AI News You Probably Missed Over the Last Month

You can’t take your eyes off of the AI space for a minute without something new and extraordinary being done. 

Whether it’s testing new ideas, discovering novel solutions or disrupting the market – there’s always something crazy, yet inspiring to be found. With it, there’s always a golden nugget that brings you one step closer to making AI work for you.

Let’s take a look at some of the crazy new movements in the ever-widening world of AI.

AI Takes Baby Steps: Learning Like a Kid

We thought that, as sophisticated as it already is, AI is still taking baby steps. That it’s still in its infancy. 

While it might be the case,  AI is now quite literally learning through taking baby steps – by learning through the lens of an actual baby.

While your average AI, like the brainy GPT-4, feasts on trillions of words to get smart, a baby manages with a mere million or so words a year. There’s a huge difference in those ratios, making us wonder if AI can really give us clues about brain development.

Researchers at New York University threw us a curveball with their latest project

They decided to see what would happen if they trained an AI using only what a single child sees and hears. Thanks to some harmless headcam footage from a little one’s daily adventures, of course.

And guess what? It worked! 

This AI, by peeking through the lens of a child’s experiences (we’re talking only 1% of their awake time), began to pick up words and ideas just like a human infant. By using a concept known as contrastive learning, they were able to get the AI to connect the dots between what it sees and hears. 

This isn’t just child’s play; it’s a big leap into understanding how we can teach AI to learn more like us humans, using the vast and messy real world instead of neat, tidy databases. Supported by some serious backers like the U.S. Department of Defense and the National Science Foundation, this study is more than just academic eye candy – it’s a potential game-changer in making AI understand the world in a way that’s a little more human and a little less, well, robotic.

AI + Robots = Smoother Warehousing

Picture this: a bustling warehouse filled to the brim with robots zipping around, fetching items like there’s no tomorrow. 

Sounds like a scene from a futuristic movie, right? Well, it’s the reality for many industries today.

But there’s a catch. Managing an army of robots without obstacles isn’t easy.

Even the smartest path-finding algorithms break a sweat trying to keep these robotic speedsters from crashing into each other. 

It’s a massive headache that’s got the big brains at MIT thinking.

By taking a page out of their traffic management playbook, an MIT team applied it to warehouses. They whipped up a deep-learning model that’s pretty much a maestro at directing robot traffic. By breaking down the robot swarm into smaller groups, this AI system can quickly figure out where the digital congestion is and how to clear it up, making the whole operation run smoother and more efficiently. 

Beyond just making warehouses more efficient, this deep learning trick could change the game in other complex areas, like designing computer chips or figuring out the best way to lay pipes in a building. 

It’s all about making big, complicated systems work better, faster, and smarter.

Sponsored by Amazon and the MIT Amazon Science Hub, this project is more than just academic fun and games—it’s about bringing a new level of efficiency to the hectic world of e-commerce and beyond.

Google’s AI Hypercomputer Gets a Major Upgrade

Remember Google’s AI Hypercomputer

Well, Google just turned up the volume at their Cloud Next 2024 conference in Vegas, unveiling some juicy upgrades that has techies excited.

Getting robots and algorithms to play nice and smart is key to keeping the e-commerce and tech industries thriving. 

Google Cloud is already a playground for big players like Salesforce. Now, with these enhancements, Google’s setting the stage for an even more powerful AI training and serving experience:

  • Custom TPU v5p chips are going mainstream, ready to be unleashed on the masses.
  • A3 virtual machines are being boosted with NVIDIA’s latest H100 Tensor Core GPUs. 
  • Nvidia’s Blackwell GPUs are being added into the mix, to handle the heaviest AI, data analytics, and high-performance computing tasks.

Character.AI, a chatbot company worth over a billion bucks, is already singing Google Cloud’s praises. They’re leveraging Google’s tech to make their AI models leaner, meaner, and more efficient. With Google’s new gear, they’re expecting to double down on performance without doubling their spend.

And for those worried about the logistical nightmare of scheduling and managing AI workloads, Google’s introducing a Dynamic Workload Scheduler. This nifty feature lets users reserve GPU firepower on demand, promising to make the whole process more budget-friendly.

Get Your Fix of AI Content

If you’re a fan of AI, machine learning, cloud and other disruptive technologies, then we’ve got some great news for you.

As an AI-enablement professional services business, we understand the complexity behind the technologies that drive the AI value chain. Whether its cloud, data, or ML engineering, we’ve got you covered!

Check out our blog for more captivating tech content.

More in the Blog

Stay informed on all things AI...

< Get the latest AI news >

Join Our Webinar Cloud Migration with a twist

Aug 18, 2022 03:00 PM BST / 04:00 PM SAST