Conversation has been an unmistakably human ability, differentiating us from the rest of creation, being the basis for culture and civilization, and defining our unique level of intelligence as a species. It serves many purposes in our daily lives: communication, coordination, social connection, completion of complex tasks, education, comfort, and entertainment – ​​just to name a few. There's nothing like a good, lively dialogue to get us all excited and creative about a topic or... about each other. A quick return fire makes us laugh or convinces us who to choose for our next president! Humans are really good at having conversations, whether it's for work or just to chat. When we talk to each other, we constantly leverage contextual information and knowledge to convey sarcasm, read between the lines, and express our personalities. Conversation is a high-bandwidth channel in which knowledge, instructions, behavior, emotions, willpower and many other messages are transmitted through language (written or spoken) and its structure. In our daily lives, we typically engage in spoken conversations, while recent technological innovations have introduced us to the habit of almost instantaneous chatting and typing. Say no to plagiarism. Get a tailor-made essay on "Why Violent Video Games Shouldn't Be Banned"? Get an Original Essay The complexity of the presentation, the structure, the conduct of the interaction, and the amount of information it carries are all measures of the intelligence of the participants. We are surrounded by an animal world of inarticulate cries that convey simple messages, and a stone-cold world of systems and machines that require specialized and limited forms of instruction and manipulation. Therefore, humanity's dream has been to naturally interact with tools that do its bidding. Popularized in science fiction novels, the concept has existed since the time of Homer (see Ulysses Rhapsody Σ where Vulcan is served by human-looking maids, golden maiden). Artificial intelligence arrives thousands of years later with the promise of realizing humanity's dream: to have “intelligent” conversations with our machines, meaning that we will be able to obtain information, pass instructions, acquire education or even receive advice in a natural way. In reality, the measure of success of these conversational AI systems is none other than the Turing test, which is also a measure of intelligence: conversing with a machine about a topic or task that would be indistinguishable from talking to another human being. However, human narcissism and creativity aside, what further fuels this effort for intelligent conversational systems and assistants is a long list of enterprise applications and a strong market need for personalized conversations between companies and their customers. Until now, the complexity and limitations that existing dialogue tools and resulting conversational systems suffer from have provided companies and their customers with disappointing experiences. They eventually get the job done, but after great effort, high development and maintenance costs, and relatively limited human-level interaction experience. That's because today's automated conversational systems aren't actually intelligent! Designers with domain knowledge and computing experience define and program each conversation with the written responses that users can expect when interacting with automated dialogue systems. Therefore, conversational systems are built based on tree logicdecision-making, in which the response given by the bot depends on a dialogue state defined by specific intents and keywords identified in the user input. IF the user input contains "buy" or "purchase" (intent); And "cellular" or "mobile" (product type); THEN send a message with the cell phone list Typically, designers need to program 3 main components to create an automated conversational system: a) the natural language understanding part, i.e. the part that analyzes and analyzes human language and identifies the parts important to the task, b) the dialogue management part, which basically identifies a state for the dialogue based on the history and current parsed input to decide what to do next, and c) a response generation part, in which the designer typically programs responses via system scripts. This means that the resulting systems will seem as smart as the effort (and patience) put in by the designers who created them: capturing and anticipating a large number of potential use cases and inputs, creating appropriate and natural responses. Furthermore, adapting and maintaining such rule-based dialogue systems with changing information or new information about a task is a time-consuming and labor-intensive programming job. To mitigate these shortcomings, next-generation dialogue systems must be capable of learning. To begin with, there are two easily accessible sources of knowledge: a) examples of human-to-human interaction and b) existing data (books, websites, manuals, databases). After all, this is what business also has at its fingertips. Companies have collected huge amounts of sample conversations from the interaction between their agents and their customers via phone or other channels (online, Twitter, etc.). Likewise, companies have abundant organized (databases, knowledge graphs, websites) or raw (documents, manuals) data that contains knowledge related to business operations and a variety of business services (travel booking) or customer-oriented activities. objectives (maintenance, repair). However, so far several attempts at example-based automated systems have been unsuccessful and have led to unexpected or even comical results (remember the Microsoft Tay offensive and the funny challenge of Facebook negotiator robots). The reason is that you have to be careful about what data you put into the system when you train it and also what mechanism you use to produce responses. Such systems will be accepted in business only when their responses are not free-form, but can be limited within a set of responses acceptable by the Business. What emerges from the above is a deeper truth: today's automated conversational systems lack the connection to available knowledge. This connection and transfer of knowledge from data is so far provided directly by human design and programming. Therefore, when we build an automated dialogue system, we have to recreate existing functionality from scratch, that is, recreate the experience of our website in a different way, while the underlying information is the same. This means extra work for companies to create a completely different channel to handle customer requests, which also results in a lack of consistency for the user. In contrast, human agents do not have this problem. They are able to continually acquire knowledge by observing others or from documents, assimilate new information, and organize themselves appropriately so as to increase their ability to conduct new goal-oriented conversations. Transfer this human ability, even to some extent, to our automated dialogue agents