🧨Nvidia's AI Grip Under Fire

Reading time: 5 minutes  

 Newsletter | News | Tools | Grow My Tool  

Today we will discuss-

• 🏹Tech titans target Nvidia's AI dominance

• 🥳MIT's solution for home robot success

• 🍿OpenAI unveils Sora's latest creations

• ⚙️9 amazing AI tools you might not have heard of

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

Intel, Google, Arm collaborate on open-source solution to end Nvidia's AI dominance

🤠Context of the news - Last month, Nvidia became the first chipmaker to hit a $2 trillion market capitalization. This growth was driven by its focus on AI hardware like the H100 and upcoming H200 GPUs. But that's not all; these powerful chips also lock developers into using Nvidia’s CUDA architecture. As a result, major tech companies are now trying to eliminate this software advantage that has helped Nvidia dominate the AI market.

👨‍💻News - According to a recent report, a group formed by Intel, Google, Arm, Qualcomm, Samsung, and other tech companies is developing an open-source software suit. This initiative is aimed at preventing AI developers from being locked into Nvidia’s proprietary technology, enabling their code to run on any machine and with any chip.

The group, known as The Unified Acceleration Foundation (UXL), has confirmed that technical details for the project are expected to reach a "mature" state by the second half of this year, although they have not provided a final release target.

The project currently includes Intel's OneAPI open standard, which aims to remove dependencies on particular coding languages, code bases, and tools, thereby preventing developers from being locked into specific architectures like Nvidia’s CUDA platform.

🎯What's their goal? The project's initial goal is to broaden options for AI apps and high-performance computing applications. However, the group intends to eventually extend support to Nvidia’s hardware and code, too. 

Additionally, UXL is reaching out to other chipmakers and cloud-computing companies such as Microsoft and Amazon for assistance, aiming to ensure the solution's compatibility with any chip or hardware. 

MIT researchers showcase how LLMs can help home robots recover from errors without human help 

🤖News - It's no secret that home robots since Roomba have seen little success, with challenges such as pricing, practicality, form factor, and mapping leading to repeated failures. Even if some of these issues are resolved, the question remains: how to handle inevitable mistakes? Fortunately, new research from MIT demonstrates that large language models (LLMs) offer a promising solution.

🕵️‍♂️What's the solution? The solution proposed in a study set to be presented at the International Conference on Learning Representations (ICLR) in May aims to introduce "common sense" into the process of correcting mistakes in home robots.

Unlike traditional imitation learning methods, which struggle with environmental variations, the new approach breaks demonstrations into smaller subsets. This strategy, coupled with the use of large language models (LLMs), enables the system to automatically understand and label subtasks, eliminating the need for manual intervention by programmers. 

👨‍🔬Has this theory been put to the test? Yes, and the solution did hold up in practice! A demonstration featured in the study involved training a robot to scoop marbles and pour them into a bowl, where LLMs were used to accurately list and label the process into subtasks. Researchers then introduced disruptions, such as bumping the robot off course and causing marbles to spill from its spoon. Despite these challenges, the system promptly self-corrected the errors within the smaller tasks (as labeled by the LLMs) instead of resetting entirely, showcasing its ability to recover autonomously without additional human intervention.

OpenAI shares first professional shorts created with Sora

🎥News - It's been a little over a month since OpenAI released Sora, its text-to-video tool. During this time, a select group of artists, designers, and filmmakers have had the opportunity to experiment with it. Yesterday, OpenAI shared some of their creations and first impressions. 

The videos submitted by these early testers will help OpenAI assess the capabilities, limitations, and potential risks associated with Sora. In its blog, the company wrote that while there are many improvements to make to Sora, the videos provide a glimpse of how the model can assist creatives in bringing ideas to reality.

📀Is it actually any good though? While all the films, ranging from 20 seconds to a minute-and-a-half, are visually striking, many lean towards the abstract. However, one standout is the multimedia production company shy kids' "Air Head," which could arguably be deemed the most entertaining. 

The clip portrays a man with a hot-air-filled yellow balloon for a head, seamlessly blended with a seemingly human body and realistic surroundings. Walter Woodman of shy kids remarked that while Sora excels at generating realistic imagery, its true potential lies in creating surreal elements — which, we agree, it indeed does! Watch for yourself.

shykids_

View more on Instagram

🙆🏻‍♀️What else is happening?

• Apple to use Baidu's AI for iPhone 16 series in China, Google’s Gemini to power devices in other regions

• Meta joins the AI talent war with quick offers, emails from Zuckerberg

• Bloomberg asks US court to toss copyright lawsuit over AI training

• Doctor Who promos spark massive backlash for using AI, BBC responds

• Canadian firm launches AI-powered ETF targeting AI investments

👨‍🏫Did you know?

• ChatGPT doesn't actually "understand" the information it generates. It bases its understanding on statistical patterns rather than true comprehension. 

• Google Assistant is the most accurate voice assistant, with an accuracy rate of 98%. Amazon Alexa follows at 93%, while Apple Siri lags behind at 68%.

• During the China Hi-Tech Fair in Shenzhen, a robot named Xiao Pang, or "Little Fatty," damaged a display booth and injured a young boy. Witnesses also observed that the robot, which was designed for children aged 4 to 12 appeared to be frowning after the incident.

👩🏼‍🚒Discover mind-blowing AI tools

1. Chatbit - A customizable AI chatbot tool designed to enhance website engagement by answering questions and capturing leads ($19/month)**

2. Bluedot - A chrome extension for Google Meet that allows users to record, transcribe, and summarize meetings ($14/month)

3. GPT Prompt Engineer - A tool that uses GPT-4 and GPT-3.5-Turbo to generate and test prompts for specific tasks (Free)

4. Klap - A tool that uses AI to identify engaging topics within YouTube videos and creates snippets that have the potential to go viral on TikTok ($23.2/month)

5. Photo2Math - Assist users in solving complex math problems by providing step-by-step solutions (Free)

6. ProductBot - A tool to help you decide what to buy on Amazon and get recommendations based on your preferences (Free)

7. Twinning - A social media tool that allows influencers to create an AI clone of themselves which their followers can chat with (Free for influencers with 1M+ followers, $299 for influencers with 100K+ followers) 

8. Skeptical Tom - An AI assistant that helps users make thoughtful choices while shopping (Free)

9. Appy Pie  - A no-code development platform that allows users to create web and mobile applications without any coding ($190/app/year) 

🎮AI-based games

• Alien: Isolation - It is a survival horror game set in the Alien universe. You play as Amanda Ripley, daughter of Ellen Ripley, as she investigates the disappearance of her mother on the space station Sevastopol. The game is praised for its atmosphere, suspense, and faithful recreation of the Alien universe.

• Halo: Combat Evolved - It is a first-person shooter game where you play as Master Chief, a cybernetically enhanced supersoldier. The game's AI is used to create a sense of atmosphere and immersion, making you feel like you are actually in the game world.

• Rocket League - A vehicular soccer game where you control rocket-powered cars to hit a ball into the opponent's goal. Here, AI has been used to create challenging and intelligent opponents, as well as to provide training tools for players.

[The tool marked with ** is a sponsored feature.]