Welcome back!

The age of affordable humanoid robots is no longer a distant dream. Unitree Robotics just announced that their $16,000 G1 humanoid robot is entering mass production. Robotics is becoming increasingly within reach for home users. Let’s see how this unfolds...

Today’s Summary:

• 🤖 $16,000 humanoid robot enters mass production

• 🔧 Fine-tuning now available for GPT-4o

• 🎥 Luma AI updates Dream Machine 1.5

• 🛠️ Hugging Face tutorial: build your own robot

• 🧑‍💼 Stability AI names new CTO

• 📰 OpenAI to train on Condé Nast content

• 🩺 Google AI HeAR model detects diseases by cough sounds

• 🛠️ 2 new tools

TOP STORY

$16,000 Humanoid Robot Enters Mass Production

The Summary: Unitree Robotics has unveiled a production-ready version of its advanced G1 humanoid robot, featuring capabilities such as jumping, stair-climbing, and dexterous hand movement. Priced at $16,000, the G1 makes advanced humanoid robots more accessible beyond research labs and factories.

Key details:

• G1 foldable humanoid stands 1.32m tall, weighs 35kg / 77 lbs

• Equipped with 3D LiDAR depth camera, mics, 5W speakers

• Move at speeds up to 2m/s (4.5mph), handle a 2kg arm payload

• 9,000mAh quick-release battery for 2 hours of runtime

Why it matters: The G1 capabilities and price point could make it an attractive option for robotics researchers and developers, potentially accelerating innovation in the field. Although this is still an early version, it offers a glimpse of how humanoid robots might soon become accessible to home users, as the hardware matures and costs decrease.

Specifications

OPENAI

Fine-tuning now available for GPT-4o

The Summary: OpenAI launched a fine-tuning service for GPT-4o, enabling developers to customize the model for specific use cases, resulting in increased performance and accuracy.

Key details:

• Fine-tuning for GPT-4o costs $25 per million tokens, with inference at $3.75 per million input tokens and $15 per million output tokens

• OpenAI is offering 1M free training tokens per day through September 23

• Fine-tuning was previously available only for GPT-3.5 and 4o mini.

• Fine-tuning improved performance of Cosine Genie AI software engineering assistant, achieving a new state-of-the-art 44% score on SWE bench

• Distyl fine-tuned GPT-4o to rank 1st on the BIRD SQL benchmark with 72% execution accuracy

Why it matters: The ability to fine-tune GPT-4o allows developers to further enhance the capabilities of this advanced AI model, customizing it to meet specific needs.

Fine-tuning documentation

VIDEO AI

Luma AI Releases Dream Machine 1.5

The Summary: Luma AI has released version 1.5 of its Dream Machine text-to-video generator, featuring enhanced realism, improved motion tracking, and smarter understanding of prompts.

Key details:

• Now capable of rendering text within generated videos, ideal for dynamic title sequences, animated logos, and on-screen graphics

• Improved handling of non-English prompts

• 5x speed boost, generating 5 seconds of high-quality video in 2 minutes

Why it matters: The Dream Machine 1.5 update opens up new creative possibilities, enabling the creation of high-quality title sequences, infographics, and multilingual learning materials.

QUICK NEWS

Quick news

• Hugging Face releases a tutorial on building your own robot

• Stability AI appoints new CTO

• OpenAI signs deal to train on Condé Nast content

• Google AI HeAR model helps to detect diseases based on cough sounds

TOOLS

🥇 New tools

• Mynd - 24/7 AI bestie for mental wellness

• AgentQL - Painless data extraction and web automation