How do you see the future with Machine Learning & AI?
There is a general acceptance that machines will eventually exceed human performance in different aspects; optimists even believe it can happen within 10 – 15 years time. We are in the 3rd wave of the AI rush. It is fuelled by the volume of data that we can generate and with IoT. We can collect even more real time, multi-dimensional data which helps us to understand our context better. Though existing algorithms are successful in tackling some very challenging tasks, these technological achievements are largely based on the availability of data with significant drawbacks like, data hungry and slow training and learning.
In the future, we would expect improved algorithms which will greatly shorten the time for training the model, and solutions that are less data hungry. And by then, more complicated and skilled tasks provided by professionals can also be augmented or automated.
Do you agree with Elon Musk that AI should be regulated?
Having it regulated mostly comes from fear. Unknowns scare us when the machines grow to become smarter. But fear does not simply come from the unknown and technological advancement. To me, the true fear comes from the differences between human cognition and machine cognition. Even though we try very hard to mimic and to create sensors for the machines, they are not using the same cognition or value systems as we have to interpret different environment parameters. Machines are trained to improve efficiency. That’s why we have to either make the machine reason / infer based on the intelligence or values learned from humans or have an emergency stop button to prevent things going wrong that could harm mankind.
What do you see as the future of KAMI?
As technologies evolve we would like to have a more intuitive way for human-machine interaction. Major technology companies put much of their energy into improving natural language processing and understanding. At KAMI, we believe the best way to understand humans is through conversation because it’s the only portal to human thoughts.
With neural network and deep learning, we can see great progress in Natural Language Understanding (NLU) with advanced syntax or semantic analysis. Applying them in the virtual assistant area, we can now easily operate our systems and complete business use cases with simple verbal / text commands. However, machines still lack a natural understanding of different entities. To make a truly smart machine which can learn through reasoning with facts gathered from ongoing machine-human interaction, we have to bring machine cognition to the next level, and pragmatic competence and machine reasoning are inevitable. KAMI focuses on representational learning which empowers systems with reasoning capability, so it can not only learn from humans but since it reasons, knowledge can be generalised and eventually we make intelligence truly transferrable.