Facebook AI Innovation Summit

Building a better world with AI

A celebration of six years of Facebook AI innovation in Europe

Six years ago, Facebook AI Research (FAIR) Paris was created to advance the state of the art of AI in Europe through open research for the benefit of all. On the occasion of its sixth anniversary, FAIR Paris brings its European partners together to celebrate this moment during an online conference around its leaders, researchers, and institutional partners. We will revisit the lab’s key achievements, current research interests, and look towards future challenges around the theme “Building a better world with AI”.

Please join us on June 30th from 2:00 PM to 6:00 PM CEST.

Register here for the event

Featured Speakers

Join our Facebook AI leaders as they deliver keynotes on the future of FAIR Paris research


All times listed are in Central European Summer Time

2:00 - 2:05 PM


Pierre-Louis Xech, Research Program Manager, Facebook AI, Master of Ceremony

2:05 - 2:15 PM

FAIR in Paris and beyond


An overview of the last year at FAIR and the recent achievements that have been made possible by our Facebook AI team.
- Antoine Bordes, Co-Director of FAIR Accel


2:15 - 2:30 PM

Discovering novel drug combinations to treat diseases


Finding new ways to repurpose or combine existing drugs has proven to be a powerful tool to treat complex diseases like cancer and provide more personalized treatments for patients. Recognizing this, Facebook AI and Helmholtz Zentrum München built the first single AI model that predicts the effects of drug combinations and dosages with the goal of accelerating the process of identifying optimal combinations of drugs and other interventions that could lead to better treatments for complex diseases.
- Fabian Theis, Helmholtz Research Center in Germany
- David Lopez-Paz, Research scientist, FAIR Paris

2:30 - 2:35 PM

Accessibility with Cap’Falc


Every day, we are inundated with information. And much of it remains incomprehensible to people with learning disabilities. Unapei, Inria and FAIR have joined forces to develop a tool called Cap’FALC under the patronage of the French Secretary of State for the Disabled. By facilitating transcription through an ergonomic interface and an innovative algorithm, Cap'FALC aims to significantly improve access to information for all and to promote the long-term participation of people with intellectual disabilities in society.
- Louis Martin, Research scientist, FAIR Paris

2:35 - 2:50 PM

Using AI to forecast the spread of COVID-19


Facebook AI has built an adaptive model and collaborated with experts to help the world better understand the spread of COVID-19. The COVID-19 forecasts produced by the models gives researchers and public health experts information that can help them with resource planning and allocation and early outbreak detection. These forecasts are developed using public, non-Facebook data, and serve as a tool to support our global efforts to keep people informed as the pandemic evolves.
- Enric Alvarez, professor, UPC
- Maximilian Nickel, Research scientist, FAIR New York

2:50 - 3:00 PM

Using Data for Good and AI to combat COVID-19


The Facebook Data for Good team has been partnering with nearly 500 institutions across 70 countries since the onset of the COVID19 pandemic to leverage real-time insights from the Facebook platform to improve COVID19 containment and response. In Europe, these collaborations have informed mask mandates in Poland, re-opening strategies in the UK, and are now helping inform targeted strategies for vaccination outreach.
- Alex Pompe, Policy Research Manager, Data for Good

3:00 - 3:10 PM

Questions and Answers


3:10 - 3:25 PM

Expire-Span: Teaching AI to forget


Our brains have evolved over thousands of years to forget most of the information that is not relevant to us. However, machines so far have struggled to figure out what information to keep around and what to discard. Therefore most AI models to date have had to memorize information without distinction. This is where Expire-Span comes into play: it helps AI forget by adding an expiration date to information. Once that time has passed, that bit of information is gradually decayed and eventually forgotten from an AI model’s memory. This forgetting mechanism allows the important memories to stay for a long time, forgetting the unneeded ones.
- Sainbayar Sukhbaatar, Research scientist, FAIR Paris

3:25 - 3:40 PM

Image Understanding Based on Text


Getting our machines to behave closer to how humans do implies giving them a more natural and free way of communication, one that can capture the multiple ways of communicating a message. For example, humans use multi-modal understanding - they extract insights from a combination of sources like text, image, audio, and more. This work proposes to capture visual concepts expressed in free text form by reasoning jointly over text and images.
- Gabriel Synnaeve, Research scientist, FAIR Paris

