OpenAI-backed 1X’s Robot Could Pass as Human
Plus: Google releases ‘improved’ trio of Gemini models, OpenAI to unveil ‘Project Strawberry’ this fall, New AI model simulates DOOM video game, and more.
Hello Engineering Leaders and AI Enthusiasts!
This newsletter brings you the latest AI updates in just 4 minutes! Dive in for a quick summary of everything important that happened in AI over the last week.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
🧑🏫 OpenAI partners with ASU to use ChatGPT in classrooms
🔍 Google releases ‘stronger and improved’ trio of Gemini models
🍓 OpenAI to unveil ‘Project Strawberry’ this fall
🎮 New AI model simulates DOOM video game in real-time
🎥 Qwen2-VL beats GPT-4; analyzes 20-min long video
🤖 1X’s new robot is strikingly human-like
📚 Knowledge Nugget: Top AI tools people actually use by
Let’s go!
OpenAI partners with ASU to use ChatGPT in classrooms
Arizona State University (ASU) is integrating OpenAI's ChatGPT Edu, a version designed for universities with enhanced privacy and security features, into over 200 projects across its campus. The university has launched an 'AI Innovation Challenge' for faculty and staff, which has received an overwhelming demand for ChatGPT to enhance teaching, research, and operations.
Key projects include an AI writing companion for scholarly work, an AI-powered chatbot named 'Sam' for medical students to practice patient interactions and AI-assisted research recruitment. The partnership has inspired other institutions like Oxford and Wharton to pursue similar collaborations and make ChatGPT a core part of their educational ecosystem.
Why does it matter?
While some universities resist AI, ASU embraces ChatGPT to unlock new academic possibilities. As education evolves in the age of AI, this partnership will be instrumental in transforming the education system and equipping students with the skills they need to thrive in an AI-powered world.
Google releases ‘stronger and improved’ trio of Gemini models
Google has just released three new Gemini 1.5 AI model experimental versions. These updates include:
Gemini 1.5 Flash-8B: A smaller, faster version of the Gemini 1.5 model that can handle text, images, and other data types efficiently for quick responses.
A Stronger Gemini 1.5 Pro model: Better at writing code and understanding complex instructions. It has now climbed to the #2 spot on the Chatbot Arena leaderboard.
A Significantly Improved Gemini 1.5 Flash model: Better on Google's internal tests across various tasks and has jumped to the #6 spot on the leaderboard.
Google is making these experimental models available to developers on its Google AI Studio platform. The company says it continuously works to improve the Gemini models and get the latest updates into users' hands as quickly as possible.
Why does it matter?
While OpenAI is making everyone wait for its new model, Google has constantly updated and added new features to its AI ecosystem. With these significantly improved models in math, coding, and complex prompts, Google is in an advantageous position.
OpenAI to launch ‘Project Strawberry’ this fall
OpenAI plans to launch its advanced AI project 'Strawberry' later this year, possibly in the fall and as an upgrade with ChatGPT. This new AI model is said to have superior reasoning capabilities compared to current chatbots.
It could also perform tasks that today's AI systems struggle with, like designing market strategies and solving intricate word puzzles. The Strawberry model is being developed under a secretive project called 'Q*.' The company has also demonstrated the technology to US national security officials.
Why does it matter?
Strawberry could establish a new benchmark in AI reasoning capabilities, pushing OpenAI towards Stage 2 in its five-stage roadmap to AGI. For a while now, ChatGPT has been getting simple math problems wrong. This could be an improvement Project Strawberry aims to make.
New AI model simulates DOOM video game in real-time
Google and Tel Aviv University researchers have developed a new AI model called GameNGen that can interactively simulate the classic 1993 game DOOM in real time. GameNGen uses a modified version of the Stable Diffusion image synthesis model to generate new frames of the DOOM gameplay at over 20 frames per second on a single TPU.
This allows GameNGen to function as a limited game engine, where the AI can "imagine" or "hallucinate" the graphics in real time instead of using traditional rendering techniques. In tests, human raters had difficulty distinguishing between short clips of actual DOOM gameplay and the outputs generated by GameNGen.
Why does it matter?
GameNGen could be a pioneering AI model as it simulates an actual video game in real time without requiring a game engine. So, we’re at the cusp of a fascinating innovation: In the next few years, AI can create any video game on the fly with personalization for each user.
