Yesterday was an exciting day. First, at the closing of the conference, it was announced that I, along with the with my colleague Corinna Cortes (of Google), would be one of the Program Chairs of next year’s conference. This is a really great honour. Many of the key figures in machine learning have done this job before me. It will be a lot of work, in particular because Max Welling and Zoubin Ghahramani did such a great job this year.
Then, in the evening, we had a series of industrially sponsored receptions, one from Microsoft, one from Amazon and one from Facebook. I didn’t manage to make the Microsoft reception but did attend those from Facebook and Amazon. The big news from Facebook was the announcement of a new AI research lab to be led by Yann Le Cun, a long time friend and colleague in machine learning. They’ve also recruited Rob Fergus, a rising star in computer vision and deep learning.
The event was really big news, presented to a selected audience by my old friend and collaborator Joaquin Quinonero Candela. Facebook had already recruited Marc’Aurelio Ranzato, so they are really committed to this area. Mark Zuckerberg was there to endorse the proceedings, but I was really impressed by the way he let Joaquin and Yann take centre stage. It was a very exciting evening.
I’d guessed a big announcement was coming so I climbed up onto the mezzanine level and took photos and recorded large parts from my mobile phone. They’d asked for an embargo until 8 am this morning, so I’m just posting about this now. I also cleared it with Facebook before posting, they asked that I remove the details of what Mark had to say, principally because they wanted the main focus of this to be on Yann (which I think is absolutely right … well done Yann!).
A commitment of this kind is a great endorsement for the machine learning community. But there is a lot of work now to be done to fulfil the promise recognized. Today we (myself, James Hensman from Sheffield and Joaquin and Tianshi from Facebook) are running a workshop on Probabilistic Models for Big Data.
Mark Zuckerberg will be attending the deep learning workshop. The methods that are going to presented at these workshops will hopefully (in the long term) deal with some of the big issues that will face us when taking on these challenges. In the Probabilistic Models workshop we’ve already heard great talks from David Blei (Princeton), Max Welling (Amsterdam), Zoubin Gharamani (Cambridge) as well as some really interesting poster spotlights. This afternoon we will hear from Yoram Singer (Google), Ralf Herbrich (Amazon) and Joaquin Quinonero Candela (Facebook). I think the directions laid out at the workshop will be addressing the challenges that face us in the coming years to fulfil the promise that Mark Zuckerberg and Facebook have seen in the field.
Very proud to be a program chair for next year’s event, wondering if we will be able to sustain the level of excitement we’ve had this year.
Update: I’ve written a piece in “the conversation” about the announcement here.