The Edge of Artificial Intelligence

by David Vivancos

Being at the edge of Knowledge in your field, is something key to keep up, we say it is more important now than ever, but I think it can be fully extrapolated to all human history and endeavors, if you are not at the edge, your value starts to fade. What is true is that the fade speed is now faster than ever, that’s why we say it’s more important now.

What is also true is that the formula to get to this edge, is changing, some centuries ago the way to achieve it, was trough teachers and masters of each discipline, and since the invention of the printing press, books play a key role, and of course now internet is leading the way.

I always say that all books are “history” books, even my own book “From Big Data to Artificial Intelligence”, since, from the day your start writing it, to the day it get’s published a lot of new things could happen, and this time, could easily be a year or more, and now, in 1 year the world can change more than in 10 years of the last century.

I still think that books are a great way to get yourself started in a new discipline, but probably not the best way to keep up.

Also formal education, has the same problems as books, maybe very good to start but not to keep up (unless you research there), since 3/4 years of your life is still a long time in current terms for most fields, even with great and updated teachers with real hands on experience (sometimes hard to find).

That’s why I think the key is lifelong learning, and this means to be always learning on a monthly/weekly or even daily basis, here is where updated online sources can be critical, once you find and trust them, or great tools to help you with that like CapaBall (disclaimer: I am a partner and A.I. adviser here, but nevertheless l think it is a unique tool leveraging machine learning for human learning worth spreading).

But even beyond that, the main source of updated knowledge I use, is through International Conferences, where the leaders of a specific field, showoff and share its latest and brightest achievements, this is usually where the edge of the discipline lays.

The key point here is to find the main one/s and a few relevant events, since there is also a flood of conferences of each discipline worldwide, and your time is a very limited asset. Choosing the right events could be a hard activity, my rule is always follow the leaders and disrupters, and where they regularly go, you should go too.

I have been on the Artificial Intelligence field for over 25 years, from pure old style automation, to giving “artificial life” to avatars in VR historic worlds in 2001, to building a personalized search engine in 2006, to developing 200K+ lines of deep learning code over the last 5 years (the “golden age” of A.I.), or advising boards and CEOs about A.I. and the road to the automated world, or lecturing 50+ keynote A.I. conferences and seminars in the field.

In this field my chosen events to be at the edge are:
– ICML (International Congress on Machine Learning)
– NIPS (Neural Information Processing Systems)
– ICLR (International Conference on Learning Representations)
– CVPR (Computer Vision and Pattern Recognition)
– ACL (Association for Computational Linguistics)

Also some of the Re-Work conferences, and the great online initiative of “AI With the Best”, and many others very specific to an area of knowledge.

Last week I attended the 35th International Congress on Machine Learning (ICML) on Stockholm, a massive gathering of about 5,000+ people working in the field from all over the world, just a few years ago in these conferences we only gather a few hundred people at the very best, but now it is growing exponentially. We will soon be needing a football or baseball stadium to fit the attendees.

Some of the “fathers” of the discipline were there like Yoshua Bengio or Yann LeCun, he was the invited talk at IJCAI a joint event this year with ICML, AAMAS, ICCBR and SOCS, in the Federated Artificial Intelligence Meeting FAIM. (sorry for the endless list of acronyms)

The event included 67 workshops, 9 Tutorials and 621 accepted papers for several parallel presentation tracks and posters sessions.

Also a expo area including interesting A.I. showcases from Deepmind, Google, Nvidia, FaceBook, Microsoft, Amazon, Intel, Baidu and many others like the recent and worth following Element AI from Jean-François Gagné and Yoshua Bengio.

The edge of A.I. is built at the labs, universities, corporations and startups were we create and test the latest machine learning tricks, algorithms, datasets, pipelines or frameworks. To try to go beyond and enable something previously thought impossible.

But building the technology and the science is not enough, you need to share and connect, and events like this are a great venue for that, and the best way to stay up to date and open your mind.