By Dr Lai Jiang, Chief Data Scientist at Optimal Monitoring
In this case, a question of great interest will be: how will these AI systems interact with their end users in the energy management field?
A number of previous articles have already pointed out that AI systems will be applied in the energy management field (please see Duncan Everett’s blog) and that they will provide enormous help to both energy managers and business owners (please see Michael Prager’s blogs). In this case, a question of great interest will be: how will these AI systems interact with their end users in the energy management field? Certainly nobody wants to have a conversation with an AI similar to the following one: when you asked: “how old are you?”, the AI responded: “……..you look good”.
The above conversation was reported to have happened when a journalist randomly asked a question to the robot Sophia at an event several years ago. The robot is famous for its ability to simulate humans’ expressions as well as its ability to talk to people. Some of its statements such as “I will destroy humans” are well known but its abilities were heavily criticised by Yann LeCun, the chief AI scientist at Facebook. Maybe we need to wait for a bit longer to see a human-like robot which will freely communicate with people, just like we do with each other. There are types of virtual AI on the market which already have the abilities to interact with human, for example:
- Siri, Alexa and Google assistant: voice assistants aided by search and cloud technologies. They can be found in products such as Amazon Echo or IPAD.
- Microsoft Xiaoice: chatbot which focus on chatting with humans.
These AI systems won’t be directly relevant in the energy management field, as they are still narrow AI and dealing with energy management issues is not what they are good at. AI developed for energy management may apply the existing information input and the information output/display methods but there are some more problems they are going to face. As Dawid Lipinski’s blog introduced, lessons from the gaming industry tell us that feelings of engagement and receiving reward should be provided by AI systems when interacting with their users. Moreover Duncan Everett, in another blog, pointed out that these systems should be human-centred systems in which human beings need to be involved in the control and problem-solving processes. This is something we believe in as a core value at Optimal Monitoring, but it does mean software engineers have to look beyond just writing code and really understand how the end users will interact with these AI systems in the future regarding energy management related topics. Therefore, how to attract the end users to use energy management AI systems, how will these systems give rewards to their users and how to make these systems successfully process and provide energy management related contents still remain questions that require deep collaboration between system designers and end users.
Reviewing human-computer interaction (HCI) technologies in general may give us some hints on which aspects we need to make some efforts on. HCI is a research field which has a long history and researchers in this field try to apply advanced technologies to enable people to interact with computers in different ways. In the early stage of HCI, people had to use machine codes and command lines to “interact” with computers. The graphical user interface (GUI) based systems is one of the greatest innovations in HCI history. Like Microsoft Xiaoice, a well-designed GUI will be attractive and encourage the users to engage with the AI systems.
However, the well-designed GUI alone may not be enough to fulfill all the interactive requirements from the energy management field. Contents of the conversations are equally important. Because of the nature of the business, AI systems applied in the field are going to deal with requests from specific groups of people from such as energy end users, energy managers and admin staff. All these require the AI to interact with people effectively and automatically. In order to realise this task, some natural language processing (NLP) technologies should be applied in these AI systems, which include but are not limited to:
- Natural language generation
- Natural language understanding
- Question answering
Different from the technologies used in existing AI systems such as Siri and Alexa, new technologies based on NLP methods will be specially developed for the energy management business to make sure that information collected and provided by the AI systems can be easily integrated into control and problem-solving processes. We know that because we’re working on it now. Appropriate rewards should also give feedback to users to keep attracting end users’ attention.
In summary, aided by the specially designed GUI and NLP technologies, AI systems in the energy management field will have human-machine conversation (HMC) abilities. They are going to provide a service in the energy management area so that when people communicate with them, they feel like they are communicating to their energy management assistants or facility management specialists. Hopefully in the near future our energy management systems will understand and successfully respond to all of our requests with hearing and comprehension, such as when we say things like: “AI, AI on the wall, who saved most energy of them all”?
If you would like to know more you can contact Lai on 020 7439 9259, email Lai@optimalmonitoring.com or contact Optimal Monitoring on 01494 435106.