Qwen2-VL beats GPT-4o; analyzes 20-min long video
Qwen2-VL is a powerful new vision-language AI model released by Alibaba. It has impressive capabilities and outperforms GPT-4o across several benchmarks, particularly in document comprehension and multilingual text-image understanding.
Qwen2-VL can analyze images in various resolutions and even understand videos over 20 minutes long. It excels at complex tasks like college-level problem-solving, math reasoning, and document analysis. The model also supports multiple languages, including most European, Japanese, Korean, Arabic, and Vietnamese.
Why does it matter?
With the release of Qwen2-VL, a brand new contender is looking for the throne in the state-of-the-art AI model race. The model’s ability to comprehend diverse visual inputs and multilingual requests could result in sophisticated and globally accessible AI applications.
1X’s new robot is strikingly human-like
1X Technologies has revealed a new humanoid robot called NEO Beta. It is designed for home assistance and has human-like capabilities in movement, interaction, and task performance. The robot stands 5'5" tall and weighs 66 pounds. It can walk 2.5 miles/hour and run 7.5 miles/hour. It has a carry capacity of 44 pounds and can operate for 2-4 hrs on a single charge.
What's most impressive about NEO is how human-like it is. The robot uses "embodied AI" to understand its environment and learn from interactions. It can perform a wide range of tasks, from household chores to providing companionship. NEO communicates through gestures and body language rather than speech, making interactions natural and intuitive.
Why does it matter?
The race to create the most affordable humanoid robot is underway. 1X Technologies has entered this competition, joining the ranks of China's AGIBOT and Tesla's Optimus. 1X's robot, NEO, is so strikingly human that it has sparked debate about whether it's an actual person.
Enjoying the latest AI updates?
Refer your pals to subscribe to our newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Knowledge Nugget: Top AI tools people actually use
In this post,
discusses today's top AI tools. It shares data from Andreessen Horowitz's report on the top 100 generative AI consumer apps, looking at the most popular web and mobile apps.The data shows that creative tools like image, music, speech, video, and editing comprise 52% of the top generative AI apps. This reflects a growing consumer demand for accessible AI-powered creativity tools. The article also notes that while image generation tools are still popular, there is a shift towards more dynamic content like video and music generation.
Why does it matter?
Generative AI has moved beyond the initial novelty phase and is integrated into everyday productivity and creativity. So, consumers will gravitate towards AI tools that empower them to create high-quality, engaging content more efficiently while offering customization options.
What Else Is Happening❗
💻Lenovo is preparing to launch new, more affordable Copilot Plus PCs, including models powered by an unannounced 8-core Qualcomm Snapdragon X Plus chip.
🤖Amazon's upgraded "Remarkable Alexa" with new generative AI features is expected to launch in October. Anthropic's Claude AI models will power this upgrade.
🖌️Hobbyists have discovered a way to insert custom fonts into AI-generated images. It allows them to create images with specific typefaces, like chalkboard menus or business cards.
🛠️Anthropic has made its "Artifacts" feature generally available for all Claude users. Users can create and run code, visualizations, and interactive apps within the Claude chatbot interface.
🛰️Midjourney has announced that it is "getting into hardware" and has started a new hardware team based in San Francisco. It explores new form factors, like a potential "orb" device.
🔮Gemini is getting new features, like an advanced image generation model called Imagen 3 and the ability to create customized "Gems"-personalized chatbots for specific tasks.
🌐Google has integrated the Gemini chatbot into the Chrome browser's address bar. Users can access Gemini by typing "@gemini" to ask questions without opening the app or website.
🛡️OpenAI and Anthropic will grant US government pre-release access to new AI models. This will allow safety testing and feedback from the government's AI safety institute.
🚀Llama models have over 350 million downloads, a 10x yearly growth rate. Llama usage across cloud providers is 2x in 3 months, making it the leading open-source AI model family.
📈ChatGPT now has 200 million weekly active users, doubling from 100 million users a year ago. The continued growth is attributed to ongoing improvements and new features.
New to the newsletter?
The AI Edge keeps engineering leaders & AI enthusiasts like you on the cutting edge of AI. From machine learning to ChatGPT to generative AI and large language models, we break down the latest AI developments and how you can apply them in your work.
Thanks for reading, and see you next week! 😊
Thank you for the shout-out, Hiren! Great newsletter with a lot of helpful information. I was intrigued by the mention of the hobbyist who found a way to insert custom fonts in FLUX images!