On Sunday 21st of May, 2023 MIT Shanghai Club held its second AI Seminar event on AI and Robotics. The venue at 上海万科中心was generously provided by Fred Liu (PhD course 3 ‘19,  board member), which offered a panoramic view of the riverfront. Around 20 participants gathered in person, with 8 others joining remotely via Tencent Meeting.
Session was kicked off by Peter Yu (PhD course 6 ’18, board member), chief organizer, with an overview of robotics and motivation for the seminar. Charles Du (6/15 ’02, MEng Media Lab ’03, board member), then introduced the Meldd(袂得) consensus framework being used for information consolidation across seminars and reviewed key cluster topics from the previous AI and Art seminar, ranging from AI toolsets, strategy, and policy.
Speaker Daolin Ma (post-doc course 2 ’16) gave an in-depth talk on the latest research on tactile sensors, usage of gel finger tips to detect material deformation, creating a feedback model capable of fine action controls like object insertion separately from traditional camera modeling systems. Videos of the pincer test system was shown, and candid assessment of the challenges (like linear objects deforming in non-linearly).
Speaker Hang Zhao (PhD course 6) presented recent trends in autonomous navigation systems. Problems with traditional precision mapping techniques like costly photo/tagging and policy constraints were examined. Vector prediction modeling and no-map navigation were discussed as possible solutions. Challenges in off-road robotic navigation were highlighted as possibly benefiting from the new approaches.
During the panel phase, audience participated in providing industry insider perspectives, Eddie Wang weighed in on the imminence of drone delivery citing the ongoing Shenzhen service offered by Meituan drones, supporting the argument in favor of commercial scaling feasibility. Yuanting Shu (MFin ’17) talked about LiDAR vs visual only navigation. Jie Zhang (MA course 4 ‘15) raised the topic of Neuralink alternatives in China. Peter posed the question of misconceptions and challenges in robotics. Charles then led discussion around topic clusters on Meldd regarding feasibility of level 5 self-drive, medical robotic surgeries, visual prompting, NVidia Omniverse simulation, fembots, and AI robotic host chassis specifications.
Overall, the latest breakthroughs in AI does not equate to a breakthrough in Robotics, but visual prompting and new simulation tools mean potentially major extension of AI through drone delivery and RoboBee swarm systems. Presenters and participants alike came across new ideas to explore, and will reconvene in the future to share progress made.
To see the full perspectives and make contributions of your own insights, download the free 袂得 app and access the MIT AI Seminar node. Vote on clusters with contrasting perspectives to help build the ongoing dynamic consensus!
Download 袂得:
iOS from App Store, or scan:
Android  from Play Store, or scan: (then Open in Browser from “…” in upper right, then Install button)
继续阅读
阅读原文
关键词
of the
on the
of the
AI and
from the