First time I saw the great update of march 2015 in Android Google Photos I knew something was happening and I knew that it was passing far away from me. Google was recognising and classifying images as it was text how can this happen?
It took me time to barely have some notions and starting pointing my googlings to the right direction of how they do something like that. They where doing it with artificial intelligence. At this time I was trying to boost my career assisting to all the possible meetups, conventions and presentations available in Barcelona. Nobody was talking about AI but I was listening a lot about big data, data scientists, Hadoop etc… I found some courses in Coursera about machine learning but it was very hard to pass the first lesson. I assisted to one meetup in the Politechnical University of Barcelona about data science done by Dr. Oriol Pujol and I loved the spirit of the university so I overlooked the courses and found the Masters in the computer science school. On September 2016 after more than 15 years from finishing the University I started a Master of Science in Artificial Intelligence on the UPC and I will tell in another story how it was how it was.
I pass it and it has not been easy but what it is coming now to me is a lot of ideas and possibilities about how to apply it in the business what was learned. Apply Artificial Intelligence to products or services consists in cycling phases around the digital data: acquiring data, processing data, understanding data, recognising this data (finding patterns) and taking decisions with this insights AI disciplines acts in all this cycle.
For acquiring data we have Computer Vision for video and images or Natural Language Processing for text understanding and analysing. All those acquired digital data are numbers grouped in matrices or vectors and as numbers they can be treated with mathematical algorithms. I see an algorithm as a process of different steps for transforming and acting against those numbers. Computers are mathematics computing machines and for processing and understanding those numbers they use different adapted algorithms build and run by software programs. Having acquired those images or texts it is now time to use Machine Learning, Computer Intelligence techniques like Neural Networks, Bayesian networks, Classifiers or Genetic Algorithms among others hundreds of scientific techniques for learning to predict a result. I mean, if you input your software with images of beaches he has to always be able to predict other images being pictures of beach or if you you input to your software a text like “I didn’t like the food” he has to be able to inform you that your customer dislike your menu. This is automation, this makes possible to find knowledge in the immense amount of data existing and produced in digital ecosystem. Machine Learning is the fuel of lots of companies, you know the hours of sleep or kilometers walked because behind your sensor (smartband or smartwatch) there is machine learning identifying a pattern, there is machine learning behind the recommendation engine of Netflix and etc.
Besides this basic two process there are other disciplines studied in a master of Artificial Intelligence like Knowledge Discovering, Planning, Reasoning or Multiagent Systems but the three that I find more interesting for applying AI to the business are: computer vision, natural language processing and machine learning.
If we try to understand what is AI more than calling to an API driven service like Amazon Rekognition or Lex we will find an amazing amount of mathematical techniques build during dozens of years of applying science inspired in life evolution for evolving beyond its limits the computing discipline. I personally dislike the amount of terms related to human intelligence and brain as I think they biased the approach to this discipline specially for the infinite amount of applications that can be done. 15 years ago building and maintaining websites was a difficult experience and hosting your site was painful, nowadays it isn’t anymore. I think AI will have the same path in a few years it will be more accessible to everybody and it will change the way that we build services and products this can driven by companies and business to make better things.
I love this Tweet of François Chollet “Neural networks” are a sad misnomer. They’re neither neural nor even networks. They’re chains of differentiable, parameterized geometric functions, trained with gradient descent (with gradients obtained via the chain rule). A small set of highschool-level ideas put together
