Is It a Good Idea to Use Synthetic Data to Build a Model?

People often think about model understandability and explainability as like, i should be interrogating model weights. I think that's the absolute wrong way to think about it for a hundred different reasons. These are programmes that we are commanding and directing with data. And so the explanations that you are looking for are in your data. They're not in your model. In the model, you're just chasing phantoms when the data actually has the answers.

Play episode from 49:35

chevron_right

Transcript

chevron_right

Transcript

Episode notes

MLOps Coffee Sessions #64 with Slater Victoroff, The Future of AI and ML in Process Automation.

Get the newsletter: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTNewsletter⁠⁠

// Abstract
The Unstructured Imperative
Recent advances in AI have dramatically advanced the state of the art around unstructured data, especially in the spaces of NLP and computer vision. Despite this, the adoption of unstructured technologies has remained low. Why do you think that is? How have the dynamics changed in the last five years?

Multimodal AI
Historic AI approaches have generally been constrained to one data modality (i.e., text or image). Recently, a wide range of papers in image captioning and document understanding have emphasized the need for more sophisticated "multimodal" techniques that can fuse information from multiple modalities. What is multimodal learning, and why is it so promising? Why are we seeing such an explosion of activity? What is Indico doing in this space?

Machine Teaching
As methods of supervision become more complex and multifaceted, many researchers have begun investigating the inverse problem. That is how do we design supervision systems that more naturally follow human processes? What are some interesting trends in "the space", and where can we expect this field to go in the next few years?

// Bio
Slater Victoroff is the Founder and CTO of Indico, an enterprise AI solution for unstructured content that emphasizes document understanding.
Slater has been building machine learning solutions for startups, governments, and Fortune 100 companies for the past seven years and is a frequent speaker at AI conferences.

Indico’s framework requires 1000x less data than traditional machine learning techniques, and they regularly beat the likes of AWS, Google, Microsoft, and IBM in head-to-head bake-offs.

// Relevant Links
https://indico.io

--------------- ✌️Connect With Us ✌️ -------------
Join our Slack community: https://go.mlops.community/slack
Follow us on Twitter: @mlopscommunity
Sign up for the next meetup: https://go.mlops.community/register
Catch all episodes, Feature Store, Machine Learning Monitoring, and Blogs: https://mlops.community/

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Vishnu on LinkedIn: https://www.linkedin.com/in/vrachakonda/
Connect with Slater on LinkedIn: https://www.linkedin.com/in/slatervictoroff

Timestamps:

[00:00] Introduction to Slater Victoroff

[03:52] Slater's journey into ML

[06:52] Birth of Indico

[07:04] "Unstructured data"

[09:47] Historical perspective journey of Indico

[11:13] Adoption of unstructured technologies

[16:05] Technology techniques

[21:40] Unstructured data challenges

[26:35] Changing Supervision

[28:35] Synthetic data and its role