NLP and Deep Learning in Japan

6:30 pm - 7:00 pm (JST)

​

Transformers

Thomas Wolf, Co-Founder and CSO at HuggingFace

​

xxx​

​

Speaker info

​

​

7:00 pm - 7:30 pm (JST)

​

Large language models in production: opportunities and challenges applying AI to patents

Sam Davis, CEO, Amplified AI

​

At Amplified, we use deep learning to quickly and efficiently compare new inventions and products to 140 million patent documents. Building around large language models presents unique and exciting opportunities for working with patent data but at the same time, working at scale from day one as a startup brings with some unique challenges. We’ll share why we started with deep learning and some of the successes and failures in our journey building an AI-first product and getting customers worldwide that include government offices, universities, and Fortune 500 companies.

​

Speaker info

Sam Davis is the co-founder and CEO of Amplified. Since 2015, Sam has been an angel investor and advisor to deep learning focused AI start-ups. Before that, he led global business development for a leading patent research company and is a recognized speaker on AI’s impact on the future of IP work. Outside of Amplified Sam is a loving father of two and enjoys learning foreign languages, playing guitar, and writing mediocre songs that are probably best kept to himself. 🎸

​

👉 Changing the patent system for good

​

​

​

7:30 pm - 8:00 pm (JST)

​

Good ML UX needs "AI"

Chris Gerpheide, CTO, Bespoke

 

For many machine learning scientists, improving the system means optimizing model performance. But, in many real-life applications, a decent model alone often doesn't imply a good user experience. In this talk Chris will give examples of how we combine ML/DL with non-ML strategies to create a machine learning product that's successful in the wild.

​

Speaker info

Chris leads the engineering team at chatbot startup Bespoke in Tokyo. She's passionate about software quality and using machine learning where it matters (and not where it doesn't). Previously she was an engineering manager at AWS. 

​

👉 Find Chris on Twitter 

👉 Bespoke 

​

​

​

8:00 pm - 8:30 pm (JST)

​

Politeness-aware Machine Translation for Japanese

Anja Austermann, Software Engineer, Google Japan

​

The talk gives an overview of our experiments to generate Japanese translations with different levels of politeness. The Japanese language has a complex system of rules to express politeness in interpersonal communication. We trained neural machine translation models to take the user’s desired politeness level into account when translating into Japanese.​

​

Speaker info

Anja is a Software Engineer on Google’s Translate team in Tokyo and works on machine translation. She was born in Winterberg, Germany and studied Computer Science in Aachen and Paderborn before moving to Japan in 2006 to join a PhD program and do research on human-robot interaction. Outside of work, she enjoys tinkering with arts, crafts and electronics. 

​

​

​

​