Cutting-Edge AI Used to Train Crisis Counselors

This week in AI & Machine Learning: computer vision in e-scooters, gpt-2 training counselors, monitor models in production, PyTorch3d, commercial ML opportunities, and more!

Artificial Intelligence News:

The AI field is growing fast! Here are some of the most interesting or thought provoking AI stories and applications that made the news this week.

E-scooters Embrace AI to Cut Down on Pedestrian Collisions

This isn’t the first time we’ve seen E-scooter companies using computer vision to help manage how shared scooters are used within crowded cities. But instead of being used to monitor if customers illegally ride or park the scooters, this company aims to prevent pedestrian collisions by detecting pedestrians with computer vision techniques similar to some autonomous vehicles. 

It will be interesting to see how much this improves rider safety and potentially increases scooter ride sharing adoption in wary cities. 

AI Used for Training Counselors to Help Teens in Crisis

A crisis hotline is using a chatbot trained with OpenAI’s GPT-2 natural language processing (NLP) algorithm to help train employees for complex situations with teens in crisis. The idea of using NLP as counselors or therapists is a hot topic. This approach of using machine learning to train humans is intriguing and could help save lives!  

Artificial Intelligence to Take Drive-Thru Orders at Englewood Restaurant

A restaurant is switching over to speech recognition software to take orders at a drive-thru restaurant. Think about using Alexa or Siri to place your order using your voice. Do you think it will improve order times or end up being more frustrating to customers? 

Developer Tools & Education:

Keep up to date with the latest relevant developer tool releases & educational posts.  

Lyra: A New Very Low-Bitrate Codec for Speech Compression

Checkout how Google AI is using generative machine learning algorithms to improve compression that can allow voice communication even on the slowest networks.

OpenCV Connected Component Labeling and Analysis

This week’s computer vision tutorial from PyImageSearch you’ll learn how to implement connected component labeling and analysis with OpenCV.

Rendering Volumes and Implicit Shapes in PyTorch3D

PyTorch3D added some interesting new features: implicit shape rendering, multiple ray sampling implementations, new volumes data structure, and more. I’m excited to see where 3D deep learning libraries take us!

Upcoming Online AI & Data Science Events:

Live tech events are a great way to learn skills, meet new people, and stay connected to amazing communities around the world!

Monitoring Machine Learning Models in Production

Mar 02, 06:00PM PST

Learn how to monitor and maintain a model after deployment in productions to ensure performance. 

Practical Approaches for Efficient Hyperparameter Optimization

Mar 16, 10:00AM PST

This talk will walk machine learning practitioners through guidelines for efficient hyperparameter optimization based on Oríon, an open source HPO framework

Intro to Computer Vision: Building Object Detection Models and Datasets

Mar 24, 10:00am PST

Build your own object detection model from start to finish. Includes how to do data annotation and model training on your own dataset.

Interesting Podcasts & Interviews:

Listening to podcasts is a great way to stay updated with the ever changing world of AI, and often provides deeper insights into how brilliant people solve problems. 

Commercial ML Opportunities Lie Everywhere | SuperDataScience

In this episode of SuperDataScience with Michael Segala (Co-Founder & CEO of SFL Scientific), learn how GPUs can be used to accelerate machine learning, the biggest government policy holding back machine learning, and more.

Low code, no code, accelerated code, & failing code | Practical AI

The hosts of Practical AI discuss low code/no code development and new learning opportunities for machine learning. 

Evolution and Intelligence with Penousal Machado | TWiML

Listen to Penousal Machado, Associate Professor and Head of the Computational Design and Visualization Lab in the Center for Informatics at the University of Coimbra discuss Evolutionary Computation and more.

Innovating Neural Machine Translation with Arul Menezes | TWiML

Hear what Arul, a 30-year veteran of Microsoft and a manager of machine translation research at Azure has to say about the evolution of machine translation, including seq2seq and transformer models.

Notable Research Papers:

Every week there are many interesting papers published to arxiv. I highlight some of the ones I find interesting below. 

• IBRNet: Learning Multi-View Image-Based Rendering

• Learning Generalizable Visual Representations via Interactive Gameplay

• SPINN: Sparse, Physics-based, and Interpretable Neural Networks for PDEs

• Retrieval Augmentation to Improve Robustness and Interpretability of Deep Neural Networks

• Investigating the Limitations of the Transformers with Simple Arithmetic Tasks

• Zero-Shot Text-to-Image Generation

• Transformer is All You Need: Multimodal Multitask Learning with a Unified Transformer

About Plainsight Technologies

Plainsight dramatically reduces the cost and complexity of AI-powered computer vision across many verticals and enterprise use cases, including manufacturing, retail, food service, theft, and more. As companies automate their operations with AI and Agents, Plainsight brings vision capabilities to watch and analyze events in the real world and generate structured data for analytics and reporting. Headquartered in Kirkland, Washington, Plainsight Technologies operates as a distributed team, delivering cutting-edge solutions worldwide. To learn more, visit plainsight.ai.