Coffee: check. Pastry: check. Experts from globally leading AI and Deep Learning projects: check.
Yesterday morning we kicked off the Deep Learning Summit and AI Assistant Summit with Fabrizio Silvestri from Facebook and Nikola Mrksic from Cambridge University giving us an introduction to the busy and diverse schedule of the two days.
The goal of this event is to bring together people with diverse backgrounds and to foster discussions and interactions around DL, not just to listen - Fabrizio Silvestri
The event is made up of two tracks with a diverse array of speakers presenting their most recent work. The morning in DL explored the theory and landscape of deep learning along with business applications where we heard from Andreas Damianou from Amazon, Shubho Sengupta from Facebook AI Research (FAIR), and Jonas Lööf from NVIDIA. Andreas spoke in the morning about probability and uncertainty in deep learning, drawing on the importance of being able to predict an outcome regardless of the model having experienced the exact instance before. For example, Andreas said:
I managed to give my colleague a lazy eye even though he doesn’t have a lazy eye, but I managed to find the data to alter that - it’s the way of letting data to imagine something based on what you already know that’s important to drive these progressions.
Over on the AI Assistant track, Nikola Mrksic, compere for the day and Machine Learning Researcher at University of Cambridge summarised the landscape of personal virtual assistants in his presentation. As we are aware, the principal problem with voice assistants is their understanding (or lack of) and he highlighted this well:
‘Speech recognition has got really big! The AI Assistants needs to understand what the user is saying. At the moment, it can write down what you’ve said but not action that correct.’
Continuing the discussion and focusing on the importance of NLP was Vijay Ramakrishnan from Mindmeld. He shared his work in deep learning techniques for named entity recognition from an enterprise perspective, whilst Yariv Adan from Google Assistant explained how the age of personal assistants is very young and still has a long way to go. Although it feels like these products are super advanced and complex today, they’re in their early stages. Look at how far we’ve come so quickly, “mobile first changed our world in less than 10 years. In the next 10 years, we will skip to a world where AI is put first”.
After a busy morning of discussions, it was time for a coffee break where attendees, speakers and exhibitors had the opportunity to network, so we took the opportunity to catch up with new and returning guests to see how they were finding the summit so far:
It’s been fascinating, i’ve got loads of new ideas to take home with me - Daphna Idelson, GSI Technology
The last talk on speech to video and video speech in DL was so good it was crazy! - Michael Akintunde, Imperial College London
Brilliant ! Loads of great speakers! The balance between technical and applications is perfect! Janet Bastiman, Story Stream
I absolutely loved Andreas Damianou presentation on Probability and Uncertainty in Deep Learning. Manju Rangam, McLaren Applied Tech
After everyone was fed, watered, and had the chance to start their networking, the Deep Learning Summit continued with Marta Garnelo from DeepMind who spoke about how deep learning algorithms have achieved impressive results on a variety of tasks ranging from super-human image recognition to beating the world champion at the game of Go. Whilst representations are not at all new, before deep learning people were doing them by hand so the efficiency has come on unrecognisable amounts. We then heard from Fabrizio Silvestri from Facebook, and Sebastian Riedel from UCL:
As well as hearing from industry and research experts, the RE•WORK team have been busy behind the scenes interviewing speakers for exclusive talks and fireside chats on our Video Hub, as well as recording exciting new episode for our podcast, Women in AI. Elena Kockhina from University of Amsterdam, who discussed her work in debunking false news stories and rumours in social media spoke with our podcast host, Yaz, about the challenges in identifying falsehoods and how she’s using Machine Learning to overcome this. She explained the difficulty of automating the flagging of rumours.
Elena told us how Twitter only provides the opportunity for a tweet to have one sentiment label, it says that ‘a classification task of emotional polarity (on each tweet) is split into either positive, negative, or neutral - a tweet can only express one sentiment because it's short'. This isn't true, and Elena shows us examples of how multiple sentiments are crammed into a short sentence and discusses how to overcome this. Subscribe to the brand new Women in AI Podcast to hear Elena’s episode amongst others.
Back in the presentation rooms we heard from Adi Chhabra, Senior Product Manager - Artificial Intelligence at Vodafone who explained how ‘chatbots are awesome!’ and are hugely improving the customer service capabilities of their company. There are so many queries from bills, to data usage, to new phones - chatbots are the answer, however with the model they’re currently using, if you fall outside the categories that the bot understands, customer satisfaction isn’t achieved - Adi explained how they’re working to overcome this with their deep learning model.
The conversation of personality and psychologically aware AI continued throughout the afternoon covering topics such as emotional AI and ethics, trusting AI devices, as well as end to end testing for virtual assistants.
Covering natural language processing in more depth, Nikolaos Altreas, Applied Scientist at Amazon shared some of his most recent work on labelling topics using both images and textual inferences. He explained that the importance of topic labelling in producing ‘automatic methods for giving short and concise labels. In text it can be a short phrase, or these topics can be represented with images.’
Whilst we were engaged in discussions, the team were busy recording more interviews, and Sam Shead from Business Insider spoke with Davide from Facebook about his most recent work and progressions in feep learning.
We wrapped up the day with some exciting discussions about deep learning in space with Digital Globe revealing how they are planning to sweep the whole globe once every 20 minutes and analyse it with deep learning. The end of the day saw attendees, speakers, and team members coming together for the Deep Learning Dinner in County Hall where we enjoyed an evening of networking, a champagne reception, and a three course meal and wine - a great end to the day!
We heard about some amazing technological advancements and groundbreaking progressions yesterday, and are incredibly excited tfor what we have to hear today from more AI and DL experts.
For now, we’ll leave you with our quote of the day from Fangde Liu from Imperial College:
Think of AI not as experts, but as interns. You can have millions, but you have to teach them to cooperate and learn team building.
Missed out on the Summits in London?
We’ll be back in Montreal with the Deep Learning Summit this October 10 & 11. There are limited tickets remaining, so register now to guarantee your place and learn from global pioneers including Geoffrey Hinton, Yann LeCun, and Yoshua Bengio.
23 January 2018, San Francisco
Leading minds in machine intelligence will come together for an evening of networking and keynote presentations. Join us for a three course meal to support women in AI and Machine Intelligence.
25 January 2018, San Francisco
The next generation in predictive intelligence. Anticipating user & business needs to alert & advise logical steps to increase efficiency. The summit will showcase the opportunities of advancing trends in AI Assistants & their impact on business & society. What impact will predictive intelligence have on business efficiency & personal organization?
25 January 2018, San Francisco
The Deep Learning Summit is the next revolution in artificial intelligence. The increasingly popular branch of machine learning explores advances in methods such as image analysis, speech and pattern recognition, natural language processing, and neural network research. This summit will explore how deep learning algorithms and methods are being applied to solve challenges in industries including healthcare, manufacturing, transport, security and communications.