A keynote on borderless AI at DSPT Day in Porto
On 26 October 2019 Tarry Singh took the DSPT Day stage in Porto and made the case for borderless AI to a packed auditorium of Portuguese data scientists.
DSPT Day 2019 filled an auditorium in Porto on Saturday 26 October. Data Science Portugal, the community behind it, put its own description on the opening slide: a community of data science enthusiasts who share knowledge and experiences. A lit set piece spelling DSPT DAY 2019 sat stage left all day. Talks ran from chart selection for analysts to representation learning, and the room stayed full through the afternoon.
Tarry Singh worked from a deck built for a general audience rather than a technical one. It opened by separating artificial intelligence, machine learning and deep learning, then put one question to the room: where will your company be in the 2025 AI economy? A ranked chart of the top 25 global brands sat behind him while he asked it. The middle section was titled Borderless AI. It laid out the arms race between states investing hard in the technology, the surveillance limits some European countries were writing into their own law, and the trade in behavioural data that had made a handful of internet companies enormous.
Two client stories carried the second half. The first was breast cancer detection with the healthcare division of GOPA, aimed at women in Afghanistan who reach diagnosis late. The deck credited Dr Sohaila Niazi, described there as the first oncologist in the country and written up in Elle, and set out a four-month pilot: annotations from German pathologists, a dataset organised and pre-processed from scratch, and a prototype at roughly 82 per cent accuracy on a very small sample, with 98.5 per cent reached later. The second was diabetic retinopathy in China. Working with Zhongda Ophthalmology and researchers in Australia and Singapore, the team standardised the grading of 70,000 fundus photographs against the British diabetes screening system. The resulting self-service fundus examination was reported at 95 per cent accuracy and already held SFDA approval.
deepkapha.ai ran from Assen in the Netherlands. The management slide listed three people: Tarry Singh as chief executive, Foke S. van der Helm as chief operating officer, and Jie Mei as chief AI officer and head of learning. The same slide put the number of professionals trained worldwide at more than 25,500 by that October.
