- AI Weekly Insights
- Posts
- AI Weekly Insights #21
AI Weekly Insights #21
Robotic Revelations, Open-Source Controversies, and Gaming Innovations
Happy Sunday!
Gear up for "AI Weekly Insights #21," where we're exploring the latest AI news and discoveries with the excitement of explorers on a new planet. This week we’re diving into Figure’s impressive humanoid robots, Elon Musk’s move to open-source AI, and Google’s versatile agent for video games.
The Insights
For the Week of 03/10/24 - 03/16/24 (P.S. Click the story’s title for more information 😊):
What’s New: Figure, a robotics startup established in 2022, has unveiled a compelling demo of their humanoid robot powered by the technology behind ChatGPT.
Figure 01: Dubbed the “world's first commercially viable autonomous humanoid robot,” the demo highlights the robot, equipped with cameras, adeptly handling tasks like dish storing, trash cleanup, and even offering an apple upon request.
Why It Matters: The demo is a must-watch, showcasing moments that genuinely impress with the robot's capabilities (there were a few moments that I said “wow” out loud). While the speech is not perfect and speed can be improved, it's a solid debut, hinting at future refinements for the 01 model.
What's New: Elon Musk announces the open-sourcing of Grok, xAI’s chatbot, positioned as a rival to ChatGPT.
xAI & Grok: Founded by Musk, xAI introduced Grok to offer an AI alternative focusing on truth over everything. Initially available to Premium+ subscribers on X, Grok contrasts ChatGPT with its unique, arguably more candid approach.
Why It Matters: This is a direct response to the ongoing legal battle between Musk and OpenAI. It is a critique of OpenAI's pivot from “an open-source, non-profit entity to a for-profit one under Microsoft’s influence”. While ChatGPT generally outperforms Grok, the latter’s distinct nature appeals to a niche audience.

Image Credits: Jakub Porzycki / Getty Images
What's New: Google DeepMind has showcased an AI agent, SIMA, capable of executing tasks in diverse video game environments through natural-language commands.
SIMA: Standing for Scalable Instructable Multiworld Agent, SIMA's training across nine games enables it to adeptly navigate in-game tasks like driving and climbing, outperforming game-specific agents. The aim is to evolve SIMA for complex strategic planning and multi-tasking.
Why It Matters: This breakthrough shifts the AI training paradigm from reward maximization to versatile task handling. Google’s research is creating AI to handle tasks in multiple game environments. Imagine a copilot that can help you in any video game to complete a task you’re struggling with. The image below shows some of the generalizations in actions made by SIMA

Image Credits: Google DeepMind
Your curiosity fuels our journey through the evolving AI landscape. Share your thoughts and questions, or just spread the word among fellow AI enthusiasts. Until next Sunday!
Warm regards,
- Kharee