'Artificial Intelligence meets human intelligence': lessons from an inspiring evening
On September 5th, 2018 Trends, element61 and Moore Stephens Belgium (the mother company of element61) an evening-seminar around "Artificiele Intelligentie ", welcoming over 200 participants. Following the event, Bart Van Der Vurst, Data Science & Strategy Competence Lead at element61, wrote this report.
Artificial Intelligence: we can no longer imagine everyday life without it, and it is a hot topic for many employers and entrepreneurs. Therefore, Moore Stephens, in collaboration with Trends brought together three top speakers who are world-renowned in the field of AI, robotisation and what they mean for our future. With over 250 attendees from a variety of sectors, it was clear that AI is no longer just a buzzword and a hype, but that, apart from the well-known examples from films, books and articles, it is already being used in very concrete applications in various industries. We are happy to share a summary and review of this fascinating, innovative evening with you.
First speaker of the evening: Jonathan Berte, CEO & Co-founder of Robovision, a Ghent-based company specialised in applied ‘Deep Learning’. A discipline within AI in which there is no longer any preliminary data processing (i.e. feature engineering), but the algorithm itself learns, using mathematics, based on a large number of examples (a large amount of data) and/or trial and error.
Jonathan made a nice presentation of concrete applications in various sectors:
- In agriculture: smart control of robots to pick up plants.
- In public security: recognition of individuals and behaviour.
- In the medical sector: recognition of diagnoses from scans.
All these applications are based on a learning process in which AI algorithms recognise and automatically learn patterns based on a very large number of labelled examples. As the success of many AI applications is thus linked to the availability of such labelled examples, Robovision has developed a platform, similar to Amazon Turf, which uses crowdsourcing in order to enable thousands of people to contribute to the process of labelling photos, marking points and classifying examples.
Ans De Vos, Professor at the Antwerp Management School and coordinator of the expertise centre Next Generation Work, addressed the issue from a different perspective and reflected on the impact of AI on the working environment and hence on organisations. We learnt that it is also the responsibility of organisations to support these changes for workers by working on the following:
- A new career ladder, in which job rotation and training (both internal and external) help employees develop skills and experience.
- Greater ‘commitment’ through a more intense involvement of employees in changes and processes.
- Increased ‘ownership’ by giving employees responsibility at all levels.
Professor De Vos also pointed out the growing importance of soft skills, creativity and flexibility. While we, as employers, attempt to protect ourselves against these changes by looking for the ideal candidate with all hard skills we can think of, we must also – especially – invest in people based on their human and social capital.
As keynote speaker we welcomed Pieter Abbeel, Professor at UC Berkeley and, as an entrepreneur, founder of Covariant.ai and Gradescope. As an international AI pioneer, he provided us with a clear explanation of the different types of ‘deep learning’ and the specific uses of each type:
- ‘Supervised learning’: based on a large amount of training data, we can e.g. train an algorithm to label photos with a sentence.
- ‘Reinforcement learning’: based on trial & error and great computing power we can teach robots how to walk or win video games.
- ‘Unsupervised learning’: here, we start from what we want to achieve, and then go back to how this should be done.
Pieter shared with us the possibilities he sees for AI in nearly all industries, but also mentioned the challenges and dangers of the developments in AI. With the key question 'How much easier is it to build AI that benefits some vs. AI that benefits all', he established the link with geopolitical challenges in ‘the search for the best AI’ and the fact that coordination of the innovators is crucial in order to actually reach that best possible AI.
Finally, during a panel discussion, we explored the challenges of AI further and discussed questions like what AI is good and bad at today, what the role of humans is and what obstacles AI will encounter in the coming years. A fascinating discussion and for many also the perfect start to continue the discussion during the walking dinner.
At Moore Stephens, we are looking back on a lovely evening and looking forward to our next Data & AI events, at which we would love to welcome you. If you would like to continue talking about AI in your organisation, you can also reach us via firstname.lastname@example.org.