What's New in H2O, Challenges in Czech NLP & Legal Liability Aspects of AI

Dear Makers,

H2O.ai is one of the gold partners at ML Prague. Please come and check out our machine learning platforms if you are planning to attend the main event (23rd and 24th Feb).

We would like to organise a warm-up event at our office on 21st Feb (the day before the conference workshops). Please come and find out more about the ML Prague conference from the organizer. We also have three other talks scheduled for the evening. We are hiring in Prague and other cities (https://www.h2o.ai/careers/) so please come early and contact our recruiter Petra (https://www.linkedin.com/in/petra-stola%C5%99ov%C3%A1-980625b/) to find out more.

Agenda:

18:00 - 18:30 - Doors Open. Refreshments + Networking
18:30 - 18:40 - Welcoming Remarks by H2O.ai team
18:40 - 18:50 - Introduction to ML Prague by Jiri Materna
18:50 - 19:10 - Talk 1: New Features in H2O.ai's ML Platforms by H2O.ai team
19:10 - 19:40 - Talk 2: Text Classification: Challenges posed by the Czech Language by Jindra Lacko
19:40 - 20:00 - Talk 3: Legal liability in the context of AI: EU perspective by Olga Slipetska
20:00 - 21:00 - Networking

=====

Talk 1: New Features in H2O.ai's ML Platforms by H2O.ai team

The H2O.ai team
- Jakub Háva (https://www.linkedin.com/in/havaj/)
- Pavel Pscheidl (https://www.linkedin.com/in/pavel-pscheidl-19b15990/)
- Jo-fai Chow (https://www.linkedin.com/in/jofaichow/)

=====

Talk 2: Text Classification: Challenges posed by the Czech Language by Jindra Lacko

Abstract:
Current best practices of text classification using Keras / TensorFlow (word embeddings and LSTM layers) are summarized. Challenges to this approach posed by the specifics of the Czech language are discussed, with possible remedial actions. A practical example is demonstrated classifying text from Twitter timelines of real Tomio Okamura (Czech xenophobe of Japanese descent) and a parody account operated in his name by an unknown prankster. After the presentation slides and source code in R will be available at http://github.com/jlacko/ml-meetup-text

About Jindra:
Jindra is an independent consultant in the field of BI and advanced analytics.

=====

Talk 3: Legal liability in the context of AI: EU perspective by Olga Slipetska

Abstract:
With the AI technologies rapidly advancing and retaining a higher degree of discretion for the purposes of autonomous decision-making, attribution of liability for AI faults has become an important topic calling for immediate attention of the EU regulators.

For now, this segment is regulated by the Directive 85/374/EEC that aims to strike a balance between consumer rights and a business need to market innovative products. If the AI component is embedded in a product that causes damage to an individual, the liability for that damage will sit with a manufacturer of the product in question. In this scenario, the chain of causation is key in determining the proportion of liability. For instance, if the consumer negligently or otherwise contributed to the incident the manufacturer may bare a smaller proportion of liability and, as a result, will be required to pay a reduced compensation.

About Olga
Olga is a lawyer based in Prague whose expertize is focused around corporate governance, data protection as well as legal aspects of immersive technologies. Currently, she holds a position of the Corporate Legal Manager at Honeywell and prior to that worked as a Legal Advisor for the Czech subsidiary of Oracle. In parallel, Olga acts as the Of-Counsel for the local law firms pursuing ICT and TMT agenda. Olga obtained degrees in Law from the University of London and Queen Mary University of London with specialization in International Alternative Dispute Resolution (Arbitration).