Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More
Modern generative AI relies heavily on user-inputted prompts to get results. But what if there was a system that was smart enough to actually predict what the right prompt is to generate the best results?
No, it’s not Artificial General Intelligence (AGI), or some form of science fiction Minority Report-style precognition, it’s a new approach that Dataminr refers to as – ReGenAI.
Dataminr has been building out its real time AI platform since 2009, long before the concept of gen AI was as commonplace as it is today. The company’s technology has been steadily evolving over the last decade to help organizations better understand data coming from various sources of information to determine potential risks. Among Dataminr’s customers are large organizations such as the United Nations (U.N.), Airbus and Dyson.
While Dataminr has used gen AI previously to provide text descriptions of detected events, ReGenAI goes a step further. With ReGenAI the system is now also able to regenerate event alerts in real time as events change.
“ReGenAI is the next platform evolution for Dataminr,” Ted Bailey, Dataminr’s Founder and CEO, told VentureBeat.
How Dataminr is moving beyond prompts
Bailey noted that when an event occurs, knowing about it is very powerful. However, corporations and governments need to know more than just the fact that the event is happening, they need to know the scope and the potential implications.
What ReGenAI does is it can dynamically detect new developments from data sources, synthesize them in real time and update the information and recommendations as events occur.
“This is a use of generative AI that’s beyond the prompt, it’s trying to automatically reformulate itself and doing so by connecting to the real-time world,” Bailey said.
ReGenAI is a combination of predictive and gen AI
With many modern technologies including some gen AI tools, there is some form of suggested next steps which can help users to decide what to do next. Bailey emphasized that’s not what ReGenAI is doing.
He also stressed that ReGenAI is not a form of AI agent-based approach that is just chaining the results of multiple gen AI models. Instead what’s happening with ReGenAi is a bit more complex.
Bailey said that Dataminr has built and trained its own large language models (LLMs). The LLMs have been trained on Dataminr’s data archive of over a decade of events.
“We’ve done a combination of customized and open source models, training our own models, such that today all of our models are internal and we’re not relying on any third party,” he said.
Bailey explained that the way the Dataminr system works is when an event is detected, predictive AI models developed by the company are used to create a chain of related information from any number of different types of data sources.
“As we get all this streaming real time data coming into our system, we’re processing it and we’re discovering all these interim developments,” he said. “We take these interim developments and we describe them all with generative AI.”
ReGenAI works on top of the existing predictive and gen AI models to apply logic in real time to help users understand what’s important about an event at any given moment.
“It’s a combination of classic predictive AI taken to the next level in the age of foundation models and LLMs together with generative AI, fusing the two together to create this new form of dynamic, iterative live information,” he said.