Keynote: Breaking black-box AI

Machine learning and artificial intelligence are becoming wide-spread and productionalized - you no longer need a mathematics PhD and months of software development time to implement and use a machine learning algorithm. You can just call an API and you get the answer! You can treat these systems completely as black boxes and use them directly in your applications! But beware - all the algorithms have some cases when they fail to deliver what you're expecting.

This talk is packed with live demos that show failure cases of popular algorithms, from linear regression to cutting-edge deep learning. We will look at practical examples, use standard algorithms as black boxes and observe when they fail and why. You will learn that although you can treat the algorithms as black boxes, they can still fail silently and what to do about it.

 

Speaker

 

Evelina Gabasova
Evelina Gabasova is a machine learning and data science researcher. She works as Principal Research Data Scientist at The Alan Turing Institute, the UK's national institute for data science and artificial intelligence. She is a member of the research engineering team where she is connecting academic research with real-world applications. Her passion is to make data science understandable and accessible to everyone. When not writing software, wrangling data or training machine learning models, she is an active member of the F# open source community, Microsoft MVP and a technical speaker.

Gold-Sponsoren

HMS
Structr

Silber-Sponsoren

codecentric
Phytec
*um

Bronze-Sponsor

incontext.technology GmbH

data2day-Newsletter

Sie möchten über die data2day
auf dem Laufenden gehalten werden?

 

Anmelden