In partnership with

Good morning, human brains, and welcome back to your daily munch of AI news.

Here’s what’s on the menu today:

• Hippity Hoppity, this bot won't stoppity 🏃 💨

  Tiangong is a humanoid robot that runs, goes up stairs, and more.

• GPT-4 Turbo got PWND by a Chinese model 🤯 👀

  SenseNova 5.0 outperformed it in almost all key benchmarks.

New here? Subscribe!

MAIN COURSE

When bots Jazzercise... 🏃 💨

Yesterday, The Beijing Humanoid Robot Innovation Center unveiled Tiangong. It’s a full-size electric humanoid robot that can run up to 3.7 miles per hour.

No hands? Must be where the flamethrowers go. 🤖🔥

How fast can it run?

Tiangong stands over 5 ft tall, weighs about 95 lb, and can maintain a running speed of 3.7 miles per hour.

What’s under the hood?

Tiangong is equipped with multiple vision perception sensors, a high-precision Inertial Measurement Unit, 3D vision sensors, and six-axis force sensors for accurate force feedback.

What else can it do?

In the demo, Tiangong jogged, went up and down steps, and more.

Isn’t Atlas from Boston Dynamics’ better?

Oh, yeah. China presented Tiangong as the world’s first full-electric humanoid robot, but Atlas earned that title first. Plus, it runs faster, has a broader range of motion, and more.

For more on Atlas, click here.

SPONSORED BY AE STUDIO

Have an AI Idea and need help building it?

When you know AI should be part of your business but aren’t sure how to implement your concept, talk to AE Studio.

Elite software creators collaborate with you to turn any AI/ML idea into a reality–from NLP and custom chatbots to automated reports and beyond.

AE Studio has worked with early stage startups and Fortune 500 companies, and we’re ready to partner with your team. Computer vision, blockchain, e2e product development, you name it, we want to hear about it.

Book time to tell us all about your idea

SIDE SALAD

Chinese model wrecks Gpt-4 Turbo 🤯 👀

Last week, a Chinese firm called SenseTime launched SenseNova 5.0. It’s an update to its LLM, which outperformed GPT-4 Turbo in almost all key benchmarks.

What is it?

It’s a 600B model with a 200k context window that was trained on over 10 TB of data.

How exactly did it beat GPT-4 Turbo?

In the live demo, SenseNova 5.0 came out on top in creative writing, mathematics, logical reasoning, and coding. It even crushed GPT-4 Turbo in a game of King of Fighters.

What’s unique about it?

It utilizes a specialized hybrid expert architecture utilizes curated datasets to improve its ability to understand and reason.

Wow. Anything else?

Yupp. They also announced new enterprise-level solutions that reduce AI inference costs by up to 80%.

RECOMMENDED READING

The Town Crier delivers satire on politics, business, sports, art, and science with a unique twist. Each issue presents a fresh, humorous perspective on the world's happenings.

Experience the lighter side of current events with The Town Crier. It’s free!

YOUR DAILY MUNCH

Tools 🛠️

Uptrends.ai — monitors thousands of sites for trending stocks.

Bundle of Joy — generates potential names for your baby boy or girl.

Homestage — add virtual furniture to any uploaded picture.

Think Pieces 🧠

What are the pros vs. cons of Sora? The intensive post-production and challenges of using OpenAI's realistic but limited video generation tool.

How does the public view Meta AI? It’s free, but it regurgitates search results without excelling in any particular area.

A review of the Rabbit R1. It’s $199, has a touchscreen and innovative design by Teenage Engineering, and is a fun and accessible alternative to Humane's Ai Pin.

Startup News 💰

Alphabet, Google’s parent company, earned $80.5 billion in Q1 2024. Sundar Pichai, the CEO, claims AI is a key driver of its success.

Sam Altman's Worldcoin seeks a collaboration with OpenAI. This is despite regulatory controversy for both parties.

Microsoft reported $61.9 billion in YoY (year-over-year) sales. Satya Nadella, the CEO, says it’s due to Microsoft’s AI strategy.

Research 👨‍🔬

LayerSkip — a method to increase the efficiency and accuracy of LLMs by allowing them to make predictions using fewer layers.

List Items One by One — enhances the visual reasoning skills of multimodal models by teaching them to describe visual tags on images sequentially.

ALPHALLM — enhances LLMs by combining them with Monte Carlo Tree Search, streamlining their ability to perform complex tasks without extra data.

MEMES FOR DESSERT

TWEET OF THE DAY

The memory feature. Yay, or nay?

Tag us on Twitter @BotEatBrain for a chance to be featured here tomorrow.

AI ART-SHOW

Until next time 🤖😋🧠