🤖 Tesla Unveils Optimus Gen 2

On Tesla's new groundbreaking robot, trending spaces and top models on HuggingFace

Author

Lior
December 14, 2023

AlphaSignal

Hey ,

Welcome to this week's edition of AlphaSignal.

Whether you are a researcher, engineer, developer, or data scientist, our summaries ensure you're always up-to-date with the latest breakthroughs in AI.

Let's get into it!

Lior

In Today’s Email:

  • Tesla’s new robot: Optimus Gen 2

  • Trending spaces on HuggingFace

  • Trending models on HuggingFace

Read Time: 3 min 03 sec

NEWS
Tesla Unveils Optimus Gen 2

What's New?

On Tuesday, Tesla released a demonstration video of its latest humanoid robot prototype, Optimus Gen 2, marking a significant advancement from its initial public showing. The previous version, introduced over a year ago, exhibited basic functionalities with noticeable limitations in movement and stability.

In contrast, the new Gen 2 robot demonstrates enhanced hardware and capabilities, as confirmed by Tesla Senior Staff Software Engineer Julian Ibarz: "Everything in this video is real, no CGI. All real time, nothing sped up. Incredible hardware improvements from the team."

The video displays the progression from Tesla's first humanoid prototype 'Bumblebee' to 'Optimus - Gen 1', culminating in the introduction of 'Optimus - Gen 2'. This latest model showcases various tasks, including slow walking, squatting, and the careful manipulation of delicate objects like eggs, indicative of significant improvements in balance, motor control, and tactile sensing.

Key technical specifications of the Optimus Gen 2 include Tesla-designed actuators and sensors, a 2-DoF actuated neck for enhanced head movement, integrated electronics and harnessing within the actuators for streamlined functionality, and a notable 30% increase in walking speed.

The robot's feet feature foot force and torque sensing, articulated toe sections, and are designed to mimic human foot geometry, contributing to a more natural gait. The overall weight of the robot has been reduced by 10 kg, improving its agility and energy efficiency. The hands of the robot are also upgraded, now featuring faster, 11-DoF hands with tactile sensing on all fingers, enabling delicate object manipulation such as holding an egg without causing damage.

While the Optimus Gen 2 shows promising progress, it remains a prototype and is not intended for immediate production or sale. It represents Tesla's ongoing efforts in developing a functional humanoid robot, envisioned by CEO Elon Musk to perform tasks that humans prefer to avoid.

The design aims to replicate human shape and size, intending to seamlessly replace human labor in various applications. However, given the complexities of engineering such advanced robots, the timeline for achieving this goal remains uncertain.

Manual annotation is dead. One-shot is the future!

SuperAnnotate just dropped an object search capability that allows selecting a reference object and searching anything that resembles that object in the dataset.

Such a new approach in labeling can fully change the traditional ways of doing image annotation and would further accelerate the AI development for computer vision.

This approach is particularly impactful for solving the long-tail distribution problem in computer vision, where rare objects can be found instantly.

TRENDING MODELS

A generative Sparse Mixture of Experts Large Language Model, which has a superior performance compared to Llama 2 70B in various benchmarks. It offers compatibility with vLLM serving and the Hugging Face transformers library.

Playground v2 is a diffusion-based model generating 1024x1024 resolution images from text prompts. It surpasses Stable Diffusion XL in aesthetic quality and introduces the MJHQ-30K benchmark for evaluating image aesthetics.

A top-performing 7.04 billion parameter text generation model, excelling in both accuracy and efficiency with support for 8K-token sequences. It employs Grouped-Query Attention and AutoNAC-optimized architecture, while offering exceptional computational efficiency and is suitable for both commercial and research applications in English.

TRENDING SPACES

Marigold is a model for monocular depth estimation, leveraging the extensive priors of generative diffusion models like Stable Diffusion for more generalizable results. It achieves state-of-the-art performance, even on unfamiliar content, and can be fine-tuned efficiently on a single GPU.

NexusRaven is an open-source and commercially viable function calling LLM that surpasses the current state-of-the-art approaches in function calling capabilities.

LaVie introduces a novel approach to text-to-video generation, building on pre-trained text-to-image models. It combines base T2V, temporal interpolation, and super-resolution models to create visually realistic, temporally coherent videos while retaining creative generation capabilities.

How was today’s email?

Not Great      Good      Amazing

Thank You

Want to promote your company, product, job, or event to 150,000+ AI researchers and engineers? You can reach out here.