The PolyU seminar named “GenAI: Opportunities
and Challenges” was organized by Faculty of Science, PolyU on 8 Jan 2024. Dr.
Dennis Ong is guest speaker who is a Technology Executive at Amazon Web
Services (AWS) focusing on AI and cloud transformation for Fortune 100
companies.
This seminar aimed to provide a comprehensive overview
of Generative AI (GenAI), fostering a deeper understanding of its capabilities,
challenges and potential for positive transformation while emphasizing the
critical need for ethical and responsible implementation in an ever-evolving technological
landscape.
Firstly, Dr. Dennis Ong introduced AI from expert
system to neural network and deep learning. There were four types of learning included
supervised learning, semi-supervised learning, unsupervised learning and
unsupervised learning in real life.
Then Dr. Ong briefed the Generative AI from word
vector to LLM and adaption, then application such as ChatGPT.
Word vector is important for the response by ChatGPT
based on prompt leading to which cluster. The important paper from Google named
“Attention is all you need” in 2017 explained the transformer.
And then he briefed the how large language model (LLM)
functioned. It predicted the next word based on your input statement.
After that Dr. Ong explained ChatGPT 3.5 through
Reinforcement Learning Human Feedback that used human feedback to optimize ML
models to self-learn more efficiently.
He also mentioned Retrieval Augmented Generation (RAG)
that combined the strengths of both retrieval- and generative-based artificial
intelligence (AI) models. That prompt plus enhanced content from knowledge
sources as input to LLM.
Finally, he briefed different potential applications
of GenAI including marketing, code generation, etc.
However, Generative AI had associated with different
risks including poorly understood, data privacy, hallucination, biased, job
loss and copyrights, etc.
Lastly, Dr. Ong encouraged us to unlock innovation opportunities
from GenAI such as co-pilot working with GenAI. Life-long learning is most
important to adopt AI era he said.
Before end of seminar, he asked us to discuss the
application topics with other participants in the class.
During Q&A, I had asked questions about patent
using OpenAI and the accuracy of Dall-e for visualizing the new idea.
At the end, I took a photo with Dr. Dennis Ong for
memory.
