The Applied Machine Learning Days are back at the Swiss Federal Institutes of Technology of Lausanne for the fourth time. A five-day extravaganza took place with 31 hands-on sessions, 29 tracks of presentations, over 2'500 visitors. And 1 very important virtual speaker, Edward Snowden. Excellent in-depth coverage was done (in French) by media partner Heidi.news, who ran extra interviews on site. The previous two years have already been discussed on my blog (#038 #050 #051).

Ready to dive in after my gig in Bern on Friday with lots of ideas and inspiration, such is my luck, that I missed most of the conference this year on the road. Consolingly, there was an enormous amount of coverage on social media, and it was easy to get caught up in the enthusiasm and quality insights being shared.

Get ready for a barrage of tweets {:-)

#AMLD2020 featured far too many incredible references to list in a single blog. You will find tons of content for health, climate change, policy, etc. I will share a few samples from just one domain of focus - urban data. There was, for instance, an introduction to scikit-mobility, a sweet Python package for analysis of data about people on the move:

A clearly intriguing discussion of sustainable urbanization took place, posted here by Datamap - who also presented their own ML-based mobility tracking application and data producer cooperative that I'm very excited to be involved in.

There were eye opening perspectives from Mohamed Kafsi, who works on embedding climate change thinking into machine learning at Swisscom, and with whom I had a pleasant conversation at the stand:

Thank you to Konstantin Klemmer for spending some time with me in the break to share some key takeaways from the track on A.I. and Cities that he (very capably, by all accounts) facilitated. "We must not separate social systems thinking from investigation of cities", he said, and in this thread, shares insights from the people behind transformational projects:

A lot of themes fundamental to the engineering side of machine learning were discussed. I recalled the data quality and consistency issues that Martin Veterli memorably introduced last years event when I saw this:

From "systematic confusion about technology with intention", to stunning glimpses of where the next breakthrough might lead us ...

And the increasing influence of ML on the social sciences ...

An incredible amount of ground was covered in this short time. But even this was not the extent of AMLD. In fact, building on the spirit of hackathons like the one we organized in 2018, for the past several months five independently run challenges ran to culminate in presentations at the event. I enjoyed talking to the organizers of the Flatland challenge - and interactively sampling their results in the form of a retro game arcade.

The last day of the conference was also an opportunity to experience the EPFL Extension School, a groundbreaking initiative inspiring hopefully thousands of people to take a plunge into Data Science and related "Digital skills" - as well as being a model for next-gen remote learning. In the photo, I can see myself in the workshop - quite like back in my student days, resolutely in the back row.

Speaking of speeding things up...

In the meantime, the future of humanity was on the radar:

I commented regularly on Twitter during the duration of the day, so I'll also summarise using the same tweets here:

So. I'll try to shake out some clearer insights over the days ahead, as the content settles in. Keeping eyes peeled for the video releases, till August do we part, AMLD. Увидимся!

Learning is not finding out what other people already know, but is solving our own problems for our own purposes, by questioning, thinking and testing until the solution is a new part of our lives.
The Age of Unreason, Charles Handy