Data/Label-Efficient ML (Active Learning), Transparent ML, Deep Metric Learning, Robust ML Theory and Practice, Semi-supervised ML, Noisy/Missing labels.
Here is a great scientific publication to impede AGI hype: https://t.co/Dut8c7JMMk. On MMMU where advanced perception and reasoning with domain-specific knowledge is needed, the powerful GPT-4V and Gemini Ultra only achieve accuracies of 56% and 59%, respectively.
Therefore, I expect it will take several years or a few decades of multinational AI/ML/DL experts/specialists to push the boundaries of AI in terms of reliability/robustness, accuracy, transparency/interpretability, latency, and throughput, etc
@SharonYixuanLi @uwcdis @WisconsinCS@UWMadisonLS This relates to robust learning and data quality.
In large-scale learning, no data is perfect in terms of coverage, diverse sampling, label noise, etc.
Therefore, it is super important to understand and guide what to learn:
* https://t.co/JrJD1EIY5C
* https://t.co/q6ZYUPrTAB
I just posted "[R] IMAE ICLR2023 RTML: loss function understanding and design for the purpose of robust and reliable ML" on Reddit
https://t.co/WIZYjTKLTo
I cannot agree more! "It is not about any one person's efforts, it is a collective endeavour for a collective purpose!"
In my past 2.5-year startup experience, I have been understanding deeply and comprehensively the importance of stepping out of our com…https://t.co/or4zt10Brr
Registration discount can be offered, if you can attend and share a related presentation. Please feel free to DM if you are interested.
A great chance for networking and show-and-tell sharing.
Join us at International Summit on Biotechnology & Bioinform…https://t.co/hKMfmZtvMF
For the scientific questions, ChatGPT is great, but still needs improving.
I have impression that the large language models are replacing the large text databases, e.g., AlphaFold2 are replacing the PDB database, ChatGPT can repla…https://t.co/c92VyAkmX6 https://t.co/EwKiSoZMPZ
When talking about artificial/machine intelligence/learning, it is always great to revisit and make an analogy about how the human learn statistically or probabilistically:
1. A person becomes over-confident after being exposed to the same information/c…https://t.co/7F1HeCFtOG
Not All Knowledge Is Created Equal: Mutual Distillation of Confident Knowledge (arXiv: https://t.co/ir6k0rde1k) (OpenReview: https://t.co/Nml8xHLMr7)
I am excited to share that this work on Mutual Selective Knowledge Distillation for Robust ML/DL, mainly…https://t.co/VT27aJvr9Y