The Power of Predictive GenAI: Combining Generative AI with Predictive Machine Learning
Before generative AI became the major industry trend it is today, there was predictive AI, which focused on providing predictions for future events based on data. However, imagine the possibilities if you could combine the power of both technologies into one seamless solution.
Pecan AI: Bridging the Gap
This is where Pecan AI comes in. This eight-year-old startup has already established itself as a provider of a predictive analytics platform for enterprises, raising an impressive $116 million in funding since its inception. Now, Pecan AI is taking things a step further with the launch of their new tool, Predictive GenAI.
Predictive GenAI is a groundbreaking tool that combines the capabilities of modern generative AI with predictive machine learning. Zohar Bronfman, the CEO and co-founder of Pecan AI, explained the inspiration behind the tool: “While we were working in our side of the neighborhood on the classic machine learning predictive analytics solutions, on the other side of the neighborhood the entire gen AI revolution happened.”
“One thing gen AI is terrible at is creating predictions.”
– Zohar Bronfman, CEO and co-founder, Pecan AI
While generative AI may not excel in making predictions, predictive machine learning techniques can be complex and difficult to use. Pecan AI’s Predictive GenAI aims to bridge this gap by enabling data scientists to build and generate predictive AI models more easily.
Democratizing AI
Pecan AI has a goal to make machine learning and AI accessible to companies in the simplest way possible. Traditionally, AI platforms have been primarily used by data scientists. However, Pecan AI aims to democratize AI capabilities and bring them closer to the business side of things within companies.
“Historically, data scientists were the primary users of AI platforms, and in particular, predictive machine learning technology.”
– Zohar Bronfman, CEO and co-founder, Pecan AI
The Predictive GenAI capability of Pecan AI consists of two parts. While generative AI is known for its versatility in tasks such as building chatbots and summarizing content, it is not ideal for making predictions. According to Bronfman, the datasets used by generative AI tools during training are not in the proper format required for predictive modeling.
“For a predictive model, the dataset needs to have each row as a distinct entity, with each column representing a specific feature and a label column for the target variable.”
– Zohar Bronfman, CEO and co-founder, Pecan AI
In real business scenarios, obtaining datasets in the required format involves significant data engineering work. Generative AI models struggle with transforming raw tabular data from different sources into the necessary flat, two-dimensional format for predictive modeling, a task that typically requires an experienced data scientist.
Additionally, Bronfman noted that using a vector database alone is not sufficient for comprehensive predictive AI modeling. While vector databases and embeddings can support basic predictive capabilities with a limited set of features, more complex predictive models require data scientists to perform intricate feature engineering to prepare the data in the proper format.
Automating Data Preparation and Feature Engineering
Pecan AI is also focusing on automating data preparation and feature engineering to enhance model accuracy and address issues like data leakage. Data leakage refers to the utilization of information during the training process that wouldn’t typically be available during predictions.
“It’s not trivial to identify leakage, especially if you’re not a professional data scientist. So we have, for example, automated ways of identifying leakage.”
– Zohar Bronfman, CEO and co-founder, Pecan AI
