Axel Springer AI - the past, present and future

Slides of my talk at the the German-Chinese Association of Artificial Intelligence 2nd Annual Meeting 2019 in Berlin where I talked about my plans to turn Axel Springer into an AI-first company.

Transcript

1. Dat Tran - Head of AI @ AS Ideas Engineering (@datitran) 15 june 2019 ~ berlin ~ Axel Springer AI the past, present and future
2. echo $(whoami)
3. echo $(whoami)
4. echo $(whoami) https://www.volkswagen-newsroom.com/de/pressemitteilungen/volkswagen-konzern-it-setzt-bei-digitalisierung-auf-neue-recruiting-und-arbeitsmethoden-542
5. echo $(whoami) Face2Face Hydroplaning Prediction Image Quality Assessment
6. echo $(whoami)
7. The past
8. That’s us
9. In the early 2000s Total circulation of newspapers in Germany 2003-2019* * Statista 2019
10. organic 2007 2008 2010 2009 2011 2013 2012 2006 2014 2015 2016 2017 2018 Organic growth and acquisition by acquisiton
11. The present
13. 74% of revenues and 87% of adj. EBITDA from digital activities
14. Leading digital publisher and digital classified worldwide
15. Over 250 brands
16. A family of many strong brands
17. approx. 16k employees worldwide
18. Media & Tech company
19. The Problem: "Software is Eating the World, but AI Is Going to Eat Software" Jensen Huang (CEO of NVIDIA)
20. My top three apps Apple Photos Google Maps Spotify
21. What do these apps have in common? They are smart...
22. Smart Photo Search
23. Traffic Prediction and ETA
24. Personalized Music Recommendations
25. Morale of the Story
26. #1: User these days expect products to be smart
27. #2: AI is only part of the entire product: you also need great software, great user experience, great business model etc
28. #3: Creating an AI culture is important to create the foundation for smart products
29. Current State of AI@Axel Springer
30. #1: A lot of companies with no or some strategy/vision for “smart” products
31. #2: Team sizes ranges from 1-2 people to larger teams
32. #3: No “standardized” AI strategy
33. #4: “PoC” with deep learning projects and only a few in production
34. #5: We’re really good at making “buy” decisions but not how to “make” it
35. #6: A lot of churn; people leave Axel Springer and their units because they can’t do cool stuff
36. Our Solution: Axel Springer AI
38. Vision
39. Turn Axel Springer into an #AIFirst company which will shape the way how we make product tomorrow differently
40. Mission
41. #1: We make AI more accessible within Axel Springer and hence drive innovations within the group
43. #2: We deliver end-to-end machine learning projects and establish best practices
46. Computer Vision Natural Language Processing Recommendation Service Axel Springer AI Cloud Services
47. Footer 26 April 2019
48. Extreme Programming Practices for Data Science Pair Programming CI/CD - API First Test-Driven Development Lean startup
49. Technology Stack PyData Deep Learning Big Data Computer Vision NLP Production Machine Learning Visualization Data Preparation
50. #3: We conduct state of the art AI research
51. Applied Research • Why: fundamental research is important to stay innovative as a company • Regular knowledge sharing with leading universities and also tech companies like Amazon, Google, Tencent, Zalando etc.. • Long-term goal: collaboration with research institute ◦ Research focus on NLP, Computer vision problems: ▪ Explainable AI ▪ AutoML ▪ Compression ▪ Edge computing ◦ Shared partnership: e.g. Axel Springer provides data, infrastructure and people ◦ Leverage government grants (EU, Germany)
53. https://letsenhance.io/
54. #4: We set and communicate the AI branding strategy for Axel Springer
55. Sounds nice? What now?
56. Will start with a central team...
57. AI Use Cases@Axel Springer News Media • Content Recommendation • BI/Churn • SEO title generation • Text generation • NER Visual Meta • Image aesthetic • Product classification • Image aesthetic • Product classification • Recommendation • Recommendation Classifieds • Instant job match • Profile completion • NER • Pricing/Price prediction And many more use cases...
58. Initial steps - Apply use cases to other units Hotel image ranking for both aesthetic and technical quality Low-to-high resolution Product catalog categorization
59. The future
60. Outlook • Machine learning is ingrained in every unit of Axel Springer (i.e. in production) • More “make” than “buy” decisions • Product manager/owner with a better understanding of machine learning • Axel Springer AI cloud service • Research lab
61. Questions? url: www.dat-tran.com Twitter: @datitran