The Washington Post recently published an article highlighting the most important technologies that have changed our way of life over the past decade.We are at the end of another decade and the beginning of the new will bring with it new technologies and interesting evolutions of the previous ones.
We therefore decided to retrace how Artificial Intelligence (AI) and Natural Language Processing (NLP) contributed to the creation of some of these technologies. CELI has been operating in this area of business for 20 years and has contributed to the completion of projects that have changed the world forever and above all the way of accessing technology.
But let’s get to the technologies that changed our lives.
1- Smart speakers and voice assistants
Whenever we talk to Siri, Alexa or Google Home we have direct interaction with AI.
The rise of voice assistants began in 2011, when Apple released Siri for the iPhone. Google then launched Google Now in 2012, followed in 2016 by Google Assistant. However, voice assistants have been often confined to smartphones.
Amazon Alexa, introduced in 2014, has led to the explosion of the voice assistant market and has brought AI to many of our homes. The introduction, in 2016, of Google Home further pushed the growth of this sector.
A few figures: in the third quarter of 2019 alone, Amazon shipped 10.4 million smart speakers worldwide that use Alexa, conquering almost 37% of the global market for these gadgets (Canalys 2019).
Voice assistants have definitely changed our relationship with technology: being able to communicate with a machine has been one of the objectives of computer science since the dawn (“Computing Machinery and Intelligence” 1950, A. Turing) and a suggestion of science fiction (just think of the 1968 film, “Space Odyssey” by Stanley Kubrick).
What seems new today is that every time we speak to a voice assistant, our voices are recorded to improve their NLU (Natural Language Understanding) system. Therefore, the previously unidirectional relationship has now become mutually beneficial: these tools offer us their services while we directly contribute to their improvement.
Alexa, Siri and Google Home also helped us feel comfortable with the idea that there is only one answer to a question. With traditional search engines, each search query corresponded to a series of pages that probably contained the answer to our question. If, however, we ask a voice assistant “I would like to cook a soup”, the assistant will give us directions following a single recipe, chosen automatically.
From research and analysis of requirements, through design to the creation of all-round Voice First products: CELI helps companies manage complex and constantly evolving contexts that include the use of voice as a tool for interacting with end customers.
2- Recommender systems
Netflix, the streaming movie and TV series service, was born way back in 1997 but only in the last decade has seen its biggest development. How do Netflix use AI to improve our experience?
First of all through the use of our watching history: through recommendation algorithms (Recommender Systems) we are shown the titles that we will like and will convince us to pay our monthly subscription. In addition, AI is also used to optimize the use of regional servers that guarantee faster loading times even in peak times, choosing the servers to be used based on the history of the visualization data.
This same technology has been used for some time by Amazon, Spotify and Sky to improve the user experience and push customers to see the next TV series, or to buy a new product.
Historically recommendation systems are divided into two families:
- based on similarity between users: user profiles are automatically created with clustering algorithms. In this way, having identified two “similar” users, it is easy to offer the former the products the latter liked
- based on the similarity of content: information extraction algorithms allow to enrich the products of the catalog with metadata (actors who participated in the movie, classification by genre, date of publication, etc.). At this point the main document is considered as a query in a semantic search engine and the first results become the suggestions of the system.
CELI has experience in both types of systems.
To create a content recommendation system, a client can rely on CELI technologies that allow to extract structured metadata from an archive of unstructured data, be it text, audio or video. In addition, our proprietary search technologies are easily customizable to build the engine of a content-based recommendation system.
3- Automotive: Tesla model S and Mercedes MBUX
The automotive world has always been the area in which technological innovations are designed and applied in advance of other sectors. Without a doubt, the last decade has been studded with great innovations that promise to bring us to autonomous driving in the next decade.
Tesla is the first example that comes to mind: a car conceived more like a smartphone whose software is updated via wi-fi. Tesla collects data from all its vehicles and their drivers, with internal and external sensors that can collect information on the positioning of the driver’s hand, the instrumentation and how they use it. The data is used to generate high-density maps that show everything, from the average increase in traffic speed on a stretch of road, to the location of possible dangerous elements, so as to induce drivers to act.
