Our guest this week, Dr. Rajiv Shah, recently generated a lively discussion on Reddit. Subject of the debate was reproducibility of an article published in the notable science journal Nature. In this case unlike episode 7 of 26.1, where Dr. Rachael Tatman touched on her efforts to get industry AI folks to follow academia’s standards for reproducibility, Dr. Shah shared how using the same dataset as the article authors, he was unable to reproduce the results. In this case the aim of the paper's authors were to predict aftershocks of Haiti's disastrous earthquakes, a disaster the country and its people continue to struggle with today. Though we open our discussion about the buzz Raj got on Reddit, like every episode, we manage a discursive discussion in 26.1 minutes that touches on AI in manufacturing, C-level leaders getting to speed with AI practices, and the need for lay people to acquire a literacy about AI methods.
This week’s episode may be our most cerebral to date, probably thanks to hosting our first University of Washington Husky, Ph.D. graduate. Dr. Rachael Tatman shares snippets of her experience at Kaggle, stochastic approaches to ML models, errors in ML models, understanding prediction, and importance of reproducibility. A big takeaway references Dr. Tatman’s PyCon talk, “Put down the deep learning: When not to use neural networks and what to do instead.” We explore how many businesses benefit from a simple linear regression model instead of investing millions of dollars of compute time to a deep learning approach. Included with our conversation with Dr. Tatman is a discussion of linguistics and will Don ever get accurate translations of his Korean friends’ tweets in real-time. So far, our conclusion is that we’re far away from the singularity. Episode seven of 26.1 AI Podcast delivers lots of value for business leaders contemplating their future AI strategy.
This week we have Algorithmia founder and CTO Kenny Daniel. Algorithmia has grown tenfold over the past six years, and the startup just completed a $25m Series B raise in May. Perhaps our densest episode to date, we manage to pack a lot into a nearly perfect 26.1 minutes. We cover in this episode -- Algorithmia’s founding story with Kenny’s fellow founder Diego Oppenheimer, challenges bridging the culture between traditional software professionals and the current practices of AI professionals, needs for analogs for prosaic software tools tailored for AI deployment, and will the data engineer job title survive as AI matures.
This week’s guest Mark Hoffman shares how exposure to data analysis in a nuclear physics lab led to him joining years later NASA Jet Propulsion Laboratory as a data scientist. At NASA, he’s repurposed ML work he delivered on fraud detection for one of the country’s largest healthcare orgs for NASA’s projects where spacecraft sensors deliver many false alarms. Included in this episode is Mark’s experience launching a successful startup applying ML for black car services. This startup was successfully sold. Some of the computing infrastructure challenges our guest solved as part of his startup foreshadows our future guest Kenny Daniel’s episode. Kenny a founder of Algorithmia is working hard to save data sciences from the overhead of wrestling with computing infrastructure and concentrate on delivering more data science, ML, and AI, faster.
In this interview we speak with Venice, CA based tech insider Everett Carney on broad range of tech topics including blockchain, GANs for players to generate custom goods within games, and compliance around AI. Along with some future thinking and optimism about AI/Blockchain, out guest sprinkled in some precautions as well.
Meet Ramkumar Hariharan head of applied AI at macro-eyes. A startup, macro-eyes has won two Bill and Melinda Gates Foundation Grand Challenges Explorations awards. Learn how appearing on Indian television and radio contributed importantly to Ram’s journey becoming an AI practitioner and educator. Also how Ram and team deploys AI in India, Africa, and Seattle to improve medical outcomes for access to care and effective vaccination programs. Once again, learn about AI in an easily digested 26.1 minute session.
Don and Brian interview a machine learning veteran Avilay Parekh. He was one of the 60 original creators of Microsoft Windows Azure platform, has helped an IoT startup use machine learning on circuit printed socks, was an engineer at Amazon, and now is launching the ability to create 3D simulations of real life cities and places on the Unity platform for training AI models dependent on visual data.
Terrazas is the AI LA President @aila_community and Founder @grassy_vibe. Don and Brian ask Todd about his progression toward becoming a story teller in the AI Space, current involvement, travels, the LA AI Tech scene, and future events.