How #AI are you? Or is it just the hashtag?
I am sure you feel like I do, when reading articles or hear news about #AI products and when you meet with #AI experts. I find myself often talking to colleagues, clients and students, entering into discussions about how much #AI are we! Is this normal, or do we live up to the hype with buzzword symptoms? In this story, I am making an approach to define the fair use of #AI and I am looking forward to see your thoughts in the comments.
My first experience with AI started back in 2001. I was coordinating an R&D project, in which we were trying to add smartness to our system, to be able to compare an observation sequence with historic patterns, select the appropriate techniques to improve pre-defined KPIs and fine-tune them. We were obviously looking for something more advanced than satisfying conditions in an IFTTT-way. We were looking for something that will learn and evolve with time. The selection of the quasi-AI algorithm of Case-Based Reasoning was the right choice. Lightweight and very efficient, it helped us to create a smart system that was able to support cellular networks of 2nd and 3rd Generation to adapt to highly congestion situations and significantly improve their capacity and performance. Our results raised a lot of interest from the industry and some investors. (We failed however to do the next step, but this is another story).
Another practical experience of smartness I was involved with, was while working in the design of smart home environments for healthy adults and patients. The challenge in such cases is to make a smart home learn the behavior of the inhabitant. Home domotic systems are in general considered to be smart, but are they really adapting to human behavior? For these activities I was also involved with IoT networks and M2M communications, aiming to make devices and machines talk to each other, under a framework of intelligent decisions making that learns from its mistakes.
As the years were passing, I started to get involved more and more in eHealth systems and platforms for Assisted Living. The need for smartness was much bigger, but this time Internet technologies were more mature to support Service Oriented Architectures running in the cloud, giving us a variety of services to develop advanced reasoning and self-learning mechanisms. In the eWALL project that was a key milestone in my academic and business career, we developed a cloud-based eHealth platform that was supporting intelligent decisions-making based on IoT data fusion and advanced reasoning, with the ability to hot plug micro-services and provide the environment for development of real intelligent apps. In the years that followed the successful completion of this project, we developed CloudCare2U in Innovation Sprint based on eWALL open-source, as a platform for continuous research. Under these activities we enhance several features, among them artificial intelligent services that can be consumed by medical and clinical applications.
At the same time, the hype of #AI was at its climax, with numerous stories that could make you laugh in the best case, or even get pissed in the worst case. From marketers up to researchers, we ended up having AI as the top suggestion in Google and phone keyboards when typing the letter ‘a’.
I am not an AI expert, but I have a good experience, including practical implementations and therefore, I am allowed to speak and wonder “How AI are we?”. I am also allowed to be critical to those trying to fool people saying that they have AI systems, while these are only supported by IFTTT mechanics. I am also allowed to raise or answer geeky questions trying to see the level of artificial intelligence, if it is based on a neural network and what type of machine learning mechanisms it is utilizing.
I am opening a parenthesis to strongly encourage you to read two books that I read this summer. I am obviously influenced in this story from “Life3.0”, written by Max Tegmark from MIT, who presents all attributes of AI (and AGI), while avoiding to name drop technologies. And that because he knows and he doesn’t need to give any wow effect in his story. He knows the background and he focuses on the real outcomes, for the year’s and decades to come. The book is awesome and brings up aspects of AI that sound like science fiction, but maybe they are not; maybe they are yet to come. I was recently comparing this book with another one that I read this summer, namely “Sapiens — A Brief History of Humankind”, by Yuval Noah Harari. Harari, after presenting the history of evolution of the past tens of thousands of years, is ending the book with some projections based on today’s technological achievements, such as cyborgs. Tegmark on the other side is considering cyborgs and uploads as a natural evolution, in some of the future AGI scenarios. In any case, having followed the evolution of the Homo Sapiens, it is really challenging to see how the future will be formed.
But, let’s close this parenthesis and conclude, by focusing on the definition of AI and how we should address this term. In my view AI systems should have at least the following characteristics:
- Support intelligent decisions making, meaning decisions will not be based on per-programmed rule engines;
- Ability to learn and evolve, based on observations, cases, errors and evaluation of actions;
- Adapt to the behavior of the users, in a way to support personalized services to be delivered;
- Have the need to be trained during its initialization phases;
- Be supported by technologies and algorithms the allow scaling of systems abilities.
At the time I am writing this story, I am watching my lovely Alexa in front of me blinking and I am thinking of the infinite times I hear my kids asking the very same questions and receiving the very same answers. How much #AI is Amazon Echo (Alexa)? If we are wondering about Alexa’s level of smartness, which indeed we should, then we have better to think twice before talking about AI systems.
