Blog: My Take on the O’Reilly Artificial Intelligence Conference — NYC
My Take on the O’Reilly Artificial Intelligence Conference — NYC
What comes to your mind when you think about an AI conference? Fast pace speakers, companies with cool ideas and demos, interesting little swags, the eye-opening road maps to move toward a fancier future? To be fair, the O’Reilly AI conference had it all.
This year, I was honored to be selected as one of the Intel Student Ambassadors who were supported by Intel to travel to NYC and attend O’Reilly AI Conference. (If you are a student and interested in AI, make sure to learn about this program, you might be selected to go there next time!)
During the two days of the conference, what I liked the most were the keynotes. Over the course of 90 minutes, 10 minute long talks discussed the main AI achievements and challenges, inspired and bombard me with new ideas, pointers, and topics to learn next.
I’m not going to list all the talks and topics here but I share those that I liked the most:
- Martial Hebert overview of challenges in AI for robotics and developments in the current research. (watch the gist of it here)
- The Computational propaganda presented by Sean Gourley, not that it was new, but it was interesting the way he put it in the context
- Aleksander Mądry’s discussion on whether the AI is human ready or it needs more time to become ready for us (watch it here)
- If you are not reading the Netflix’s blog posts, you might also find Tony Jebara’s talk of how Netflix is using AI to personalize the user experiment interesting (here). They deliver a unique experience through personalization of movie covers, text and descriptions!
- and as a data scientist, I was also interested in Carlos Morales talk on overview of Nauta, an open source multi-user platform that lets data scientists run complex deep learning models on shared hardware.
Overall, what did I learn?
A couple of main points are:
- High level road map: AI solutions so far assumed that the customers will have the training data somehow, so focused on how to make the AI applicable for the consumers. However, now that the tools are semi-mature and hardware and software is easy to be accessed through cloud services and Automated ML solutions, consumers got to know that their data is neither clean nor representative as they have expected. As a result, there was a new trend of companies who were delivering data cleaning, labeling, and preparing services to address such need. But still, a lot left to be done! (Startup opportunity!)
- Tools and frameworks: Get to know more about BOHB technique (and their implementation here) for hyper-parameter tuning of neural network models which work more efficient than the grid, random, and Bayesian search. Related to that is the AutoML to be checked.
Also learned about the bqplot which is a 2-D visualization system, interactive widget for Jupyter notebook offered by Bloomberg.
Well, any resources for those who were not in the conference?
Even I didn’t get a chance to capture the whole interesting points there. I was struggling with people who came to sit next to me, then decided this is not the ideal row and turned back and stood right in front of me and specially those who were busy handling their hot coffee cup and in some cases shared it with my jacket :D But… I got you covered! I have resources for you:
To get a sense of what was shared during the keynotes, take a look at publicly available speaker slides & video shared on the O’Reilly website. Also, in case you are looking for a job, here is the job opportunity blackboard, which is a premium asset shared with you here:
Have you been there or get a chance to watch and read the resources I shared?
share with me what is your take on the conference materials, on where the AI is and is headed, and how are you benefiting, contributing, and/or improving this domain?