THE NEWS

🦾 Figure had an AI breakthrough

Figure, a small yet mighty team of 70, has been buzzing in the tech scene, especially since they dropped their cool Figure 1 robot last year after just a year of tinkering.

Two days ago, their CEO caused a stir with a tweet hinting at a major breakthrough.

Yesterday they finally revealed it. Their robot can now… make coffee.

I know… not what you expected. While it might sound simple, the cool part is how it learned: by watching videos of humans doing it.

This is groundbreaking because it means robots can be trained for various tasks like making coffee, folding laundry, or warehouse work, just by feeding them human data. There's no need for task-specific programming; the AI just figures it out.

Adcock, the CEO, shared his excitement in a recent post on X: “The last 90 days I’ve witnessed industry leading AI in our lab running on humanoid hardware, and frankly it’s blown me away.” 

He admits his thinking's done a 180. At first, he figured they'd sort out the robot's body before getting the brain part down. Now, he's betting advanced AI might be ready even before the robots are mass-produced.

Looking ahead, he predicts that 2024 will be a pivotal year for the world of robotics.

💸 Stanford grad students prove robots can be cheap

Three Stanford grad students have made a small breakthrough in robotics, crafting a robot from off-the-shelf parts for just $32,000 – a third of the cost of typical commercial robots. To cut down on cost and complexity they ditched the conventional 5 finger robot hands for simple clamps.

By repeatedly showing the robot a task 20 to 50 times, they trained it to perform everyday chores such as wiping spills, opening cupboards, and even calling an elevator.

The real game-changer in their study, though, is proving that even with these basic clamps, the robot can handle complex tasks like making the bed, folding laundry, and cooking. This finding suggests that full humanoid robots aren't always necessary, paving the way for more affordable robotic solutions in various fields.

Here’s a video showing that (Click on the tweet to watch it):

ONE MORE THING

You can use a local LLM on your iPhone Pro

AI is going mobile in 2024 and this seems to be the first step. An app that just launched in the App Store runs Mistral 7B 0.2 model locally on your iPhone Pro. Mistral is the best small open-source model right now, outperforming even the larger 13B models.

Why is this cool? It’s totally private, you can be sure nobody is using your chat history to collect data or use it for AI training. And it runs on your phone without any internet connection. You can be in the middle of nowhere or arriving at an exotic country where your SIM card doesn’t work and you’ll still have a chatbot with the vast knowledge of the world at your hands.

⌚ If You Have One More Minute

• 💻 What to expect at CES 2024

• ⚖️ Judges in England and Wales are given cautious approval to use AI in writing legal opinions

• 🇨🇳 Nvidia to launch China-focused AI chip in Q2 2024

• 🤖 Deloitte rolls out artificial intelligence chatbot to employees

AI Art of the Day 🎨

This app shows how much we can trust video calls now. Pretends+ turns you into a celebrity in real time.

🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🌊🏄🌊🏄🌊🏄

That's all for today, folks!

If you enjoyed this, please consider sharing this hand-crafted newsletter with a friend.