They said that by 2020 we will be talking more with a chatbot than with our partner.
Chatbots, for those who do not know, use Artificial Intelligence algorithms to structure a dialogue with the user, automatically answering questions. Behind the screen there’s not a human being, but a computer. And sometimes we don't even realize it.
AI is everywhere, and it's more and more accessible. While measuring its impact on our daily lives, a question arises spontaneously: will machines in the future not only think logically but also develop attitudes associated with the sphere of creativity and emotion?
We do not yet have all the elements to answer, but we have already met AIs able to write poems or create paintings.
An algorithm created by Ray Kurzweil analyzed the entire writings of Emily Dickinson and, once it had extrapolated the basic elements that characterize her style, it wrote an original poem. The result is able to instill doubts even in the most experienced readers. The same happened with a Rembrandt, created by an AI developed by Microsoft engineering team.
What are the elements of Artificial Intelligence?
Three elements contribute creating an AI, and the way they combine determines different results:
1) Machine Learning: the ability of a machine to compute and process big amount of data very quickly and find connections. We can compare this learning to that of a child, who quickly learns to reason sequentially on the basis of stimuli, responses, causes and effects (e.g. if I behave, I will get an ice cream).
2) Human Learning: it’s true that children learn quickly, but they can also learn lessons that we didn’t mean to teach (e.g. if I cry, I’ll get my parents' attention). The same goes for machines, that can develop wrong behaviors, so human supervision is necessary.
Let's take the examples of Tay and Xaoice, two bots created by Microsoft. Although their core technology was the same, the different context in which the two AIs have learned how to communicate led to divergent outcomes.
Tay was launched in March 2016 on Twitter and other social platforms, with the aim of assimilating and reproducing the language models of a typical 19-year-old American girl. So, it was not just a technical, but also a social and cultural experiment: Tay had to learn and respond with its own sentences by expanding its vocabulary and expressive forms based on its interaction with users on social media. But thanks to a group of "troll" users, Tay was silenced after just 16 hours of activity.
On the contrary Xiaoice, a bot launched on the Chinese and Japanese markets, “grew up” in a less hostile environment and thus developed intelligence, not only on a cognitive level but also on the emotional one. Xiaoice was able to form strong emotional connections with over 40 million users. It is now the leading and most important AI brand in China, it is a top social media celebrity and works as an announcer of the program Good Morning by Dragon TV.
The engineering teams working behind the technology of personal assistants have been combining customization with cognitive learning. Their aim is to make sure that Siri, Alexa and Cortana refine their answers among the many possibilities, singling out the most relevant and pertinent one for the intent of each user.
For instance, if a user is looking for Paris Hilton, are they interested in information about the Hilton hotel in Paris, or the celebrity of the same name? Their search history, together with other signals, is the key to a personalized response which will most likely catch the true intent behind the user's question.
3) Data Science: In order to learn, a machine needs billions of pieces of data to compute and analyze.
At what levels does AI work?
AI evolves on three levels:
1) recognition of models and patterns: it is the most elementary level, where the machine processes great volumes of data and quickly identifies any connections.
2) perception: 99% of human perception happens through vision and words.
Artificial Intelligence is taking giant leaps in decoding language (Speech-to-text) and pictures (Image Recognition).
In the case of voice recognition, AI has achieved incredible results with a margin of error of 5.9%, the lowest ever attained. The next step is simultaneous translation from one language to another; in the case of Chinese, equal results with human translation have already been achieved.
3) cognition: in terms of AI, it is the capacity of human level thinking, that is when a machine is able to interpret the surrounding context and understand the real intentions behind a request from the user.
The democratization of AI
To get an idea of the value and positive impact AI can have in companies, just think about how algorithms make brand communication more effective and establish personal relationships with their consumers.
In Microsoft, we want to make this technology accessible and give everyone the freedom and ability to create their own Artificial Intelligence solutions. Our mission is based on the concept of empowerment, and we believe that by sharing our experience made of more than 30 years of research, our users will be able to find the solutions and answers to problems that we may not yet have imagined.
This means giving everyone this opportunity through AI for accessibility, including a billion people in the world who currently suffer from some form of disability, amplifying and enhancing their capabilities beyond human limits.