Machine learning in the cloud is concerned with “educating” the entire fleet, while at the individual car level, edge computing decides what action the car should take. There is also a third level of decision making, with cars capable of forming networks with other Tesla vehicles nearby in order to share local information and insights. In a scenario in the near future in which autonomous cars will be widespread, these networks will most likely also interface with cars from other manufacturers, as well as with other systems such as traffic cameras, road sensors or cell phones.
Mercedes MBUX is instead the best example that shows how natural language can be successfully applied in situations where hands are busy.
The Mbux system works like Siri, Alexa and Google Aassistant, thus distinguishing itself from the systems present in other cars that are purely executors of basic voice commands such as making a phone call.
MBUX, equipped with “Hey Mercedes”, is an artificial intelligence system that interprets the command, processes it and interacts with the driver. For example, if you just say: “Hey, Mercedes I’m hot!” automatically the system will turn on the air conditioner and lower the car’s temperature.
The excellent results obtained by Mercedes are also the result of a long-standing cooperation between CELI and Nuance (now Cerence for its automotive branch).
4- Knowledge Graphs
Knowledge Graphs are datasets that try to enclose all the common sense information by structuring it in a graph of knowledge according to a precise formal logic: geographic places, famous people, works of genius (film, music, etc.) are linked together by arches that define its relationship.
With graph support, for example, Google helps to disambiguate searches by identifying real world entities rather than text strings (for example: Verdi, the composer, the party or the color, are three different entities in the graph). Furthermore, once the entity sought is identified, the graph allows to facilitate the discovery experience, navigating the semantic relationships starting from the entity: for example from the film sought, to the leading actor, to the other films shot by that actor or to news about him.
Google builds its graph “at home” but there is one in the public domain, supported by the organization that also maintains Wikipedia and populated by the crowd: Wikidata (www.wikidata.org). Google’s knowledge graph is populated starting with wikidata and Google itself is one of the major funders of the project. In 2010, Google donated $ 2 million to the Wikimedia Foundation. The initial development of Wikidata was financed with a donation of 1.3 million euros, a quarter of which came from Google (source: Wikimedia).
Voice assistants, in particular Amazon, also use knowledge graphs, for example every time we ask Google Assistant or Alexa who is a famous person, information about a film, or indications on a geographic location.
5- Instant messaging
Chat communication technologies have undoubtedly exploded in the last decade. The concept of instant messaging began to develop in the 90s, allowing friends, acquaintances, colleagues and related thinkers from all over the world to connect in real time.
Since then, instant messaging has revolutionized the way we communicate and today over 2.5 billion people are registered with at least one messaging app. The current instant messaging experience is fluid and intuitively integrates features such as videos, photos, voice, e-commerce and games with simple messaging.
However, despite the extraordinary features of dominant apps like Snapchat, Facebook Messenger and Whatsapp, today’s technology would simply not be possible without the previous discoveries of their most rudimentary predecessors.
Facebook is one of the big ones who believed in it more, given that in 2014 it acquired Whatsapp, a service for which it also offers the business version, and in 2016 it announced at its annual conference that it would focus heavily on the Messenger platform to facilitate the communication between companies and customers (a bit like WeChat was already doing in China).
The messaging platforms have been open for some years also to the possibility of using chatbots. A chatbot is able to answer most questions from users of a service without having to involve human operators, usually contactable through call centers. However, he is also able to realize when there is a need for more complex interaction and automatically forward the conversation to a human being.
In a world where new technologies seem to be able to change everything that has happened before them, the CELI team has learned how to look to the future without ever stopping and without being fooled by easy enthusiasms that often lead to failures.
The technologies we have talked about have been disruptive over the past decade. For this reason, we have written this article convinced that they will also last in the coming years, having now reached a level of maturity such that not only the big names in the sector can afford to experiment with them by taking on the risk.
At the same time, our engineers and linguists are already experimenting with the latest discoveries to offer our clients solutions to their problems that are both innovative and reliable.
Would you like to know more? Get in touch with us!
Francesca Alloatti, Computational Linguist
Matteo Amore, Computational Linguist
Mariella Borghi, Marketing Manager
Riccardo Tasso, Software Engineer
Raffaella Ventaglio, Senior Software Architect