Twenty sixth of June it was so far the first Chatbot summit in Europe. And even better it was in my favorite city Berlin. In short 2 good reasons to go.
In the openings keynote from Yoav Barel, the CEO of Chatbot summit, made clear the future of chatbots are to become the personal assistants. Assistants for health, banking, travel and everything you can imagine. Peer Bentzen from Deutsche Post confirmed it in the panel discussion. Peer is even a heavy user of an health assistant.
My highlights from the Summit (mainly focused on Development & AI a little less on UX.
As soon as it works, nobody calls it AI anymore
AI or Artificial Intelligence, the term is used and abused allthough the definition is clear and probably different from 1956. So as soon as we will not hear about AI in chatbots anymore it means AI is working.
The hybrid approach: Unsupervised pre-learning combined system actions defined by humans
Björn Gülsdorff – Creative Virtual
This the approach we are using within ContentForces in combination with a human in the loop (see below).
The Hybrid Model nr 1
A human in the loop
Chris van Pelt – Crowd Flower
A nice example with the Covfefe tweet from Donald Trump
But in the end this is the model to start and improve. Human interaction, directly used a training.
The Hybrid Model nr 2
Unsupervised pre-learning for pizza toppings
Jairson Vitorino CTO from E.life
An nice example of the The Hybrid Model nr 1
Cold start the the Natuaral Language Understanding AI
Lack of data forces Chatbots to start rule based, to coldstart. Multilingual is even a bigger challenge. For example in Finnish every word can have many forms which increase the dimensionality (deep learning).
5 million users for an Indian dating chatbot
Haptik by Aakrit Vaish
It was all about UX, and Aakrit mentioned that they feed the bot with interesting content and the users react on it. I am wondering if the behavior of European users would be the same.
Change the habit to use the chatbot.
Meekan by Doodle (Google)
Interesting UX, between more convenient and straightforward there is a big challenge to change the behaviour. In this case to move users from (Google) Calendar to Meekan.
Natural Language Processing on Steroids
And Karsten Schmidt from SAP Labs put it technically nice to together.
If you are interested to share your idea’s or hear more about my vision, just contact me via the form.