Explore the transformative potential of multimodal generative AI in this insightful presentation by Steven Hoi at SuperAI 2024. As a leading expert and a driving force behind Singapore-based company Hyper GII, Hoi dives into the latest advancements in multimodal foundation models and their far-reaching applications across industries.

​

Hoi's talk underscores the tremendous strides made in training large-scale models that can process a variety of inputs, such as text, images, video, and audio. He highlights the development of multimodal large language models (LMs) for understanding and multimodal diffusion models for generating creative content. Notably, models like GPT-4V and emerging open-source counterparts, such as D-Tree and Stable Diffusion, exemplify the burgeoning capabilities within this space.

​

Hyper GII introduces cutting-edge solutions through proprietary technologies like the High Pre-train Transformer (HPT) framework, designed to close the modality gap and enhance model scalability. Hoi showcases achievements in personalizable content generation and presents pioneering models like HPD 1.5H, optimized for mobile and edge devices, marking a new era in AI accessibility and application.

​

Hoi also discusses the far-reaching implications of multimodal generative technology, from personalized marketing and intelligent medical diagnostics to innovative e-commerce and autonomous AI agents. His vision anticipates a future where digital and physical spaces are redefined by highly adaptable AI systems.

​

Join the movement shaping tomorrow's AI landscape. Stay informed and inspired—like, comment, and subscribe for more updates and insights on the forefront of AI advancements.