Hong Kong Information Technology Industry Council (HKITIC) of the Federation of Hong Kong Industries (FHKI) (FHKI Group 25) conducted AI & Big Data Workshop on 22nd Jun 2018. It provided another extraordinary opportunity to explore more about their applications in commercial world. Per FHKI invitation, I attended the workshop. I met my EngD Cohort Dr. Sunny Chai (Vice Chairman, FHKI) and we took a photo for memory.
I recognized speaker Mr. Samson Tai and his colleagues in IBM.
(Left: I, Mr. Kin Tsang (Executive, SaaS, Cognitive Solution, IBM Greater China Group) and Mr. Samson Tai (Distinguished Engineer & CTO IBM Hong Kong))
In the beginning, Dr. Sunny Chai (Vice Chairman, FHKI) gave a welcome speech. He said FHKI had successfully applied Professional Services Advancement Support Scheme (“PASS”) for FHKI International Forum on promoting AI, Big Data and Cloud Computing. Now, we were honor to invite IBM CTO to share the latest tools and advanced technologies to shape our innovative and data-driven city.
Then Dr. Sunny Chai and Mr. Chris Huang (Group 25 Vice Chairman, HKITIC) presented a souvenir to the guest speaker Mr. Samson Tai.
Mr. Samson Tai (Distinguished Engineer & CTO IBM Hong Kong) presented title named “AI Training – Putting AI to work today!” His topics included deep learning, neural networks, deep learning frameworks and common use cases, as well as, future of AI.
He briefed the history of deep learning and it had a breakthrough in 2012 because of Graphics Processing Unit (GPU) where deep learning was one of disciplines of machine learning. Deep Learning is equal to Training Artificial Neural Networks.
The following diagram showed deep learning set of functions for Speech Recognition, Image Recognition, Playing Go and Dialogue System. Then Mr. Tai briefed its application in various industries such as Healthcare Biology (e.g. Cancer Detection), Consumer Web Mobile, Retail (e.g. Image Classification), Media Entertainment (e.g. Video Captioning), Autonomous Cars Transformation (e.g. Vehicle detection) and Security Public Safety (e.g. Face Recognition).
And then Mr. Samson Tai mentioned deep learning process and supervised learning. He said data without labeling had no meaning. Therefore, supervised learning was important. After that Artificial Neuron was introduced.
Mr. Tai introduced 3 paths to AI System and they were:
i) Pre-trained model (e.g. Per-trained visual recognition AI models)
ii) Transfer learningiii) Customer models (e.g. create customer AI models by data scientist)
The Transfer Learning is used Pre-Trained Model with customer’s domain data to build transfer learning model (modified the existing model) for new application. We could develop own AI model using the common deep learning framework available in the market.
Mr. Samson Tai then briefed the Convolution Neural Network model coded in Keras. One of example was finding information of your favorite brand of Whisky. He also briefed the improving Enerprise AI life cycle in Open Source to us.
There were four Deep Learning Networks as follows:
- Feed Forward Neural Network (FNN) (for structure data calculation)
- Convolution Neural Network (CNN) (for image)
- Recurrent Neural Network (RNN) (for language with time series)
- Long Short-Term Memory Network (LSTM) (for context (前文後理))Finally, he introduced the Debating Technologies “Project Debater”.
During Q&A session, Mr. Samson Tai demonstrated the Watson Studio. He also showed some free learning paths:
He used Dr. Sunny Chai photo as example to demonstrate the visual recognition tool to identify the characteristic of the photo. (It also detect the age as well!)
At the end, all FHKI Group 25 committee members took a group photo for memory.
20180329: FHKI International Forum: Cloud Computing, Big Data and AI in Realizing an Intelligent Society - https://qualityalchemist.blogspot.com/2018/03/fhki-international-forum-cloud.html20180213: FHKI Seminar on Bay Area Soar by Innovation and Industrial Upgrading Opportunities (大灣創新騰飛, 工業升級良機) - https://qualityalchemist.blogspot.com/2018/02/fhki-seminar-on-bay-area-soar-by.html