Apple's MM1 Model Signals a New Era in AI Assistants


Jim Miller

Apple Quietly Advances in AI with New MM1 Model Amidst Industry Buzz

Welcome back to AI Hungry, where the latest developments in artificial intelligence are always on the menu. In this update, we delve into Microsoft's strategic move to hire DeepMind co-founder Mustafa Suleyman, stirring the pot of industry concerns. Meanwhile, Apple's AI advancements with their new MM1 model hint at a future where Siri might be outshone by its own kin.

From executive shake-ups to technological breakthroughs, these stories not only reflect the dynamic nature of the AI landscape but also raise questions about the future of AI governance and innovation. Continue reading to uncover the implications of these pivotal moments in AI history.


Microsoft Hires DeepMind Co-Founder to Lead New AI Division Amid Industry Concerns

Mustafa Suleyman, a co-founder of the AI research company DeepMind, has been appointed as the CEO of a new division at Microsoft, named Microsoft AI. Along with Suleyman, his Inflection co-founder KarΓ©n Simonyan and several team members are joining Microsoft to focus on consumer AI products and research, including Copilot, Bing, and Edge. Microsoft CEO Satya Nadella expressed his admiration for Suleyman's visionary work and his history of founding innovative teams.

Despite the positive announcement, there are concerns in the tech community about Suleyman's past allegations of bullying at Google DeepMind and the potential for increased consolidation of power in the AI industry. These concerns are amplified by Microsoft's significant investments in AI, including its partnership with OpenAI, and the recent hiring of prominent figures in the AI field.

Main course

Apple Quietly Advances in AI with New MM1 Model Amidst Industry Buzz

Apple, traditionally reserved in its AI ventures, has made a significant leap with the development of a new generative AI model named MM1, capable of processing both text and images. This model, detailed in a research paper by Apple engineers, demonstrates abilities similar to ChatGPT, including answering questions about photos and displaying extensive general knowledge. MM1 is a multimodal large language model (MLLM), which sets it apart by its capacity to understand and interact with visual data as well as text. The model's smaller size in terms of parameters suggests Apple's strategy might be to refine its AI technology before scaling up.

While Apple has been in talks with Google about possibly integrating Google's Gemini AI model into iPhones, the emergence of MM1 indicates that Apple is also developing its own generative AI tools. This could lead to a new generation of AI assistants, enhancing or even replacing Siri, which now seems outdated compared to recent advancements by competitors. Apple's control over both software and hardware could give it an edge in the AI race, especially with its commitment to user privacy and on-device processing.


πŸŽ₯ Stability AI Unveils Stable Video 3D for Advanced 3D Video Generation. Stability AI introduces Stable Video 3D, a generative AI tool for 3D video rendering from images or text prompts. It offers commercial and non-commercial access, and is touted as a significant advancement over previous models for creating coherent 3D meshes and multi-view videos. (Link)

πŸš€ Nvidia Unveils Blackwell GPU to Revolutionize Generative AI Across Industries. Nvidia introduces the Blackwell GPU, promising massive performance and efficiency gains for generative AI. The platform will be used by major tech firms and is named after mathematician David Harold Blackwell. (Link)

πŸ€– Nvidia Unveils Project GR00T: The Future of Humanoid Robotics. Nvidia introduces Project GR00T, a multimodal AI designed to power humanoid robots with advanced capabilities, including understanding and processing text, speech, and demonstrations to perform tasks in real-world environments. (Link)

πŸš€ Elon Musk's AI Firm Unveils Grok-1 Model Weights to Rival ChatGPT. Elon Musk's company xAI has released the base model weights for Grok-1, a language model with 314 billion parameters, under the Apache 2.0 license. This move challenges OpenAI's less open practices and could enable broader AI development. (Link)

πŸš€ OpenAI's GPT-5 Expected to Launch in Mid-2024. OpenAI is anticipated to release GPT-5, a more advanced AI language model, in mid-2024. The model promises improvements over its predecessor, GPT-4, and is currently undergoing testing and demonstrations to select enterprise customers. (Link)

⚽ TacticAI: Revolutionizing Soccer Strategy with Deep Learning. Researchers have developed TacticAI, a tool that uses player-tracking data and geometric deep learning to optimize soccer strategies. It predicts player movements during corners, offering tactical advice to coaches, and could soon support natural language queries. (Link)

Enjoying this newsletter?

Subscribe to get more content like this delivered to your inbox for free.