- AI PlanetX
- Posts
- OpenAI Builds Robotics Team
OpenAI Builds Robotics Team
Google's Benchmark Reduces AI Hallucination
Welcome to another edition of AI PlanetX.
OpenAI launches robotics division; DeepMind targets hallucinations with new benchmark for AI accuracy.
Inside This Edition: đź’Ž
Hottest AI News
Top AI & SaaS Tools
Top AI & Tech News
Interesting Uses of AI
Top AI Video Tutorial
Launch Your Newsletter & Earn $20K/Month
Hottest AI News
OpenAI
OpenAI Begins Robotics Team with Key Hardware Hires
OpenAI is venturing into robotics hardware, a bold step announced by robotics lead Caitlin Kalinowski, marking a major expansion of its AI vision.
Details:
OpenAI is building its first robotics team, focusing on general-purpose robotics and AGI-level intelligence, and hiring senior engineers for sensors, mechanical design, and lab management
Previously, OpenAI worked with Jony Ive and partnered with Figure on humanoid robots. This new effort signals plans to create its own hardware
The team will combine advanced hardware and software, test various robot designs, and establish a data-gathering lab for development
OpenAI's move into robotics technology reflects its ambition to blend digital and physical AI, potentially competing with partners like Figure, similar to its dynamic with Microsoft.
Writer RAG tool: build production-ready RAG apps in minutes
Writer RAG Tool: build production-ready RAG apps in minutes with simple API calls.
Knowledge Graph integration for intelligent data retrieval and AI-powered interactions.
Streamlined full-stack platform eliminates complex setups for scalable, accurate AI workflows.
DeepMind
Google Unveils Benchmark to Reduce LLM Hallucinations
AI development struggles with hallucinations—factually inaccurate responses from LLMs. Google DeepMind’s new FACTS Grounding benchmark aims to improve factual accuracy, especially for complex tasks.
Details:
The benchmark tests 1,719 examples across various fields, with models processing up to 32,000 tokens and providing fact-based responses. Gemini 2.0 Flash leads at 83.6%, with Anthropic, OpenAI models scoring above 61.7%
Evaluation has two phases: models must meet user requests and ensure responses without hallucinations, grounded responses, judged by three LLMs, with final scores averaged
FACTS dataset includes prompts, requests, and context documents. Models must extract relevant details, avoiding vague or unsupported answers
The initiative is a major step in addressing LLM factuality, with an active leaderboard on Kaggle to track and evaluate new models.
Top AI & SaaS Tools
Airbrush (Life-time Deal): Create stunning images and artwork in seconds with AI models like Stable Diffusion, Midjourney, and FLUX [190+ five-star reviews]
Gemini Search: Perplexity clone using Gemini 2.0 + Grounding—search anything, get sources, ask follow-ups [F-R-E-E]
21st dev: Ship polished UIs faster with React Tailwind components built by design engineers [F-R-E-E]
Kokoro: An 82M text-to-speech model that produces one hour of high-quality audio per minute [F-R-E-E]
cobalt: Save video, audio, and GIFs hassle-free—no ads, tracking, or paywalls. Just paste the link and go [F-R-E-E]
Top AI & Tech News
NovaSky unveiled the Sky-T1-32B-Preview model, demonstrating impressive reasoning in math and coding for under $450
Elon Musk's xAI is increasingly shaping X (formerly Twitter), using it as a testing ground for his AI initiatives
OpenAI, Google paying thousands for YouTubers and content creators’ videos to train AI models
Jasper Zhang says AI agents are already renting GPUs on their own and doing AI development in PyTorch
Jensen Huang: "The technologies necessary to build general humanoid robotics is just around the corner"
Interesting Uses of AI
AI Art Spotlight
Model: Midjourney
Prompt:
photography, vase, large bouquet of faded peonies, earth, cup of tea, a bird, torn dirty tablecloth, smoke, soft light, dark textured background, high def, 8K
AI Prompt of the Day
Email Welcome Sequence
This prompt is about crafting a 3-email sequence for new subscribers to build a connection, showcase a flagship product/service, and offer an incentive. (Don’t forget to give ChatGPT information on your brand, offerings, and discount.)
Prompt:
Craft a 3-email welcome sequence for new subscribers. Briefly introduce our brand in the first email, highlight our most popular product/service in the second, and offer a discount code in the third.