OpenAI Leadership Shifts, LLaMA 3.2, and New Arabic Models in the Spotlight

Welcome to the last edition of September 2024's AI Weekly Update. It’s been an eventful month in the world of artificial intelligence, with significant changes in leadership, product launches, and regulatory discussions shaping the future of the industry. Here are some of the top stories you need to know:

OpenAI’s Leadership and Restructuring

OpenAI is undergoing a major organizational shift. The company announced its plan to restructure its core business into a for-profit benefit corporation, separating from its nonprofit board. The restructuring could potentially value OpenAI at around $150 billion, with CEO Sam Altman reportedly receiving up to 7% equity in the new entity. These changes come as the company faces a leadership shakeup. Three key executives—CTO Mira Murati, Chief Research Officer Bob McGrew, and Vice President of Research Barret Zoph—have announced their departure. Their exits come amidst this strategic transformation, with new leadership appointments expected soon. Altman acknowledged the departures and emphasized that the company remains focused on building AI that benefits society at large

LLaMA 3.2 Release: A Multimodal Leap

Meta has released version 3.2 of its open-source LLaMA model. This version stands out as it’s now multimodal, meaning it can process both text and images, which significantly enhances its capabilities in creative and analytical applications. LLaMA 3.2 is available on Hugging Face, where users can explore its features and test out its capabilities for free. This development furthers the trend of open-source AI models becoming more accessible to the public and institutions alike

Google Gemini Models Get Major Updates

Google’s Gemini suite has received significant updates, making its models even more competitive in the AI landscape. Notable changes include a reduction in the pricing for Gemini Pro models, along with an increase in rate limits, making it easier for developers to experiment with and integrate these models into their applications. With these updates, Google aims to offer a more affordable and accessible alternative in the competitive AI model space, targeting a broad range of use cases from chatbots to content generation(

Meta's Ray-Ban Metaglasses and AI Clone Feature

Meta continues to innovate in the AR/VR space with the release of new features for its Ray-Ban Meta glasses. Among the most exciting is the AI clone feature, which allows users to create personalized AI avatars for enhanced interaction. These glasses are a part of Meta's broader vision to integrate AI seamlessly into daily life, merging physical and digital experiences

Middle East Spotlight: Advancements in Arabic NLP

A groundbreaking development has emerged from the MBZUAI France Lab, the Paris-based research arm of Mohamed bin Zayed University of Artificial Intelligence (MBZUAI). The lab has introduced two new large language models (LLMs) specifically trained for Moroccan Arabic (Darija), named Atlas-Chat-2B and Atlas-Chat-9B. These models were trained as part of G42 Group’s Jais LLM project and focus on enhancing Natural Language Processing (NLP) tasks in Darija. The new models, built on Google's Gemini 2 platform, reportedly outperform previous Arabic-focused models, including Jais 13B, improving Darija-specific NLP tasks by 13%. This achievement represents a significant step forward in AI for regional dialects and is expected to contribute to the broader Arabic AI ecosystem

Other AI Developments: Industry Moves Toward Accountability

In regulatory news, a new bill has been introduced that would hold model developers responsible for any catastrophic harm caused by AI models, even if they were not directly involved in the incident. This could have a profound impact on the accountability frameworks for AI developers and companies. Additionally, Snapchat has announced it will now use Google’s Gemini models to power its chatbot, and AMD has unveiled a small yet powerful language model, broadening the landscape of AI solutions

I remember when I first started in the world of AI and Data—it wasn’t as glamorous or straightforward as it may seem today. Like many, I struggled to find the right guidance, faced countless roadblocks, and questioned if I was even on the right path. But after years of working with some of the biggest companies in the world, I realized something: success in AI isn’t just about knowledge, it’s about real-world experience, having the right mentors, and building projects that truly matter.

That’s why I created the AI Residency Program—to offer aspiring AI professionals and entrepreneurs the support and real-world experience I wish I had when I started. This program gives you access to hands-on projects, industry experts, and personalized mentorship to help you succeed, whether you’re aiming for a career in AI or looking to build your own AI product. The new cohort is starting soon, Apply Now

Closing Thoughts

The AI industry is moving at an unprecedented pace, with advancements in technology and significant organizational changes reshaping the landscape. As we close out September, it's clear that the next few months will be critical in determining how AI companies, both for-profit and open-source, navigate the challenges and opportunities ahead.

Stay tuned for more updates, and feel free to explore some of these new models and technologies!

Thank you for subscribing to our AI Weekly Update. See you in October!