and, thus, learn fromexample dialogues and existing knowledge sources, it is necessary to make progress and make it clear that the market wants it. If we can appropriately represent knowledge on a goal-oriented task and automatically integrate it into neural network or traditionally programmed dialogue systems, we expect to gain several benefits. First of all we expect to improve the performance/response accuracy of the systems. No more dependence on how thorough and experienced the system designer is; the system will learn everything it needs to know from the available data. Even more so, we will be able to instill some “common sense” into systems that can switch between applications. Systems will be able to quickly adapt to new and changing information, as well as function and evolve in the absence of exemplary dialogue. Yet this connection between dialogue and knowledge remains elusive: it is a difficult problem for machines. Leveraging knowledge and limiting conversational systems to the set of acceptable responses will greatly automate their development and maintenance, but that doesn't mean we shouldn't still put effort into programming or improving their language understanding capabilities. Conversational interfaces represent a big shift in how we're used to thinking about interacting with our "dumb" computers and "smart" phones! Conversational computing is a paradigm shift that requires designers to change their thinking, their deliverables, and their design process in order to create successful bot experiences. We therefore expect that great progress will also be made in the systems and tools that allow the composition and integration of the different components linked to dialogue. Designers must be able to leverage the strengths and facilities provided by different AI technologies when analyzing human language or conducting dialogues learned from different sources. Therefore, over the next couple of years we will see a proliferation in the market of tools that will not only support building conversational systems from scratch, but also enable evolution towards similar tasks, knowledge transfer and dialogue management, and the seamless incorporation of new data. Such conversational interfaces will become the new operating system or digital network that holds technologies together. Our future will be inundated with digital assistants, drones, robots and self-driving cars. Therefore, we must also look for innovative ways to converse with these new devices. This means not just providing one-way instructions or questions, but conducting two-way interactions that meet our needs. This is where conversational computing comes in. We need conversation not just to fill out forms or step-by-step instructions, but we need it because we don't know the ever-changing options (e.g., available tickets and dates, or new situations encountered) and the systems don't know our needs , preferences or do not have our special training and wisdom at a given time, to complete a task. Big companies are investing heavily in the conversation sector: Google, Apple, Amazon, Facebook, IBM, Baidu, just to name a few. And by mastering the conversation, they can dominate the world. The next Alexa will be your home assistant or your hotel concierge. The next Siri or Google assistant will be your personal assistant in the office and at home. Facebook will interact with you just like one of your friends. But apart from dominance in our daily lives, conversational systems will take over in the business world, as they will deliverfaster, better and cheaper customer service. This is where companies like IBM and many startups are making their play. Gartner estimates that AI will account for 85% of customer relationships by 2020, and recent market analysis indicates that today 60% of repeat customers (that's you and me) would rather speak to an automated system than to a human to complete simple tasks, if it is faster and more informative. However, most of us (over 70%) still don't trust automated systems for complex tasks or with our money. Furthermore, according to the survey, most of us don't want to rely on automated assistants making decisions for us. These cases still require the human touch, someone who understands our needs, can negotiate, is able to explain and lead the conversation towards a win-win solution. We therefore expect great progress in systems that demonstrate more human characteristics and that take into consideration more ways of interaction than simple typed messages. Microsoft, for example, is working on a natural user interface (NUI) that combines natural language with gestures, touch and glances, to help deepen system conversations. Everything can be “heard” by sight, touch or sound. This is the type of multimodal conversation that will be automated into more human-like conversational systems. Google recently unveiled Duplex, a concept assistant that makes appointments and reservations for you – it looks and interacts very human-like. Artificial intelligence can play an important role in this. Research is already prototyping deep learning for increasingly “deep” conversational systems: instead of learning dialogue from examples of textual dialogue, new AI systems are in the works that will learn directly from spoken interactions. MILA, the UN Research Dialogue Team. of Montreal, Facebook, Samsung, Microsoft and Google are already working in that direction,,,,. It will be very powerful. We remember from our youth that spoken dialogue was our primary and only skill for learning, playing, and teaching others. We were able to negotiate with our parents before we could read and write, and we were able to describe how to play a game to our peers or coordinate ourselves to play it, before going to school. Spoken dialogue interaction is a rich form of intelligent communication that only becomes more sophisticated and complex over the years: it continually incorporates what we learn from our interaction with others or from sources of knowledge (books, documents, articles, manuals, etc.). close to seeing such systems anytime soon? “You must have patience, my young Padawan. " Our current experience concerns linked systems, i.e. systems that have a speech synthesis component that transcribes our speech which is then passed to the conversational system and the response is read to us. These systems often make mistakes and get confused, resulting in negative experiences for customers. One reason is that they are not tightly coupled: information about what the conversational system expects is not returned to improve the transcription system, and vice versa an error the transcription is passed blindly to the conversational system. Learning directly from examples of spoken dialogue can solve many of these problems and also incorporate new aspects into the intelligence of automated dialogue: emotions, attitudes, styles of expression, vocal inflections. We will speak with a machine that is able to better understand some elements of our human nature and it has been programmed or “trained” to respond and evoke.