Kyle Polich

Kyle studied computer science and focused on artificial intelligence in grad school. His general interests range from obvious areas like statistics, machine learning, data viz, and optimization to data provenance, data governance, econometrics, and metrology.

Posts
20+
Linkedin
Deploying Machine Learning to Production with MS SQL Server
January 29, 2018

Deploying Machine Learning to Production with MS SQL Server

A dirty secret of machine learning in industry is how often people re-invent the wheel figuring out how to deploy their models into a production environment. Largely, this is due to the lack of canonical tools that have all the necessary bells and whistles. In this episode, I discuss the process with Tobias Ternström. We have a detailed discussion about the questions a practioneer would have when considering how they might use SQLServer as the right tool for their production machine learning model deployments.
Colossus: The Forbin Project
January 22, 2018

Colossus: The Forbin Project

This 1970s film from Universal Pictures, based upon the 1966 science fiction novel Colossus, by Dennis Feltham Jones, is required watching in the genre of artificial intelligence movies. It's the story of what happens when a military supercomputer comes on line and is given control of important systems.
Word Clouds Revisited
January 15, 2018

Word Clouds Revisited

Listeners to the Data Skeptic podcast will recall an episode in which I put forward a call to arms suggesting: Let's kill the word cloud. I gave my strongest arguments for why I hate word clouds, but I didn't provide any empirical support for my point of view. Lucky for me, a new study fills in this gap.
Are Thermostats Artificially Intelligent?
January 1, 2018

Are Thermostats Artificially Intelligent?

In the last episode of Data Skeptic, Kyle provided a definition of artificial intelligence that was broad enough to include thermostats. This post expands on that discussion with a few finer points. Please share your thoughts in the comments!
January 1, 2018

Data Skeptic T-Shirt Sizing

We've gotten some questions about the manufacturing and sizing of our t-shirts. Here are the measurement details.
December 5, 2017

NIPS Day 2

The Neural Information Processing Systems conference (NIPS) reached a record 7850 registered attendees this year. 679 papers were accepted out of 3590 submissions (18.9% acceptance rate). The accepted papers where 90% academic and 10% industry submissions. This is my day 2 summary.
December 5, 2017

Reprogramming the Human Genome Using AI

The second day of NIPS kicked off with this presentation from Brendan Frey. In the first minute, he put forward the bold claim "without artificial intelligence medicine is going to completely fail".
November 22, 2017

Computational Complexity Theory for Data Scientists

As was announced a while back on the Data Skeptic podcast, the show is going to have running themes for the foreseeable future, in which we'll spend a few months talking about one topic or area. We more or less did this earlier this year when we focused on deep learning following by a few episodes on deep learning in medical applications. These themes will become more formal and more obvious going forw
November 20, 2017

The Formal Statement of P not equal to NP

In the last episode of Data Skeptic, I asked Lance Fortnow about whether or not is was possible P vs NP was ill-posed. He correctly pointed out that, while some surprising result come emerge (like showing its impossible to solve it), it can't be ill-posed, because the problem has a formal mathematical statement. That statement is be
November 15, 2017

Microsoft Connect 2017

I attended Microsoft Connect today. These are my notes about the announcements I found notewor
October 22, 2017

Skepticism in the Black Box World

In recent years, much discussion has taken place about machine learning models being black boxes. Some models are said to be so complex that a human can't comprehend the function these models comp
October 13, 2017

T-tests, Normality, and the Central Limit Theorem

A listener wrote in and said: "I was just listening to an old episode of Data Skeptic on the 'Student's t-distribution'. In the episode Kyle strongly emphasized that the populations that the samples are coming from need to be normally distributed... I thought that one does not need to care what distribution governs the population (whether it is normal or not) because the sampling distribution will always be normally distributed due to Central Limit Theorem." This post is a response to this excellent question.
September 28, 2017

Recommending a Guest

Details for anyone interested in recomending a guest.
September 26, 2017

Feature Engineering Github Commits

A listener question about github commits and machine learning feaure engineering.
September 15, 2017

A Musing on the Difference Between Humans and Machines

The more I have learned, the less convinced I have become that there is anything special about the human brain which cannot eventually be replicated in a machine. If one wants to claim that humans are intrinsically more special then ANYTHING which could ever be engineered, I believe the burden of proof is on them to demonstrate this claim.
September 13, 2017

Deep Thinking

This blog post is about Garry Kasparov's Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins. Yet, it's not a review, exactly. Nor is this a response. These are just my thoughts I had during the deeply enjoyable time I spent reading this excellent book.
September 12, 2017

Neurology and Data Science

Last week I had a chance to visit USC's LONI: Laboratory of Neuro Imaging. Thanks to Farshid Sepehrband (@fsepehrband) for the invite and tour! The best part for me about doing data skeptic is when I have opportunities to visit fantastic companies and labs doing really interesting work.
September 11, 2017

Voice Recognition Exploit Exposed

A recent article on TechCrunch warns that hackers are sending silent commands to speed recognition systems with ultrasound. These sorts of attacks are interesting because are sometimes plausible, sometimes outrageous, and oftentimes teach us something about security concerns we didn't know to worry about previously.
September 10, 2017

Prophecy Fulfillment

This bonus feed episode is a short recording made in Vietnam
September 8, 2017

Why didn't you grill that guest more?

I got an email from a listener (whom I will refer to only by his first name, Alex) who had some feedback on my recent episode with Andy Martin from Zillow. The email was long but I can paraphrase the core question...
August 27, 2017

Plimpton 322

I recently came across the Guardian's coverage of Daniel Mansfield's work on an ancient mathematical tablet. I found the coverage to be deeply disappointing, in no small part to the misleading way, IMO, that Mansfield presents the res