NeurIPS 2019: AlphaNPI presented on the main stage in front of thousands 

In December, six InstaDeep colleagues travelled to Vancouver to present our latest cutting-edge research at the world’s most prestigious AI conference, NeurIPS.



AlphaNPI, the joint InstaDeep-Google DeepMind research published June 2019, was last Autumn awarded a spotlight talk at the conference, being one of the top 2.4% of submitted presentations. This meant it was selected to be presented on the main stage in front of an impressive crowd, a humbling achievement testament to InstaDeep’s capabilities, and it was not lost on the team just how big an honour it was to share the stage with the top AI people the world over!  

Enhancing Deep RL 

As InstaDeep’s lead on AlphaNPI, AI Researcher Thomas Pierrot had the pleasure of presenting on the conferences’ fourth day. Ballroom A+B counted thousands of people in the audience, and without hesitation, Thomas presented the research and how it aims to enhance current Deep Reinforcement Learning methods with a focus on the methods that incorporate planning. He explained our approach, assumptions, experiments and results, as well as the algorithm and how the team adapted AlphaZero to handle compositionality, receiving glowing feedback.  

Following the talk, the team proceeded to present the work in a poster session, again attracting herds of people asking precise and insightful questions including several researchers from Google Brain and Google DeepMind.

About the experience, Thomas said; “I really enjoyed this experience. Vancouver is a beautiful city with a nice atmosphere, and the conference was great! We had the opportunity to meet and exchange knowledge with the best AI researchers and engineers in the world. It was also awesome to realise that our work was really appreciated and that InstaDeep research and engineering are at world top level.

World-leading Deep RL

As the most renowned AI Research conference across the globe, NeurIPS attracts the world’s top talent and the true stars in their fields. This created a unique opportunity for the team to meet and interact with the best AI researchers and engineers, including David Silver (AlphaGo), Richard S. Sutton (Basis of modern RL) and Oriol Vinyals (AlphaStar).

It is incredible to see our work receiving recognition on a world stage and to see InstaDeep’s research and engineering contributing to the development of Deep RL. The team travelled home full of new ideas on how to further contribute to the future of AI in industry and to keep placing InstaDeep on the AI world map.