WEBINAR | AI Prototype to Production: Operationalizing and Orchestrating AI
October 29, 2021

5 Takeaways from Perceive 2021

Table of Contents:

5 Takeaways from Perceive 2021

Perceive 2021 was our biggest event ever, with two full days of content and an international collection of speakers. If you missed the event and are looking for actionable insights to accelerate your journey into Software 2.0, we’ve got you covered. Just read on to learn about five essential takeaways from Perceive 2021.


1. People are starting to think about unstructured data

One of our most popular sessions ever was one of the most general; Fundamentals of AI for Unstructured Data. People learned that the vast majority of data created every day is unstructured image, video, text, and audio data. Deep learning enables you to start making sense of these data types at scale and lets you turn the insights into products and services. Watch the session here. 


2. Customers love technical topics like anomaly detection, testing, and evaluation

Deep anomaly detection is a popular branch of machine learning research with many use cases such as predictive maintenance, cyber security, and health surveillance. Yuchen Fama Ph.D., Clarifai’s Director of Product, gave viewers a brief overview of challenges, methodologies with uncertainty quantification using Bayesian ML, an empirical study of uncertainty quantification using Bayesian ML, as well as an empirical study of predictive maintenance using VAE with sequence-learning based encoder/decoder setup. Take a look at the talk. 


3. Automating data labeling with AI is so hot right now

One of our most shared sessions walked customers through the process of automating data labeling with AI. Attendees took a look at what goes into collecting and preparing training data to create their own AI models. They learned about automatically annotating inputs when predictions are confident, using AI to assist human labelers, and quality control. Watch it here. 

4. People recognize the importance of collaboration when building AI-based software

Another popular session featured a guest presenter from the University of Toronto, Shurui Zhou, Ph.D. Dr Zhou spoke about "Improving Collaboration Efficiency for Building AI-based Software." Attendees learned that building AI-based software systems requires data scientists and software engineers across multiple teams, who typically have different skills and educational backgrounds, to work together. We learned about the development of the entire AI-based system, including both AI and non-AI components. Collaboration challenges faced by interdisciplinary teams and collaboration efficiency were also discussed. Take a look. 


5. The World is Listening

We broadcast Perceive this year in three different time zones, and the response was tremendous. We achieved about a 50/50 audience mix between customers in the United States and the rest of the world. Many organizations attended the event, including well-recognized brands like Disney, Target, and JP Morgan Chase. The media is also hungry to understand our work, and we attracted 30+ press mentions.