The recent advancements in artificial intelligence (AI) technology is showing great promise in areas such as voice recognition, language translation, image recognition, etc.
Alex Hung, president of the Hong Kong New Emerging Technology Education Association, discusses the growing popularity of chatbots and its underlying AI technology.
Mickey Fong: After decades of building and investing, AI technology has made a huge leap forward in recent years. It is said that AI can now identify pictures better than humans, but in voice recognition, AI has not reached human parity yet. How difficult is it to develop voice recognition technology?
Alex Hung: It depends on how much time and resources are allocated to train the AI software. One of the methods adopted by Facebook and Google in developing image recognition technology is to spend five to six years training the software to recognize images provided with human assistance.
But voice recognition is more complicated, the huge difference in each person’s voice, coupled with the variation in their voice frequency ranges, makes it harder for the program to learn. For machine translation, as the same phrases or words we use under different occasions may convey entirely different meanings, it would be challenging for the software to literally judge the meaning.
Mickey Fong: We know that you have experience in Natural Language Processing (NLP) machine training; can you tell us more about it?
Alex Hung: In NLP training, texts are divided into separate words and then converted to its real meaning. The level of difficulty varies with different languages in conversion.
For English, the format is fixed, with words usually separated by spaces, so it would be easier for word segmentation. However, for language like Chinese, as it does not have a regular format with word boundaries clearly marked, it would be more difficult and complicated to separate words.
Mickey Fong: Regarding the acquisition of AI startups in the first quarter this year, majority of the acquired firms have their focus on building NLP technology. What explains the market’s favor toward NLP technology?
Alex Hung: Chatbot is becoming an increasingly popular business tool. With the enormous demand in customer service operation worldwide, everyone in the market recognizes the huge potential in chatbots and its underlying NLP technology.
Consumers calling customer service usually prefer human interaction over being handled by the interactive voice response system. But if chatbot can deliver a customer experience in which customers cannot tell if they are communicating with a human or a computer, it will be a big cost saver for businesses.
One of the advantages of using chatbots is that it remains unflappable when handling angry customers and even in the case of customer abuse. Using chatbots can also ease the rising burden of companies investing in customer service training in comprehensive product knowledge.
Having said that, there are still shortcomings in customer service chatbots by now, for example chatbots may not be capable of fully understanding humans’ context and intent.
Mickey Fong: Hong Kong is a small market for technology-based startups. Do you think we have a sufficient population of talent for developing AI technology?
Alex Hung: I heard a few companies in Hong Kong have introduced chatbots in their operations, but I believe they don’t have a deep understanding in the technology and it takes time to recruit AI talent. There are plenty of professionals with academic background in computer science in the city but the majority is looking for quick success, so they tend to switch to other professions. There are also lots of outstanding talent studying abroad, but they soon move to mainland China or the US. What a shame.
This article appeared in the Hong Kong Economic Journal on Aug. 18
Translation by Ben Ng
[Chinese version 中文版]
– Contact us at [email protected]