Early this December, we were treated to the marvelous AI congress, organized by our friends We Are Developers in Vienna. Putting aside the Weinacht feeling of having a cup of Glühwein in Vienna night, the marvelous architecture of the ballrooms-turned-conferencehalls at the Hofburg (the congress venue), and the intense congress after party in the night club (all of which deserve a separate article), I would like to point out the top talks from the congress.
While there were a lot of other nice talks, workshops, and expo, I have pointed out the most interesting ones from the data science applications point of view. Enjoy!
(Bonus) Cracking the code: neuronal networks in the brain | Sabria Lagoun
It is regular for those type of lists to count the top 10 and give a separate bonus pick. Well here is different and not only because this was one of the most fascinating talks of the congress, but also it shaded new light or (I should say) reminded us of the original light on how really the brain works. Biggest implication – a simple reminder that the artificial neuron networks are in reality a technology 50 years old, and they have not captured yet (as of today) the complexity and self-organizational nature of a real, living brain.
1. Snap ML: a highly-accelerated, scalable software library for machine learning | Haris Pozidis
A famous name among researchers and practitioners, the lecturer presented the best new tech at the congress – a free-to-use soon-to-be-open-source library by IBM, for rapid deployment and computation of generalized linear models over GPU and multi-processing framework. The results which were shown are way ahead of any deep learning or machine learning application in terms of speed of computation, but also in terms of accuracy. Again it is time to remind us an age old idea of better using multiple linear models instead of non-linear ones to model complexity.
2. Making machine learning accessible using Keras | Pavithra Vijay
Speaking of libraries – this talk has given us an excellent overview but also practical demos on implementing Google’s Keras. The lecturer went into details and was obviously captivating the audience.
3. AI and Advanced Analytics applied at Raffeisen Bank International AG | Lubomir Karlik
Here I prize the talk not so much for the technology used but for the fact that the guys at a commercial bank (typically very conservative market) are showing prime interest in developing AI to automate their services. Also it is of note that for the needs of real world business English only speech and text recognition libraries are far from enough.
4. On the importance of explainability | Holzinger
As soon as the banks go into AI business, we will have a new set of requirements on explainability of the results. The talk is what we should consider doing beyond making the correct prediction and how is this important in relation to real business services.
5. Augmented Intelligence – A Marriage between Machine and Human | Simon Stiebellehner
Speaking of explainability, the next talk here is about human interaction with the results from a AI model in two different directions – how to interpret the results (considering GDPR’s requirements for explainability) and how to improve the results using human input at the end of production.
6. How to be best in conveying the message by using NLP approach (Big Five modeling) | Tomislav Krizan
This talk moves us smoothly from personal perception to percepting a person through analysis of various psychological traits, using powerful NLP techniques. The lecture as apprized also for the easy going and pragmatical style of the lecturer.
7. Deep Learning for Recommender Systems | Alexandros Karatzoglou
The aforementioned NLP techniques are being used in wide spectrum of fields and this talk gives us a feeling of four different applications in recommender systems. The side perk of the talk is that the lecturer is a head of a research team and has obvious academic background.
8. Building a Large Music Recommender Leveraging AI, Deep Learning and Human Expertise | Òscar Celma
Staying with the recommenders here is yet another head of research who also possesses southern temperament. The business of Pandora is one of the biggest in the music streaming industry and naturally they have specific insights (structured in 5 rules) for recommendation.
9. Bridging the gap between AI and UI | Liad Magen
One of the main culprits for the existence of the AI congress and also great expert on NLP, gives this talk (final one for the whole congress) as a form of generalization of the whole event. The AI-to-human interaction is the front topic here.
10. The hand prosthesis that learns from the user | Sebastian Amsüss
Speaking of user interface, there is hardly any more important at the moment than in the filed of mechanized limbs. This talk is on automation of the process of adoption a prosthetic arm and was the best one out of the whole congress on robotics.
So here it is – my own take on the top talks from the AI Congress. Greatest regards to We Are Developers for making it happen.
P. S. Yes, we were the ones getting the creative award for making this nerdy laptop (from sticks and stickers with its own python code etc.) and later owning the dance floor for hours.