We’ve always been fascinated with artificial lifeforms, be it robot, cyborg or some other futuristic hybrid. Sometimes they’re the loyal family help. Sometimes they’re bent on wiping out mankind. But that is entertainment. This is reality.
How smart is current AI?
TAKEHIKO RYU: AI has advanced rapidly over the years, with some aspects such as computer vision, text analysis and conversational AI coming close to human ability. While many consumers tend to draw references to AI based on what they see in popular culture, in reality, AI is already prevalent in our daily lives and its applications are wide-ranging.
If you take a quick look around you – the voice assistants on our smartphones, the virtual customer service officers that we interact with online and the facial recognition technology at institutes of higher learning for attendance-taking of staff and students – all of these are AI at work in our everyday lives.
In recent months, we’ve also seen AI being widely deployed in the form of autonomous robots at parks to assist with the enforcement of social distancing requirements, in malls to assist with cleaning and disinfecting procedures, at roadsides to aid traffic ﬂow management, and at hospitals to deliver medicine, specimens and patient notes to healthcare staff.
I remember movies like War Games, Total Recall, The Terminator etc. how close are we to that? Do we need the 3 laws of robotics or is that unlikely to happen?
TAKEHIKO RYU: While the possibilities for AI are limitless, responsible AI requires us to put careful thought and consideration into how it is designed and implemented. Biases need to be taken into account in the design of AI. One source of bias can happen through ﬂawed data sampling, in which certain groups are over- or under-represented in training data.
At Panasonic, we are conscious of the risks of biases during the development of our solutions, and we handle such data imbalances through data augmentation and algorithm improvement. We also ensure that training data are universally collected from all aspects of the global population to ensure fair representation and avoid biasness as much as possible.
When it comes to AI implementation, it is important to understand that not every problem requires an AI solution. We first need to assess whether AI is feasible and suitable to be implemented for the project requirements, then consider how the AI solution can be integrated with the existing system, how it can be customised to meet the organisation’s needs, and how it will affect current operations. Another aspect to consider in AI implementation is the provision of “explainable AI”, to ensure greater transparency in the decisionmaking process and that decisions made by AI can be accounted for.
WHEN IT COMES TO AI IMPLEMENTATION, IT IS IMPORTANT TO UNDERSTAND THAT NOT EVERY PROBLEM REQUIRES AN AI SOLUTION.
We talk about smart cities, buildings etc. is that AI?
TAKEHIKO RYU: More than ever before, AI is featuring in a big way in smart homes, smart buildings and smart cities. In today’s homes, we can find AI-powered home appliances, personal assistants and even smart care solutions such as walktraining robots that empower the elderly with the ability to lead more independent lives.
On a larger scale, AI is being used in smart city development to enable a better and safer living environment for everyone. For example, Panasonic’s Smart Street solution uses AI to analyse pedestrian traffic, track speed and volume of the traffic ﬂow to help reduce accidents and improve road operations; while AI in driverless parking systems helps to reduce vehicle accidents caused by drivers’ mistakes.
IT vendors and the government have been talking about digital transformation, where does AI fit into that?
TAKEHIKO RYU: With massive volumes of data being more readily available now than ever before, businesses can leverage AI more effectively to accelerate the digital transformation process. For instance, Panasonic’s automated facial recognition gates at the Tokyo Narita and Haneda airports in Japan use AI to speed up the immigration processes. By comparing photographic data of the individual’s face in the IC chip embedded in the passport with an image taken at the facial recognition gate to verify his or her identity, this removes the need for prior registration of biometric data and improves productivity and the efficiency of immigration clearance at the airports.
We are also partnering with FamilyMart in a pilot project to utilise IoT and AI-powered facial recognition technology at convenience store chains. Not only does this help speed up payment verification processes, it also improves the overall customer experience at the convenience stores.