← THE FIELD RECORD
A KEYNOTE WITH NO ROOM

Demystifying Data Science, live from the Netherlands

30 July 2019 · Online, presented from the Netherlands · IMPACT

The New York school Metis ran its third Demystifying Data Science conference free and entirely online over two days. Tarry Singh gave the Tuesday keynote, Hilary Mason the Wednesday one.

Metis, a data science school on Madison Avenue in New York, ran Demystifying Data Science for the third year on 30 and 31 July 2019. It was free, it was live, and it was online end to end: two keynotes, fourteen further talks and six interactive workshops, with nobody in a room. Tarry Singh keynoted the Tuesday and Hilary Mason the Wednesday.

That is why the title slide reads "The Netherlands, July 2019" for a conference run out of the United States. It records where the speaker was sitting, not where the audience was. Eighteen slides, exported to PDF the same day they were delivered.

The talk opened on the talent problem. A slide on Siren Servers set up the claim that internet companies are amassing talent while enterprises are left data poor and talent poor. The next slide answered it: the talent exists, and the lab had been meeting it while training people inside enterprises and in boot camps. What was missing was a route.

The spine was a five-plateau climb. Plateau 1 is fundamentals: linear algebra, statistics, probability, programming. Plateau 2 is data visualisation and big data tooling. Plateau 3 is machine learning: regression, clustering, sampling, classification. Plateau 4 is deep learning, drawn as a dense cluster of CNNs, RNNs, autoencoders, GANs, deep reinforcement learning and capsule networks, with a footnote on the generative models that were new that summer: GLOW, BERT, GPT-2, XLNet, ERNIE 2.0. Plateau 5 is applied AI. Each plateau carries a completion percentage, 25, 34, 55, 74, 100, so a learner can place themselves on the slope rather than in a job title.

The last three slides looked past the market as it stood. One mapped roles the brain sciences might create: intention decoding engineer, dream imaging recording specialist, artificial retina implant engineer, neuro imaging and brain atlas specialist. Another did the same for space: space traffic controller, on-orbit refuelling specialist, Mars colony psychologist, space data scientist. Most of the space cells simply read "coming up". A longer version of the same deck survives, saved two days earlier, running to 32 slides and carrying the commercial material: blended learning tracks, a skills scorecard, cost and time curves, an enterprise section. The version that went out on the 30th dropped all of it. Eighteen slides, and none of them a sales slide.

There are no photographs of this one because there was nothing to photograph. The conference was delivered online, from a desk.

WITH
Metis, New York
← EARLIERThe case for borderless AI at GCAAI's annual meeting in BerlinLATER →A morning of medical AI at Tel Aviv Sourasky Medical Center