source installer for MacOS and Linux (Windows folks, use the Windows Subsystem For Linux). Install Mixtral with a simple command, but heads up, it needs a beefy 48GB RAM. Got less? Try Mistral 7B instead.

Next, get LlamaIndex and some other essentials using pip. Test things out with a quick smoke test script to make sure everything's chatting nicely. Then, load up some data (like tweets) and store it in Qdrant, a cool open source vector database. You'll embed and index this data with LlamaIndex, using Mixtral for some smart responses.

After indexing, use Qdrant and Mixtral to answer questions from your data. So, you'll learn to run Mixtral with Ollama, query it using LlamaIndex, build and use an index with Qdrant, and wrap it all in a basic web API. All with free, open-source tools, right on your own machine. Get ready for some local AI action!

READ THE ARTICLE ON MEDIUM