On Monday 11 November 2019, the Data Science Initiative and UNICRI Centre for Artificial Intelligence and Robotics ran a workshop on Deep Fakes and AI Video Manipulations. The workshop was a follow-up from DSI’s Hackathon for Good (Held in June 2019, The Hague) in which we challenged participants to create tools for the detection of manipulated videos that can support law enforcement, security agencies, courts, the media etc. to readily verify the authenticity of image and video.
The workshop was divided into two sessions:
- In the morning session there were presentations from:
- Jheronymus Academy of Data Science, UNICRI, Europol, NATO Communications and Information Agency (NCI Agency)
- The Dutch government: The Ministry of Justice and Security
- The afternoon sessions:
- Presentations of the winning and runner-up hackathon teams’ solutions
- Hands-on workshops that looked at both the value proposition of deep fake detectors, and the technology behind them.
The workshop included 40+ stakeholders from international organisations, government, the private sector and academia. Kate Dodgson, the organizer of the workshop: “It’s important to bring these diverse actors into a room together to discuss this topic, as otherwise work carries on in silos. This workshop enabled sectors and organisations to hear about one another’s work, learn from each other’s best practice and lessons learnt, and helps to reduce replication of work while encouraging cross-sector collaboration.” According to Irakli Beridze, Head of the UNICRI Centre for AI and Robotics: “UNICRI and the Data Science Initiative have worked hard to ensure that this collaboration and series of events will lead to actual outcomes with action-oriented research, initiatives, and technical solutions for this grand, alarming challenge”.
A short overview of the outcomes of the day:
Participants agreed that, to counter the negative affects of deep fakes, it will take a multi-pronged approach, including:
- Increase the general public’s awareness of deep fakes and video manipulations
- Provide guidance for law enforcement and courts when dealing with video evidence which could be AI manipulated
- Advocate for technology companies to take greater responsibility when publishing and disseminating videos which could contain misinformation through AI manipulations
- Connect policy-makers with the most up-to-date academic research and state-of-the art technology, so they are informed when making regulations and policy on the topic
- Continue developing technologies that can automatically (and through techniques such as crowd-sourcing) detect video manipulations.
DSI, UNICRI and other workshop participants are currently and actively working on how to prioritise and carry out these actions points. Keep an eye on this website and our social media for updates and more information….