How impactful are the events you’ve been attending lately? For me, it’s the actions events cause me to take that serves as my primary metric for impact. And in this regard, GTC 2017 in Silicon Valley was extremely impactful … before, during and after the event.
As a speaker at the event, I seized the opportunity to engage in the pre-conference training workshops offered via NVIDIA – who, apparently, feel that some of us techies just might benefit from a little guidance when it comes to our presentation skills … go figure 😉 As expected, I learned a lot from these sessions that I can continue to apply in future presentations – despite that fact that I’m definitely not a newbie when it comes to public speaking.
Whereas the opportunity to enhance my presentation skills was an unanticipated benefit of speaking at GTC, the need to develop and deliver compelling content was a well-known expectation – primarily based on my participation in past GTC events in the Valley, and even in Japan. Because my ‘pet project’ on extracting scientifically credible data from Twitter was accepted for presentation at this event, I spent time respectfully reframing it in a GPU technology context – as previous efforts had made use of in-memory text classification via Apache Spark on CPUs. Frankly, the outcome surprised me: As a consequence of my GTC-motivated reframing, I ‘discovered’ Natural Language Processing (NLP) – broadly speaking, the use of human languages by a computer. Moreover, by reviewing the breadth and depth of possibilities for actually doing some NLP on my Twitter data, I subsequently ‘discovered’ PyTorch – a Python-based framework for Deep Learning that can readily exploit GPUs. It’s important to note that PyTorch is not the only choice available for engaging in NLP on GPUs, and it certainly isn’t the most-obvious choice. As I allude to in my GTC presentation, however, I was rapidly drawn to PyTorch. Owing to this late-stage introduction of NLP on GPUs, however, I sadly didn’t have NLP results to share in my GTC presentation – a gap I’ve more recently closed as alluded to below.
Full disclosure: I haven’t attended a GTC event for a few years. As a consequence, I was literally gobsmacked by the engagement regarding Artificial Intelligence (AI) – engagement that went well beyond NVIDIA’s clever “I am AI” palindrome. As a vendor with a booth on the exhibits floor, I believe we had as many conversations about AI, Machine Learning and Deep Learning, as we did about High Performance Computing (HPC) – which, BTW, GPUs also excel at … in case you didn’t know. Amongst numerous other takeaways, I was able to validate that focusing on NLP was quite likely to deliver the results I was seeking in my pet project on extracting credible data from Twitter – a validation more than worth the price of admission from my perspective.
In preparing for my GTC talk, I spent a fair bit of time reading – not only about NLP, but also the mathematics underlying Deep Learning. Before I’d even left my hotel room in downtown San Jose, I reflected upon my own background in mathematics, and drafted a blog post shared previously.
GTC 2017 in Silicon Valley was held in early May of this year. About a month after the event, NVIDIA released recordings for all of the sessions. In addition to the video of the presentation I gave on Deep Learning from Twitter, you’ll find a PDF version of the slides that were presented. As a consequence of my GTC experience, I’ve continued to apply NLP to the selective extraction of data from Twitter. In this ongoing effort, I am increasingly making use of PyTorch via Jupyter Notebooks; and, as soon as I’m ready to scale my analyses to real-world use cases, I’ll be making use of various Univa-specific offerings – but more on that another time!
Net-net, GTC 2017 in Silicon Valley was, and remains, an impactful event from my perspective – and the conversation is just getting started! Although we look forward to other GTC events, your input doesn’t need to wait. In fact, we’d love to hear from you about your metrics for event impact, your HPC and Deep Learning projects, and more.