3:40 - 3:50 PM

Questions and Answers


3:50 - 4:00 PM

Self-supervised learning: The dark matter of intelligence


How can we build machines with human-level intelligence? There’s a limit to how far the field of AI can go with supervised learning alone. Self-supervised learning is one of the most promising ways to make significant progress in AI and helps machines develop a form of common sense. This common sense ability is taken for granted in humans and animals, but has remained an open challenge in AI research since its inception. This is why common sense is the dark matter of artificial intelligence.
- Yann LeCun, Chief AI Scientist, Facebook

4:00 - 4:15 PM

Introducing SEER, a self-supervised billion-parameter computer vision model


The future of AI is to create systems that can learn directly from the information provided to them, whether it is text, images or other types of data, without relying on labelled datasets. SEER (SELf-supERvised) is a new, large-scale computer vision model that can learn from any images on the internet. his model uses a technique called self-supervised learning, which means it can learn from images without the need for labelling as is the case with most computer vision training today.
- Priya Goyal, Software engineer, FAIR New York

4:15 - 4:30 PM

Chasing carbon: Reducing the environmental footprint of computing


As computer hardware and software become more powerful, they also increase their energy demand. This is particularly true when it comes to the hardware that runs advanced AI and machine learning applications and trains deep learning models. So, what can be done at the capex and opex level to reduce emissions?
- Carole-Jean Wu, Applied Research Scientist, FAIR Boston

4:30 - 4:40 PM

Training home assistants to rearrange their habitat

Demo / Presentation

Imagine walking up to a home robot and asking “Hey robot – can you go check if my laptop is on my desk? And if so, bring it to me”. Or asking an AI assistant (included in your smart glasses): “Hey – where did I last see my keys?”. Developing such intelligent systems is a goal of deep scientific and societal value. Training and testing such agents directly in physical environments is slow, expensive, and difficult to reproduce. We will present the next generation of our simulation platform for training virtual robots in interactive environments and complex physics-enabled scenarios.
- Oleksandr Maksymets, research engineer, FAIR Menlo Park

4:40 - 4:50 PM

Questions and Answers


4:50 - 4:55 PM

Why AI is important for Facebook and what it brings to the table


For more than a decade, Facebook has been making use of groundbreaking AI in its products. But what is the impact AI is having on Facebook’s apps today and what does the future hold for its products? Learn why AI is vital to building new experiences at Facebook, such as Remote Presence, AR, Integrity, and Commerce.
- Srinivas Narayanan, VP of Engineering, Facebook

4:55 - 5:15 PM

AI tech for commerce


The future of commerce is social. Facebook AI researchers and engineers are teaching machines to understand the different items in a wardrobe or an apartment, how a garment relates to an accessory, and how an online product might look in real life — and doing this for millions of people around the world. We have built cutting-edge AI techniques and systems such as GrokNet to understand precise specifics about what’s in nearly any photo. We’ve also built technology that can automatically turn a 2D phone video into an interactive 360-degree view. We’re now one step closer to our vision of making anything shoppable while personalizing to individual taste.
- Tamara Berg, Research scientist, FAIR Menlo Park
- Ana Grace, Director of Product, Facebook AI

5:15 - 5:35 PM

Questions and Answers

Read the latest publications from featured FAIR Paris researchers

June 01, 2021



The FLORES-101 Evaluation Benchmark for Low-Resource and Multilingual Machine Translation

Naman Goyal, Cynthia Gao, Vishrav Chaudhary, Peng-Jen Chen, Guillaume Wenzek, Da Ju, Sanjana Krishnan, Marc’Aurelio Ranzato, Francisco Guzmán, Angela Fan

June 01, 2021

May 14, 2021


Not All Memories are Created Equal: Learning to Forget by Expiring

Sainbayar Sukhbaatar, Da Ju, Spencer Poff, Stephen Roller, Arthur Szlam, Jason Weston, Angela Fan

May 14, 2021

July 26, 2019


Strategies for Structuring Story Generation | Facebook AI Research

Angela Fan, Mike Lewis, Yann Dauphin

July 26, 2019

Help Us Pioneer the Future of AI

We share our open source frameworks, tools, libraries, and models for everything from research exploration to large-scale production deployment.