At HUAWEI CONNECT 2018, we unveiled our AI strategy and portfolio. At HUAWEI EC0-CONNECT EUROPE 2018, we explored how we can work with our partners and customers to create an open industry ecosystem and help build pervasive intelligence in Europe.
We also spoke to Europe’s industry leaders and prominent experts about artificial intelligence, including Marco Menichelli, CTO of XSENSE. The full interview transcript is below.
What are the latest developments in AI that you’re most excited about?
Marco Menichelli: What I call Artificial Intuition. With the Deep Sensing approach, we’ve achieved incredible results, and these results are evident. With Artificial Intuition, or more technically Artificial Insight, we will create a truly intelligent, new life form.
Can you explain the Deep Sensing concept?
Menichelli: The base concept is very simple. Deep Sensing is an approach to design AI algorithms that studies and aims to emulate how the human cognitive processes work, not the functioning of brain, or at least not just that. One of the most important principles of this approach is that, with Deep Sensing, we simulate each cognitive process that’s involved in all cases that require intelligence and experience. By simulating our cognitive processes, it isn’t necessary to develop a Neural Network for each function.
If we simulate intuition, the same process can be used in all verticalizations. For example: if we develop a classical Deep Learning algorithm that can play chess, the same algorithm can’t play Go. With Deep Sensing, the same process used to win at chess can be used to play Go, poker, or other games. But, it also can be used to resolve any type of other problem like, for example, completing a sale successfully. The Deep Sensing approach is much more complex than the classical approach, because it implies a deep knowledge of the functional processes of the human mind.
By ‘the functional processes of the human mind’, I mean all types of human intelligence like linguistic intelligence, emotional intelligence, insights, and so on.
As one of your specialist areas, what kind of benefits can natural language processing deliver?
Menichelli: I prefer to speak about General Intelligence – linguistic intelligence is just a way to demonstrate intelligence, but it’s not the end. We cannot pretend to teach our language to a system that doesn’t have real basic intelligence, it’s like trying to teach how to speak to a worm. However, in my opinion, the use cases of are many.
How can challenges such as semantic ambiguities be overcome?
Menichelli: We’ve already overcame disambiguation problems, for example XSENSE is able to study any language autonomously and contextualize each word, in each sentence, to arrive at the deeper meaning. There is no dialogue without context, so the context is the base of each conversation and the context consists of many variables, variables that can be understood during the conversation.
The main challenge is to understand user intention. Why has my interlocutor asked me this question? What does he want to get out of it? By using an adaptive intelligence it’s possible to find the answers to these questions. Our UNXFLOW is able to identify the context of each sentence or question without anyone having first configured it. Let’s switch the focus from the Workflow to the Target because when we talk to each other, we always have a target that’s not just about communicating. UNXFLOW is able to reach a target autonomously, adapting its own behavior in function of the external stimuli.
What kind of use cases can we expect when General Intelligence combines with computer vision, advances in Human Machine Interaction, and brain-machine interfaces?
Menichelli: It’s a good question. In my opinion, all these technologies find perfect integration in smart glasses. Because, no one will wear special helmets to read thoughts, but everyone is willing to wear glasses. Moreover, glasses are able to give to the human mind what it needs: input in the form of big, clear images and sounds.
When this happens, it will no longer be necessary to speak or write, and in many cases, it will not even be necessary to think in order to obtain something, because this General Intelligence will be able to know us like nobody else. It will be able to learn the way we speak, write, act and think. Technically, this type of AI will learn to be us because it’s able to read, listen, view and to sense what we read, listen, view, and sense. The application fields are infinite.
“No one really knows how the most advanced algorithms do what they do.” Do you agree? Or are we on the way to making deep learning more understandable to its creators?
Menichelli: I’m very happy to have the chance to say my opinion about this.
Because this is a false problem. Imagine that we have a matrix with billions of variables. Can we predict the behavior of that matrix on each case? no, otherwise we would not need Artificial Intelligence. Then we would have to build a control matrix, but that would have billions of possibilities anyway, and of which we could not predict its behavior. This is a false problem also because the Neural Networks are not the future of AI. If we switch the focus from the structure to the process, as Deep Sensing does, this problem doesn’t exists, because, for example, I’m perfect conscious of the processes that I’ve created, and if you ask me to predict how it will behave in a particular case, I would be able to tell it to you, because I know the process that I’ve designed, and I don’t care about the data that it manages in each particular case.
Huawei launched its AI strategy at Huawei Connect in October 2018. What do you expect from us in the AI space?
Menichelli: I expect Huawei to work towards making AI more accessible and inclusive and to apply it to the betterment of enterprises and society. Your full-stack AI product portfolio positions you well you contribute to playing a leading role in the global AI ecosystem.