Date: Wedn May 15, 2024 6:30 - 8:30 PM , at ITRI 2880 Zanker Suite 103. Seminar talk will start at 7PM.
Eventbrite registration link:
Abstract: This seminar will cover some of the most recent developments in AI including: 1.) OpenAI’s Q*(Q star) project, 2.) a new more efficient Mamba architecture, 3.) Anthropic’s safe Constitutional AI, and 4.) most recent developments in humanoid robots.
1. Q* is a current initiative by OpenAI that excels in solving basic math and showcasing reasoning abilities beyond current AI models leading to advancements in general artificial intelligence (AGI). Q* mixes big language systems (Q learning) with other smart methods such as reinforcement learning and search algorithms (A* search).
2. Current Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module.
A new selective state space model (SSM) called Mamba has been developed to overcome some of the limitations of Transformers. They address Transformers' computational inefficiency on long sequences. Mamba uses a selective scan mechanism that allows its model to process sequences in a way that is both efficient (due to the parallelism) and flexible (due to the selectivity).
3. Constitutional AI developed by Anthropic consist of a set of tools and techniques that ensure AI closely aligns with human values, making sure it's helpful, harmless, and avoids deceptive practices. Anthropic created a Constitution for large language models which ensures that Claude (Anthropic’s LLM) adhere to human values.
4. Aside from the 3 new AI techniques mentioned above, the seminar will also cover most recent humanoid robots advances from Tesla (Optimus Gen 2), OpenAI (Figure 01), Sanctuary AI (Phoenix) , Apptronik’s Apollo, Unitree H1 (China), and others, as well as Nvidia’s new robotic platform called GR00T (Generalist Robot 00 Technology) .
Come and join us for this informative technical talk on recent AI advances.
Speaker Bio:
Dr. William Kao received his BSEE, MSEE and PhD from the University of Illinois Urbana-Champaign. He worked in the Semiconductor and Electronic Design Automation industries for more than 30 years holding several senior and executive engineering management positions at Texas Instruments, Xerox Corporation, and Cadence Design Systems.
Dr. Kao has authored more than 40 technical papers on IC CAD at IEEE Journals and Conferences, and holds eight US software and IC design patents. He was an Adjunct Professor at UCLA Electrical Engineering Department where he taught courses in computer aided IC design.
Dr. Kao is a Senior Member of IEEE, and was one of the founding members of IEEE-Circuits and Systems - Silicon Valley Chapter.
Dr. Kao teaches Clean Technology and Emerging Technology courses at the University of California Santa Cruz, Silicon Valley Extension.
He has given more than 100 seminars on various Emerging Technologies including Clean Technology, Renewable Energy, Big Data and Data Analytics, IoT, Smart Cities, Augmented and Virtual Reality, Robotics, Drones, AI / Machine Learning, 5G, 3D Printing, the Metaverse and Satellite Communications.