AI & Machine learning
World Summit AI 2017
- Open data,
Last 11 and 12 October, the Kernix Lab was in Amsterdam to be a part of the World’s Tech AI Summit thanks to the job done with Global Startup Weekend AI.
This event brought together over a hundred of the leading experts in artificial intelligence with more than 2000 of the world’s most active AI explorers. In this short blog post, we will report the progress made in the different fields.
The event included keynote presentations where high tech companies presented their latest projects, trending applications and perspectives on the AI field. Among those, Alibaba, the biggest e-commerce website in China, partnered with the government of the Macau special administrative region in a smart city project : the city brain project. One of the goals of this partnership is to tackle transportation issues by processing images coming from city cameras, identifying the reasons of traffic problems and finally optimizing the city’s resources. This is something that we also see here in Paris, where initiatives such as shared bikes and cars systems are a first step toward handling congestions that will be pushed further with the coming Olympic games. Indeed, in order to handle the increased traffic, the city is partnering with start-ups to better forecast traffic jams and decrease congestion with self-driving shuffles (source).
Unicef innovation team made a brilliant presentation, in which the speaker compared the determination of communities from administrative separations (citizens) and the ones created by social networks (netcitizen). This project sounds us like an internal project we have at Kernix, which enables the analysis of Twitter communities on specific themes (accounts, hashtags) and which could be used in a similar manner to connect the digital world to the real one.
Booking.com showed their machine learning applications, which encompass bidding optimizations, fraud detections but also recommendation systems. The need of recommendation systems is something we clearly see at Kernix with an increasing number of projects on recommendation systems (here one of our post on recommender systems based on graph).
Other amazing presentations in image processing were also shared. A Pintereset engineer presented his visual search method that helps users find images similar in terms of type of item shown in the image, style and colour attributes. Ebay showed his “shop-bot” that helps users find the best item with visual guidance and multi modal research (describe an item not only with words but also using images). And finally, a Nasa team presented their applications of image recognition for space exploration.
There were also workshops that helped people understand how to create machine learning applications and see in more details how these applications are implemented in the real world.Google presented a workshop about building a Tensorflow model from scratch and predict the digits from the images of the mnist dataset. The main reason for this workshop, was the promotion of their Google Computing services that state that not only the knowledge of machine learning should be democratized but also the computational resources to make calculations. The youtuber Siraj Raval also made a presentation where he presented cases where AI wasn’t always needed and alternatives optimization methods such as block chain can be used instead. At Kernix we especially focus on the pragmatism of our custom solutions and try not to use overly complex ones just in order to say “we did it with AI”…
Finally, the brightest minds from Amazon, Tencent, Accenture, IBM-Watson discussed about ethics subjects and limits of AI. Some conclusions from those discussions were that AI is a team sport and a multidisciplinary field where several subjects are combined. This was outlined by the suggestion that research labs and universities should have narrower collaborations with industry in order to create more realistic applications. At Kernix we make sure to maintain a close link with university by having PhD students in our team. Finally, Yan Le Cun closed the event talking about the state-of-art image recognition algorithms, the limitations of deep learning and also the perspectives for further development which were basically the incorporation of common sense to fill the blanks.
From the point of view of a Data Scientist, being part of this event has been an amazing experience. It’s very interesting to see how AI has been democratised and is used not only by the biggest companies but also by start-ups that catch the numerous business opportunities in this field. The major themes that we saw at the summit will help our team to demystify AI for the compagnies not used to it, to guide our clients to pragmatic goals and to develop better solutions to satisfy their needs and concerns.
Découvrez également l'article
AI & Machine learning
Gestion de la multimodalité dans les moteurs de recommandations
AI & Machine learning