AI News 18 – 23 March 2024

Posted by:

|

On:

|

Welcome to Endangered AI Weekly 📰, your beacon in the fast-evolving world of artificial intelligence. This week, we’re diving deep into Nvidia’s groundbreaking announcements, Stability AI’s new model and financial woes, Microsoft’s AI PC’s and the beginning of a new future where robots become commonplace.

This Week’s Headlines

  • Key Takeaways from Nvidia’s AI Event
  • Stability Ai’s New Model
  • Stability Ai’s Staff & financial troubles + possible savior?
  • Microsoft launches the worlds’ first AI Pc’s
  • X Open sources Grok
  • New Advancements from OpenAi’s Figure01 robot
  • New AI, Devin could replace programmers!
  • Google’s new AI could replace…Gamers?!

Nvidia’s GTC Event 2024

At Nvidia’s GTC 2024 AI event a slew of groundbreaking, and potentially world changing announcements were made

  • The biggest announcement is around their new Blackwell line of chips. These are not only built on a different die size from it’s predecessor the Hopper chips, but advancements in Nvidia’s NVLink technology means they can put two chips together to act as one. This means a single Blackwell unit can have as much as a 30x performance gain, when compared to it’s processor. What this translates to, is massive gains in not only computational power these chips have, but also how much information they can process at once. This could result in massive leaps in the capacity for understanding in AI models, particularly in Large Language Models, which can be very resource intensive. This new hardware advancement could make more advanced training methods feasable, as the time and information it takes to train models could be significantly reduced, allowing for faster iteration.
  • Nvidia also showed off advancements in it’s Omniverse suite of tools, showing how AI can be used to take 3D models of products such as cars, and shoes, and use prompts to create advertising campaigns, or apply different prints to iterate thru product design. They also showed how the Omniverse can be used to create environments that can be used to train Robots running on the also announced GROOT Model
  • Nvidia announced it’s new GROOT Model, an AI model designed specifically for robots, that can take in multi-modal inputs, meaning it can take language, visuals and sound, as well as information on previous actions, to fine tune it’s understanding of what it needs to do next. This is groundbreaking as it means the model is able to learn on the fly, rather than require continuous re-training. Furthermore, when combined with Nvidia’s Omniverse, the model can be placed in virtual robots, simulating a real life environment to allow the model to learn to be a robot in it’s destined environment, before being downloaded to the actual robot.
  • NIM’s – Also announced is a format to pre-packaged AI tools and software in Containers that are able to store the versions and dependencies for that software simplifying the distribution of AI tools and software.
  • Gaming is about to change – Nvidia also presented It’s ACE technology for creating NPC’s that can have unique dialogue in games powered by LLM’s. Showing off a toolkit to allow developers to also make NPC’s that no longer run off a pre-determined script, but rather guidelines on their character and personality. This could bring an entirely new layer to story driven video game worlds, where the player can engage, and interact with any background NPC. The best part about this, is that while Nvidia has been teasing us with this technology for years, they are finally releasing a playable demo later this year!

3D from any Image

  • Stability AI announced their newest AI Model Stability Video 3D. The model can take a 2D Image and turn it into a 3D model! Stay tuned for a deep dive video on that soon! But if the quality is any good, this could be a game changer not only for game development, but even in allowing us to bring 3D models into VR

Stability AI Financial & Team Woes + Possible Savior

  • Stability AI has been struggling as of late, with Key members resigning from their posts, and even it’s CEO Emad. A rumored reason for this could be disagreements with other shareholders seeking to make a profit due to Stability AI struggling to stay in the black. The key point of contention is suspected to be the team’s desire to release Stable Diffusion 3 as Open source, with shareholders blocking it, preferring instead to keep it closed to profit off it. Despite remaining a major shareholder, Stability AI now lacking a CEO, so close to the release of Stable Diffusion 3. However there could be a light at the end of the tunnel, as HuggingFace’s CEO tweeted an interest in buying Stability AI with the intention to release Stable Diffusion 3 as Open Source. While the loss of a major player in the AI space could be painful, considering most of the key researchers have left, such an acquisition could be a boon for Open AI and the Open source community as a whole.

Microsoft Launches the worlds first AI PC

  • Microsoft’s debut of the Surface Pro 10 and Surface Laptop 6 marks a significant leap in the computing landscape: the introduction of the world’s first AI-integrated PCs.

