Schedule
|
Announcement:
(Updated as of July 17, 2023 | 19:20 HKT) |
• | The workshop will be resumed tomorrow (July 18) following the schedule of Day 2 (July 18) as shown below with registration time from 08:30 and first talk at 09:00. |
|
(Updated as of July 17, 2023 | 07:30 HKT) |
• | As Hong Kong Observatory has issued the No. 8 SE Gale and Storm Signal, sessions of the workshop on July 17 will be cancelled.
Resumption of workshop will depend on the time when the Signal will be lowered. Please stay tuned for the latest updates on our website. Stay safe everyone! |
| 1. | This workshop will be held in face-to-face mode. |
| 2. | Unauthorized recording of any session is prohibited. |
| 3. | Please obtain the consent from the speaker concerned for any adaptation/sharing of his/her presentation materials. |
| Should you have further enquiries, please feel free to contact the secretariat by email at iasmatei@ust.hk . | |
Venue: Kaisa Group Lecture Theater (IAS LT), G/F, Lo Ka Chung Building, Lee Shau Kee Campus, HKUST
July 17, 2023 (Mon)
Time
Event
| 08:30 - 09:00 | Workshop Registration | |
| 09:00 - 09:15 |
Opening Remarks Harry SHUM & Andrew COHEN (HKUST) |
|
|
Session M1 - Theory for Emergent Intelligence Chair: Harry SHUM (HKUST) |
||
| 09:15 - 09:55 |
#1: The TinyStories Dataset: How Small Can Language Models Be And Still Speak Coherent English? [Abstract] Ronen ELDAN (Microsoft Research) |
|
| 09:55 - 10:35 |
#2: The Past, Present, and Future of Artificial Intelligence: From Black-box to White-box Deep Networks [Abstract] Yi MA (University of California, Berkeley; The University of Hong Kong) |
|
| 10:35 - 11:05 | Coffee Break (Venue: IAS Lobby, G/F) | |
|
Session M2 - Theory for Emergent Intelligence Chair: Harry SHUM (HKUST) |
||
| 11:05 - 11:45 |
#3: Feature Geometry and Applications in Deep Learning [Abstract] Lizhong ZHENG (Massachusetts Institute of Technology) |
|
| 11:45 - 12:30 |
Panel Discussion Chair and speakers from #1 to #3 |
|
| 12:30 - 14:00 | Workshop Lunch (Venue: UniQue, Conference Lodge) | |
|
Session M3 - Data and Compute for DL/ LLMs Chair: Andrew COHEN (HKUST) |
||
| 14:00 - 14:40 |
#4: On the Relationship between Generative and Discriminative Models [Abstract] Zhuowen TU (University of California, San Diego) |
|
| 14:40 - 15:30 |
#5: Technology and Large-scale Industrial Application of Task-based Intelligent Dialogue [Abstract] Xiaodong HE (JD.COM) |
|
| 15:30 - 16:00 | Coffee Break (Venue: IAS Lobby, G/F) | |
|
Session M4 - Data and Compute for DL/ LLMs Chair: Andrew COHEN (HKUST) |
||
| 16:00 - 16:40 |
#6 Compositional AI - Advanced Abilities Emerge When All Necessary Basic Abilities Are Strong [Abstract] Nan DUAN (Microsoft Research Asia) |
|
| 16:40 - 17:25 |
Panel Discussion Chair and speakers from #4 to #6 |
|
July 18, 2023 (Tue)
Time
Event
| 08:30 - 09:00 | Workshop Registration | |
|
Session Tu1 - Math for AI and AI for Math Chair: Yi MA (University of California, Berkeley; The University of Hong Kong) |
||
| 09:00 - 09:35 |
#07: Transformer Models of Language and Logic [Abstract] Michael DOUGLAS (Harvard University) |
|
| 09:35 - 10:10 |
#08: Deep Networks for Low-Dimensional Structure: Mathematical Guarantees and Network Constructions [Abstract] John WRIGHT (Columbia University) |
|
| 10:10 - 10:45 |
#09: The Unreasonable Effectiveness of Mathematics in Large Scale Deep Learning [Abstract] Greg YANG (xAI) |
|
| 10:45 - 11:15 | Coffee Break & Group Photo Taking (Venue: IAS Lobby, G/F) | |
|
Session Tu2 - Math for AI and AI for Math Chair: Yi MA (University of California, Berkeley; The University of Hong Kong) |
||
| 11:15 - 11:50 |
#10: Certifiably Robust Learning via Knowledge-Enabled Logical Reasoning [Abstract] Bo LI (University of Illinois at Urbana-Champaign) |
|
| 11:50 - 12:30 |
Panel Discussion Chair and speakers from #7 to #10 |
|
| 12:30 - 14:00 | Workshop Lunch (Venue: UniQue, Conference Lodge) | |
|
Session Tu3 - Theory and Practice for DL/ LLMs Chair: Qifeng CHEN (HKUST) |
||
| 14:00 - 14:40 |
#11: On the Adaptivity of Deep Neural Networks [Abstract] Yu-Xiang WANG (University of California, Santa Barbara) |
|
| 14:40 - 15:20 |
#12: A “Physical” Law of Data Separation in Deep Learning, with Some Ramblings [Abstract] Weijie SU (University of Pennsylvania) |
|
| 15:30 - 16:00 | Coffee Break (Venue: IAS Lobby, G/F) | |
|
Session Tu4 - Theory and Practice for DL/ LLMs Chair: Qifeng CHEN (HKUST) |
||
| 16:00 - 16:40 |
#13: Transforming Large Language Models into Interactive Agents in the 3D Visual World [Abstract] Liwei WANG (The Chinese University of Hong Kong) |
|
| 16:40 - 17:25 |
Panel Discussion Chair and speakers from #11 to #13 |
|
| 17:25 - 17:30 |
Closing Remarks by Organizers |
|
| 17:30 - 18:00 | Short Break | |
| 18:00 | Departure from HKUST IAS to Sai Kung for Dinner (For Organizers/Chairs/VIPs/Speakers only) | |
| 20:30 | Departure from Sai Kung and return to HKUST | |