NVIDIA uses AI to train robots

Welcome to Daily Zaps — Here’s what we got for you today: 🙌

NVIDIA uses AI to train robots

NVIDIA Research has developed an AI agent named Eureka that can autonomously train robots in complex tasks, including teaching a robotic hand to spin a pen proficiently. Eureka, which utilizes the GPT-4 large language model, has enabled robots to master almost 30 tasks, such as opening drawers, catching balls, and using scissors. Unlike traditional methods, Eureka doesn't rely on predefined reward templates but generates reward algorithms for reinforcement learning, leading to a performance enhancement of over 50% in robots.

The AI system is self-improving and has been applied to a variety of robots, teaching them diverse tasks. The significance of Eureka lies in its combination of large language models and NVIDIA's GPU-accelerated simulation technologies, which offers potential in robot control and generating physically realistic animations.

Amazon warehouse humanoid robots

Amazon is testing Agility's bipedal robot, Digit, for potential use in its fulfillment centers. This move follows Agility's shift in focus from last-mile delivery to concentrating on warehouse and factory applications. Amazon previously recognized Agility as one of the inaugural recipients of its $1 billion Industrial Innovation Fund, indicating the company's interest in the robot's potential.

Tye Brady, Amazon Robotics' chief technologist, expressed that while their current focus has been on wheeled robots, they see potential in legged robots, especially for varying terrains. Amazon has extensively used wheeled AMRs (autonomous mobile robots), with 750,000 deployed across its warehouses. The company's push into robotics has significantly influenced the industrial robotics industry.

From Our Partners

Elevate your customer experience with Dropchat

Here's why AI-powered chatbots are a game-changer for your business:

Website AI: Chatbots can be trained to navigate your website, offering instant guidance, answering questions, and facilitating seamless navigation for your visitors.

PDF AI: Whether it's user manuals, reports, or documentation, our chatbots can extract information from PDFs, making it readily accessible to your customers.

24/7 AI: Your customers can access information and support around the clock, improving their experience and satisfaction.

Get started with Dropchat→

Microsoft all-in on AI

Microsoft's CEO, Satya Nadella, is all in on AI, as indicated in the company's annual report. Nadella envisions AI as the most significant technological advancement in computing in over a decade. While Microsoft has faced challenges in areas like mobile and search, its cloud business remains robust. Yet, the real transformation is foreseen in the realm of AI, reshaping every aspect of software and business.

Microsoft's collaboration with OpenAI is pivotal in this vision. Through OpenAI, Microsoft has gained access to groundbreaking natural language AI, giving it an edge over competitors like Google. Nadella's belief in AI goes beyond mere technological advancement; he sees it as the next phase in human-computer interaction, transitioning from keyboards and touchscreens to natural language interfaces.

Using DALL-E to create a business logo

DALL-E 3 is now available in ChatGPT Plus and Enterprise. To utilize DALL-E 3 within ChatGPT, select DALL-E 3 from the dropdown menu under the GPT-4 option. Once chosen, you can can input text descriptions, ranging from brief sentences to detailed paragraphs, to generate art through DALL-E 3.

Below is a quick demo of using DALL-E to create a business logo in ChatGPT.

1) Create a logo for a moving company called "Move It" with slogan "We Like To Move It" and mix in a lemur that looks like King Julien from movie Madagascar

2) Create variations of image number 1, show the full van and make it aspect ratio 6:3

3) Redo number 3 to correct spelling of text “Move It" and "We Like To Move It” and change number to 867-5309

Click on the image you like and save to your computer or continue refining using conversation.