Key Innovations:

  • AI-Enabled Computing: The Surface Pro 10 and Surface Laptop 6 are pioneering the future with Intel’s Core Ultra CPUs, bringing advanced AI capabilities to the forefront of professional computing.
  • The Copilot Key: This innovative feature is a testament to Microsoft’s commitment to AI-driven productivity, providing immediate access to AI-powered tools and assistants directly from the keyboard.
  • Exclusively for Professionals: Aimed at the business sector, these latest additions to the Surface lineup are available only through Microsoft’s website or selected commercial resellers, highlighting a strategic focus on enhancing the professional computing experience.

Why This Launch Is a Game-Changer: The introduction of the Surface Pro 10 and Surface Laptop 6 by Microsoft isn’t merely about new gadgets; it’s a defining moment in the evolution of PCs. By embedding AI directly into these devices, exemplified by the groundbreaking Copilot Key, Microsoft is redefining user experience, tailoring technology to meet the sophisticated needs of professionals, and establishing a new benchmark for future computing devices.

Devin Unveiled: Revolutionizing Software Development with the World’s First AI Engineer

In a move that has the software development world buzzing, tech startup Cognition introduces Devin, the world’s first AI software engineer. This groundbreaking tool promises to revolutionize how coding, website creation, and software development are done. Devin stands out for its ability to handle a wide array of software engineering tasks with just a single prompt, from writing and debugging code to deploying fully functional websites and software.

However, Devin isn’t designed to replace human engineers but rather to augment their capabilities, making their work more efficient and less tedious. Its creators emphasize that Devin aims to work alongside human counterparts, enhancing productivity and allowing for more focus on creative and complex problem-solving. Impressively, Devin has already proven its mettle by passing engineering interviews at leading AI companies and executing real jobs on platforms like Upwork, showcasing its practicality in real-world applications.

Grok Unleashed: xAI’s Strategic Move to Open Source AI’s Future

Elon Musk’s xAI has taken a significant step in the AI landscape by open-sourcing the base model of Grok, a 314 billion parameter Mixture-of-Experts model. However, the release intriguingly omits any training code. Grok-1, designed without a specific application focus, such as conversational AI, operates under the Apache License 2.0, allowing for commercial applications. This move opens up the model for exploration and development by the wider tech community, although it notably lacks integration with real-time data from the X social network, where Grok initially debuted as a chat bot for Premium+ users.

The release has sparked interest among AI tool developers, with discussions about integrating Grok into various applications, such as Perplexity’s plans for conversational search. This decision aligns with Musk’s ongoing advocacy for open-source AI, amidst legal battles with OpenAI over the deviation from open-source principles. The open-sourcing of Grok not only challenges the status quo established by giants like OpenAI, Meta, and Google but also reignites the debate on the openness of AI development and its implications for innovation and security in the field.

By sharing Grok’s architecture and weights without the training data or social network integration, X-AI aims to foster collaboration and innovation. Yet, this move also raises questions about the balance between openness and the need to safeguard against potential misuse of AI technology. As Grok becomes available for commercial use and further development, the tech community eagerly anticipates its impact on the future of AI applications and the ongoing discourse around open-source models versus closed-source counterparts like Meta’s Llama 2.

Game On: Google’s SIMA AI Could Be Your Next Video Game Teammate

Google DeepMind’s latest innovation, SIMA (Scalable, Instructable, Multiworld Agent), is poised to redefine the gaming experience by introducing an AI that plays alongside humans, not to dominate but to collaborate. Unlike traditional game AIs focused on winning, SIMA’s design emphasizes adaptability and cooperation, learning to navigate and interact within a variety of video game environments from open-world adventures like “No Man’s Sky” to the chaotic fun of “Goat Simulator.” This advanced AI agent has been trained using games that emphasize open-ended play, allowing it to understand and execute tasks based on natural language instructions, making it a versatile gaming companion that enhances the social and interactive aspects of gaming.

Developed in partnership with several game studios, SIMA showcases an ability to understand and perform a wide range of actions across different gaming environments without needing game-specific programming or APIs. This approach not only broadens SIMA’s applicability across various game genres but also exemplifies a significant step towards creating AI that can understand and interact within complex, unstructured spaces akin to the real world. With approximately 600 basic skills already under its belt, SIMA’s potential extends beyond gaming, hinting at future applications where AI agents assist humans in more dynamic and creative tasks.

SIMA represents more than just an advancement in gaming technology; it’s a glimpse into the future where AI agents play key roles in both virtual and real-world settings. By mastering the ability to follow instructions and learn from interactions in video game environments, SIMA could soon become a foundational tool for developing AI that collaborates with humans across a spectrum of activities. Google’s initiative not only paves the way for more engaging and immersive gaming experiences but also demonstrates the potential for AI to assist and enhance human abilities in broader, more complex scenarios.

Want to stay up to date with the latest AI news? Sign up for our Weekly News Newsletter!