When I first heard “Artificial Intelligence” (AI), I thought of fully autonomous robots, like those appearing on the movie "Ex Machina". Actually, it was at my graduation project, a recommendation system for a hotel chain in Mexico, the first contact I had with the artificial intelligence world.
Briefly, (quoting Einstein: "If you can't put something in simple language, it means that you haven't understood it well enough"), it was an app with a strong multimedia component, and a recommendation system.
The rationale behind it was to provide a salesperson with all the necessary information on their IPad, at the resort's point of sale. They would then know what product to offer and be prepared with the appropriate multimedia show. The recommendation system was for the salesperson; the system was fed with a customer’s profile, and the system made recommendations to the salesperson who ultimately made the decision on what to offer the customer.
All the customer’s data and their travel history had been already stored in order to "train" algorithms. Input data consisted of information such as: age, single, couple, family, number of children, country of origin, likes/preferences (spa, sports, parties, peaceful place, yoga, among others), and if a customer had already bought there before, the point of sales would already have a record of additional information such as all their travel history, kind of customer, income band, money spent in resorts, kind of job, or position in the company. After doing research, we opted for CBR (Case-based reasoning) Jcolibri2, a free framework. Source
Nowadays, artificial intelligence is on its heyday, with a large and promising field of expansion. Already in the 90's, 80's or even earlier, the concept of artificial intelligence was being discussed, since it had been considered as a basic computational theory. Now, the question is, what has changed to become one of the most mentioned words everywhere?
To begin with, the AI phenomenon takes place after the Big Data explosion. Just think of the amount of information that is generated in devices, and sensors daily. This enormous amount of information was not generated in stages prior to Big Data. So now, we have more processing power and better algorithms, to the point that hardware already has processors with artificial intelligence modules such as Apple Neuronal Engine or Google tensor processing unit (TPU).
Technological market leaders such as IBM, Apple, Google, Microsoft, Facebook, and Tesla, among others, have already been working for a long time on artificial intelligence research and development, so much so, that they have released open source libraries for developers.
We also have IBM’s breathtaking Watson the result of impressive development and evolution which began long time ago, when Deep Blue beat Garry Kasparov, the best chess player in 1997. Quite recently, Watson beat (I would rather say destroyed) the best two Jeopardy players in the USA.
Jeopardy is an extremely popular game in the USA, that consists in someone formulating an answer and the participant asking the corresponding question afterwords. For example: "He is the CEO of Apple", and whoever touches a button first, responds "Who is Tim Cook?”
After winning Jeopardy, IBM announced that they were going to work in the field of medicine, for example analyzing substantial amounts of data related to genetic code, human genome, even X-ray analysis to fight brain cancer, and initially against lung cancer.
Pizzeria in Norway
AI has been used in other areas such as marketing. There was one case that really took me aback, and that was in Oslo, Norway. There was a screen showing advertisements, where there was also a hidden camera too, running facial recognition algorithms. The story goes that the screen was broken and the scanning that was going on behind the advertising was unveiled.
If you look at the image below, the number of people who saw the advertising, in this case has been 74 people (green box). You can also see that person 69 is a young adult woman with a 2015 attention score out of 4218 that should be a number showing how much attention she paid to the topic (red box). in the yellow box: adult male, attention (none), smiling (no), wearing glasses (yes). Source
A huge amount of data can be gathered this way, for instance, whether the person was smiling, sad, if someone paid attention or not, and if so, how much. Tools such as machine learning and Watson allow advertising to know how old you are, what your expression was when you saw the advertisement and whether you paid attention or not.
Target Case Study
Target uses AI to do statistical analysis of its customers’ purchasing behavior.
More often than not, most of us have been given a “frequent buyer” card, to have access to exclusive discounts. This point code card is used to perform marketing tracking on your purchases, and match your identity to your shopping behavior, in order to send you promotions, or gather more information on how to improve a supermarket’s sales. In fact, this is what Target applied in the following case: they sent a teenage girl in the USA, advertising for babies’ objects in general (e.g. diapers, toys, clothes for pregnant women). When her parents saw this, they got angry and sued Target. It turns out that their daughter was pregnant, and she did not even know, but Target did.
Why did Target know? Because Target had used machine learning on her, to compare her consumer behavior, not because she was buying diapers or powdered milk, but because it was a woman's normal behavior in her early months of pregnancy. Source
Interestingly enough, this took place in February 2012, more than 6 years ago. Ever since, The AI revolution is everywhere.
Watson, for example, has Marketing automation, where all the steps of a user's behavior are created in order to improve sales, from the moment they discover you, until they learn about your brand, try your product or service, buy it, how they use it and eventually, how they become influencers (people who advocate for your brand).
Even a trailer has been created with Watson, Morgan film by 20th Century FOX. Source. This is one among all the many things Watson can do.
Uruguay has been working and deepening in the area, and the government are giving guidelines on the subject. The Minister of Industry, Energy and Mining, Carolina Cosse, in her speech of July 18 emphasised that AI would be one of the axes of the country’s development . Source
At Quanam, we are doing deep research, to the point of creating song lyrics with AI using machine learning techniques and NLP (Natural Language Processing). Source
Moreover, as a result of a partnership between Quanam and GenLives for a project, a great deal of work has been done in the genetic field to detect if a person is prone to or has a disease because of their genes. The project was given the first place in the Data Scientist category at the Dataworks Summit EU 2018, Germany. Fuente
What does it take to learn AI? Apart from basic programming, in my opinion, there are two levels. One is when we can use machine learning libraries to develop machine learning, in order to generate a user experience; for example, an object classifier, recognize if a person is male or female. It is possible to reuse models that already exist on the Internet, by using and testing them in other environments such as a telephone, or in the cloud by generating a client
Level two is understanding how AI works, the mathematics behind algorithms and what makes it work, where the knowledge and its intelligence are. At this level, new algorithms and solutions can be created and developed, because many problems are both innovative and complex because there is no specific solution for them, yet.
We could spend days talking about AI because it is too large a field: there are diverse types and categories of AI, together with their own techniques and algorithms. The most well know are neuronal nets, machine learning, deep learning, among others. AI has made an impact in several industries. Let us take Tesla in the automobile industry. Their autonomous cars reach an autonomous level 3, involving various kinds of sensors whose AI recognises environment and can make decisions in real time.
Ethics is by no means a minor issue in AI. What is right and what is wrong. Making a good or a bad use of it. There is a recent case where Google ended up deciding not to renew the Maven project contract with the U.S. Department of Defense.
When the developers found out that the project was intended for drones to be able to recognize people and objects in greater detail, thus causing lethal casualties, they opposed by saying that it was against the company's values. In protest, referent developers resigned and even an open letter was signed by four thousand Google employees. Source
Clearly enough, AI is here to stay, having a major impact in the community, industries, and big labour projects.