The Neural Processing Information Systems conference is the Woodstock of artificial intelligence. Currently in its 32nd year, the conference sold out in less than 10 minutes… better than most rock stars for Madison Square Garden.
In short, NIPS is a big deal.
Well, it‘s not NIPS anymore, now that the organizers changed it to NeurIPS after a lot of petitioning and infighting (over 100 scientists from John Hopkins in the spring wrote that the „acronym of the conference is prone to unwelcome puns“ – NILS and SNIPS was discussed, but after a lot of back and forth it was settled, as NeurIPS.
The site is still NIPS, but whatever.
NeurIPS this year was held from December 3-7 in Montreal, Canada.
As customary with MSG concerts and ComiCon, the featured talks were worth the wait, among them:
– The Necessity of Diversity and Inclusivity in Tech (Laura Gomez)
– Machine Learning Meets Public Policy: What to Expect and How to Cope (Edward Felten)
– What Bodies Think About: Biolectric Computation Outside the Nervous System, Primitive Cognition and Synthetic Morphology (Michael Levin)
– Reproducible, Reusable, and Robust Reinforcement Learning (Joelle Pineau)
– Investigations into the Human-AI Trust Phenomenon (Ayanna Howard)
– Making Algorithms Trustworthy: What Can Statistical Science Contribute to Transparency, Explanation and Validation? (David Spiegelhalter)
– Designing Computer Systems for Software 2.0 (Kunle Olukotun)
It‘s always crucial to take a look at the sponsors, those that put up the money to be close to the leading AI researchers, shown below. Let‘s organize them a bit:
Among the tech, mobile and entertainment heavy hitters were Amazon, Intel AI, IBM Research AI, Alibaba, Microsoft, Facebook, Salesforce and NVIDIA, Twitter, Apple, Google AI, Qualcomm, Deepmind, SAS, Netflix, Uber, Lyft, Baidu, Huawei, Samsung, along with Graphcore and Element AI.
From the financial services side, names included Citadel, Two Sigma, DE Shaw, JP Morgan, ANT Financial, Man AHL, Morgan Stanley, Capital One, RBC, Bloomberg, Sberbank, and Jump Trading.
Other interesting and, some, surprising names, including Pony AI, Biomind (we wrote about them in our AI book), Disney Research, Stradigi AI (from the Montreal AI hub), et al.
Our clients can review case studies for each of the firms above in the client portal and the AILO database (to identify the researchers and managers for each individually).
Let‘s turn to some of the scientific breakthroughs, a hallmark for NIPS for decades.
One of the top papers presented at NeurIPS last week was Neural Ordinary Differential Equations (ODE), a new family of deep neural network models.
The MIT Technology Review had a great quote from one of the authors on what this new deep neural networks means and how it differs from traditional neural networks: (Duvenaud) „A traditional neural network is like a piano: try as you might you won‘t be able to play a slide. You will only be able to approximate the slide by playing a scale. … Switching to an ODE solver is like switching your piano to a violin. It‘s not necessary always the right tool, but it is more suitable for certain tasks.“
Today‘s computing power with the decades old scientific breakthroughs of neural networks is powering the current advancements in AI, ML and DL, so advancing the concept of neural networks is one of the the true differentiators going forward.
