Interview with Thomas Wolf (Hugging Face) | Data Science fwdays’20.
The keynote speaker of the Data Science fwdays’20 online conference Thomas Wolf — Co-founder and Chief Science Officer at Hugging Face — gave answers to our interview and told us about his childhood, work at Hugging Face, and his way in NLP.
- Tell us a bit about yourself. Who did you dream to become in childhood?
I’ve spent my childhood in a very small village in the French countryside, so I was not really thinking about the future, more playing around. That’s probably why I ended up doing so many things before settling on ML and NLP (engineering, physics research, patent attorney).
2. How did you first get involved in Natural Language Processing?
I discovered NLP pretty late when I was a patent attorney a few startups were starting to do Deep Learning. I got very interested in this which reminded me of my physics equation and started to dive deeper into this, reading books and courses on the topic. Then at some point, we decided to follow the class of Richard Socher at Stanford with Julien Chaumond and then started to build HuggingFace together.
3. Which problems are you solving at Hugging Face?
We are catalyzing and democratizing NLP. We think NLP is going to be the most transformative technology of the decade, and we want to make this vision a reality by doing a combination of open-science and open-source to develop and share the latest advances in the field. We think NLP breakthroughs should benefit everyone and be accessible to the whole community.
4. What are you going to talk about at the Data Science fwdays‘20 online conference?
I’ll start by an introduction on the revolution of Transfer Learning in NLP, talk about the limits we can see, and then, in a second part present our tools and do a quick hands-on example using all of them.
5. Which research and development activities and industry use cases you’ll be keeping an eye on over the next year?
NLP and it’s an extension to knowledge-base models and multi-modal models.
6. Do you have any advice for those who are looking into building an NLP algorithm?
You should probably use HuggingFace tools :-)
7. It seems that you read a lot of books and materials about NLP. If you are going to write a book, what exactly will it be?
I’ve been asked many times to write a book about Transfer Learning in NLP but writing a book is really a lot of work :)
8. In your opinion, what is the most useful application for society, that is using natural language processing?
That’s a tough question, in particular since I don’t really have a specific sub-field to promote, our models at HuggingFace are used in so many different applications. I think the most useful applications are where NLP is used to complement and augment humans rather than replace them.
9. What are your recommendations for beginners who are starting to learn NLP?
Think about ethics. NLP and ethics go hand in hand.
We invite you to listen to the talk of Thomas called “An Introduction to Transfer Learning and Hugging Face” during the free Data Science fwdays online conference on August 8 & 15. Registration.
