Home AI News Advancing AI: Multimodal, Safe, and General – NeurIPS 2023 Highlights

Advancing AI: Multimodal, Safe, and General – NeurIPS 2023 Highlights

0
Advancing AI: Multimodal, Safe, and General – NeurIPS 2023 Highlights

37th Annual Conference on Neural Information Processing Systems 2023

The 37th annual conference on Neural Information Processing Systems (NeurIPS) is set to kick off next week in New Orleans, USA. This conference is the largest artificial intelligence (AI) conference in the world. Google DeepMind teams will be presenting over 180 papers at the main conference and workshops, showcasing demos of cutting-edge AI models for global weather forecasting, materials discovery, and watermarking AI-generated content. One of the main highlights will be Gemini, the largest and most capable AI model developed by the team.

Multimodality: Language, Video, Action

Google DeepMind is focusing on creating more multimodal AI systems that can interact across different platforms. They are exploring how generative abilities can help models to learn across various mediums. One of the main highlights is UniSim, a universal simulator of real-world interactions. This type of technology could have applications across various industries, from video games and films to training agents for the real world.

Building Safe and Understandable AI

The research also focuses on building safe and understandable AI. Researchers found that large language models are prone to “hallucinations,” generating text that seems correct but is fabricated. They are also studying how to embed privacy into every step of developing and deploying large models to protect private and sensitive material.

Emergent Abilities

As large models become more capable, Google DeepMind is pushing the limits of new abilities to develop more general AI systems. They are introducing a new framework for language model inference called the Tree of Thoughts, which helps models explore and reason over a wide range of possible solutions. Additionally, Feature Multiplexing is being used to process billions of unique values efficiently, allowing AI models to scale for billions of users.

Fostering a Global AI Community

Google DeepMind is proud to sponsor NeurIPS and support workshops led by LatinX in AI, QueerInAI, and Women In ML, helping foster research collaborations and developing a diverse AI and machine learning community. They will also feature their Visualising AI project, commissioning artists to create more diverse and accessible representations of AI.

If you’re attending NeurIPS, be sure to visit the Google DeepMind booth to learn more about their cutting-edge research and meet the teams hosting workshops and presenting across the conference. To learn more about the future of AI, visit NeurIPS 2023 in New Orleans from December 10-16.

Source link

LEAVE A REPLY

Please enter your comment!
Please enter your name here