Wednesday Add-On Day (In-Person)

Those of you not registered are welcome to watch the presentations live by clicking on the links below:

Wednesday, October 27 will be an in-person add-on day with a technical workshop highlighting deep learning and high-performance computing with Alex Smola, VP of AWS, and Anshumali Shrivastava with Rice University at the BRC on Rice University campus. Masks are required indoors, the auditorium holds nearly 300 people with registration limited to around 125 for plenty of room to social distance, food provided in boxed and packaged format, and there will be an outdoor open sided tent with seating to allow eating/drinking to occur outside with continuous airflow.

Scalable and Sustainable AI: Overcoming the Wide Gap Between AI in Theory and AI in Production 

OpenAI recently noted that the computation cost of AI experimentation doubles every three and a half months. Existing solutions fail to catch up with this exponential demand.

Currently, the progress in AI is limited to tasks that closely resemble existing standard benchmarks and related settings such as Imagenet. Industries routinely deal with pipelines that go beyond these traditional settings. Going beyond routine pipelines results in catastrophic degradation of either the performance or the accuracy of the AI.

In the current state, even with all the hardware, scaling convolutions neural networks on ultra-high-resolution images is prohibitive. Deep Learning for Seismic Data Processing currently only caters to datasets with toy simulated settings. Model Parallelism for training extensive neural networks is needed by almost any large data processing company. However, current solutions are prohibitively slow and nowhere near satisfactory.

As a result, today, AI engineers are not designing models to solve a problem; rather, the problem is hammered to cater to well-known AI models. Scaling up the AI model remains the primary gap between AI in theory and AI in production.

In this workshop, we will explore recent advances in designing a new generation of scalable and sustainable AI algorithms that are exponentially cheaper and can cope with future demands. This workshop will feature Keynote Speakers from Industry and Academia, hands-on tutorials, contributed papers, posters, and networking sessions.

Bringing together AI and Data Science Professionals. #aidatascienceconf