Emotech, the masterminds behind the 'robot with a personality' Olly, have partnered with Cocoon Networks to create this brand-new Meet AI series in London. This monthly event series will provide you with insider knowledge and insights into AI.
The events will be delivered by researchers in different fields of AI, robotics hardware engineering, and product design. You will have the chance to explore cutting-edge AI theory, learn how to apply theory to practice, experiment and embark on steps toward developing your own AI ideas. Not only do we aim to make AI more widely understood, our mission is to kick-start a vibrant international AI community in London!
We are delighted to invite you to our first ‘Meet AI’ event focusing on Reinforcement Learning and TTS, together with our 2 distinguished speakers:
Pedro Mediano (Research Scientist at Emotech Ltd.):
In the early days, and forced by an evil mathematician father, Pedro started a physics degree at University of Valencia, Spain, truncating what could have otherwise been a promising career as a jazz melodica player. Not satisfied with the punishment of an undergrad in physics, he moved to the different, but marginally more useful field of artificial intelligence and computational neuroscience. At the tender age of 20, he had worked as a research assistant at Harvard, Carnegie Mellon and the machine learning team behind Wolfram Alpha, knowledge engine used by Siri. He somehow sneaked into a PhD program at Imperial College, London, where he found that artificial intelligence might be artificial, but is far from being intelligent. Now, working as a machine learning scientist at Emotech, he is doing his little bit to change the last part. He thinks that writing in third person looks cool.
Rasmus Dall (TTS Researcher at Jibo, Inc.):
Growing up surrounded by hippies on the countryside in Denmark, it was to be expected that Rasmus sought to do a degree in Philosophy at the University of Edinburgh. Philosophy became Mind and Language, which became Cognitive Science before he started a PhD in Speech Synthesis at the same institution. The slippery slope had begun, and with time spent at Google's TTS team in London and Nagoya Institute of Technology in Japan during his PhD it was clear. He had joined the dark side. Nevertheless, at least he was mainly trying to make computers say UM and UH in just the right way as to sound more human. Now, working on speech synthesis R&D at Jibo, it is about much more than that.
18:00—18:45 Registration & Arrival drinks
18:50—19:25 Pedro Mediano- Reinforcement Learning: Recent Progress and Future Problems
Reinforcement Learning (RL) is a field within AI that makes machines learn by trial-and-error. RL systems have improved enormously in the recent years, being able to master games like Space Invaders or Go. However, we have many challenges ahead: How can we make RL systems that live in the real world, with real people? How can we make sure these systems remain stable and well-behaved? In this talk we will explore these challenges and suggest a few ways forward to bring useful RL technology to people's daily lives.
19:25—20:00 Rasmus Dall- Character in TTS
Text to speech synthesis systems today sound very realistic, with very few artifacts and minimal signal distortion, however with little expressivity or character. While systems are certainly capable of producing a wide range of emotions and speaking styles, enabling rich character driven experiences, these are seldom, bordering to never, used in commercial applications due to one simple problem. They require human manual design. I will not propose a solution to this, but rather aim to state where the field currently is, and make a case for a more concerted effort into automating the use of human conversational elements in speech going beyond simply reconstructing the original signal as faithfully as possible.
Look forward to seeing you there!
DATE AND TIME
Thu 11 May 2017
18:00 – 21:00 BST
Cocoon Networks London
4 Christopher Street