On November 27th, I sat on a panel discussing the present and future of artificial intelligence at Nasdaq Marketplace in New York. It was an honour to sit along side Peter Voss, originator of the now-predominant term Artificial General Intelligence, which refers to an AI that has a capacity roughly indistinguishable from that of a human.
On Friday, I provided a two-hour introductory lecture on Deep Learning and Deep Reinforcement Learning (specifically, Deep Q-Learning) to the Reinforcement Learning class at Columbia University for postgraduate electrical engineers. It was a pleasure to be able to speak to a room full of folks who were already abreast of the fundamentals of machine learning (e.g., cost functions, stochastic gradient descent) so that we could dive straight into the specifics of artificial neural networks. It was also fun to fill up all of the classroom’s overlapping, sliding chalkboards :)
Last month, I had the honour of speaking to the Pintô cultural group about what Deep Learning AI is, how it’s impacting society today, and how it might in the future. I met countless fascinating people from various fields, all of whom asked terrifically clever questions.
Click through for a story from TechTarget on top data science tools. They appear to have erroneously misattributed a (perfectly reasonable) quote to me on R, but my quote about Jupyter Notebooks is the real deal.
On Tuesday afternoon, we (untapt) hosted a Think Tank on the use of algorithms for improving human-resources processes. Our audience was highly engaged and we were delighted that the average of their rankings of candidates for a given role aligned precisely with our algorithm's rankings.
In this press release issued yesterday, I commented:
Through the Think Tank's discussions and interactive exercises, we observed further evidence that both recruiters and candidates gain from the use of AI tools. The event demonstrated an inherent need for what we've built at untapt. As human-resources professionals increasingly appreciate that AI saves them time and capital while simultaneously improving the quality of the prospects they identify for open roles, we are witnessing a shift toward the technology's adoption. This fuels our passion to continue delivering solutions that make a difference.
This press release from July 19th formally launched our (untapt's) AI product, an algorithm for matching candidates with roles at scale. It offers expert-human-recruiter accuracy but is millions of times faster.
In the piece, I was quoted as saying:
The complexity and nuance of problems in the talent space make them fascinating ones to tackle with models that incorporate automation. Through the application of the latest statistical computing approaches from academia, we've harnessed the might of our large, high-dimensional data to streamline human-resource processes with commensurate subtlety. I'm delighted that these models are now available to our clients, minimizing drudgery and freeing up their time to focus on creative, interpersonal work.
Gareth Moody, my colleague and dear friend, opined that my choice of language in the quote was not un-awkward and that my verbosity was excessive.
Forthcoming research out of Columbia University indicates that, compared with human screeners, an algorithm that decides which job applicants should get interviews "exhibited significantly less bias against candidates that were underrepresented at the firm".
Beginning July 21st for five weekends, I’ll be teaching my 30-hour course on deep learning in-classroom at the NYC Data Science Academy. This will be my third time teaching the course, which I developed based on the content of my forthcoming book.
While the first two iterations of the course — offered in late 2017 and Spring 2018 — were popular, only one out of every ten participants was a woman. In hopes of reaching gender parity, the Academy has whole-heartedly supported my suggestion that we offer a 30% discount to female students — their largest discount ever. To receive the sizeable reduction of nearly $1000, registrants are encouraged to use the code DLFORWOMEN upon checkout.
Two families of Deep Learning architectures have yielded the lion’s share of recent years' “artificial intelligence” advances. Click through for this blog post, which introduces both families and summarises my newly-released interactive video tutorials on them.
Following on the back of my Deep Learning with TensorFlow and Deep Learning for Natural Language Processing tutorials, I recently recorded a new set of videos covering Deep Reinforcement Learning and Generative Adversarial Networks. These two subfields of deep learning are arguably the most rapidly-evolving so it was a thrill both to learn about these topics and to put the videos together.
With SantaCon’s blizzard-y carnage erupting around untapt’s Manhattan office, the members of the Deep Learning Study Group trekked through snow and crowds of eggnog-saturated merry-makers to continue our Deep Reinforcement Learning journey.
We were fortunate have this session be led by Laura Graesser and Wah Loon Keng. Laura and Keng are the human brains behind the OpenAI Lab Python library for automating experiments that involve Deep RL agents exploring virtual environments. Click through to read more.
At the fifteenth session of our Deep Learning Study Group, we began coverage of Reinforcement Learning. This post introduces Deep Reinforcement Learning and discusses the resources we found most useful for starting from scratch in this space.
My 30-hour Deep Learning course kicks off next Saturday, October 14th at the NYC Data Science Academy. Click through for details of the curriculum, a video interview, and specifics on how I'll lead students through the process of creating their own Deep Learning project, from conception all the way through to completion.
In February, I gave the above talk -- one of my favourites, on the Fundamentals of Deep Learning -- to the NYC Open Data Meetup. Tonight, following months of curriculum development, I'm providing an overview of the part-time Deep Learning course that I'm offering at the Meetup's affiliate, the NYC Data Science Academy, from October through December. The event is sold out, but a live stream will be available from 7pm Eastern Time (link through here).
In the twelfth iteration of our Deep Learning Study Group, we talked about a number of Deep Learning techniques pertaining to Natural Language Processing, including Attention, Convolutional Neural Networks, Tree Recursive Neural Networks, and Gated Recurrent Units. Our collective effort was primarily focused on a popular variant of the latter -- Long Short-Term Memory units. This blog post provides a high-level overview of what LSTMs are, why they're suddenly ubiquitous, and the resources we recommend working through to understand them.