Artificial intelligence startup Alembic announced today that it has developed a brand new AI system that it claims will completely eliminate the generation of false information that plagues other AI technologies, an issue often called “hallucinations.” In an exclusive interview with VentureBeat, Tomás Puig, co-founder and CEO of Alembic, revealed that the corporate is presenting the brand new AI in a keynote presentation today Forrester B2B Summit and can present again at next week Gartner CMO Symposium in London.
The key breakthrough, in response to Puig, is the startup's ability to make use of AI to discover causal relationships, not only correlations, in huge corporate data sets over time. “We mainly immunized our GenAI from ever hallucinating,” Puig told VentureBeat. “It’s a deterministic output. It can actually discuss cause and effect.”
Solution to the hallucination problem
Hallucinations have been a serious obstacle to the adoption of AI systems resembling chatbots and virtual assistants in corporations. Leading AI models can generate realistic-sounding text, but often produce false or nonsensical information, making their use for business-critical applications dangerous. By eliminating hallucinations, Alembic goals to make AI protected and reliable in order that corporations can use it for a big selection of knowledge evaluation, forecasting and decision support needs.
To accomplish this feat, Alembic built its own supercomputer infrastructure and developed latest mathematical techniques to represent corporate data as time-aware graph neural networks. “Every time we see one in all these chain reactions or levers, we are able to understand all the elemental components of what you are promoting,” Puig explained. “These all grow to be little mini-neurons, and we put them into an enormous graphical neural network. But it’s a causal-aware, time-aware graphical neural network.”
The causal reasoning engine powers deterministic AI
At the guts of Alembic's breakthrough is a novel graphical neural network that acts as an engine for causal reasoning. This AI brain takes data from a wide range of enterprise systems, from sales databases and marketing platforms to analytics tools to TV and radio, and organizes it into a posh web of nodes and connections that captures how different events and data points relate to 1 one other others over time.
“It’s almost like a 3D representation of the corporate,” Puig told VentureBeat. “Imagine having the ability to see every interaction between every customer and each a part of the corporate and the way those interactions are passed through the corporate to attain results.”
The key’s that Alembic's AI not only learns patterns and correlations from this data, but additionally identifies the causal relationships that really drive business outcomes. By understanding the “why” behind historical results, the system can predict the impact of future actions with a high degree of certainty and even recommend the optimal interventions to attain a desired goal.
The underlying neural network can extrapolate and make predictions as latest data points are added, simulating the possible future impacts. “So if you insert a brand new node, it creates the expected chain response,” Puig said. This generative capability, built on Alembic's bespoke “core model”, differentiates the corporate's approach from the “expert mix” of other enterprise AI providers, which Mr Puig dismissed as “just microservices”.
Strong interest from Fortune 500 corporations and analysts
Interest in Alembic's AI breakthrough was high as the corporate was already working with 9% of the Fortune 500, in response to Puig, following private briefings and proposals from PhD experts at Nvidia and other large, unnamed customers. “When we showed it to (Forrester and Gartner), they mainly lost it. I’ve never seen anything like this, I’ve needed to go to 26 analysts thus far, each on the IT and MarComms side.”
Alembic's technology reaches a vital point for the adoption of AI in enterprises. Spending on AI technologies is predicted to extend reach the $500 billion mark by 2024, in response to IDC, but trust issues remain a serious obstacle. If Alembic can truly deliver enterprise AI that executives can depend on without fear of embarrassing or costly hallucinations, it could help speed up AI adoption across industries from finance to marketing to manufacturing.
With strong interest from early customers and endorsements from influential analyst firms like Gartner and Forrester, Alembic appears poised to disrupt the crowded enterprise AI market. But the corporate still faces the challenge of proving that its technology can scale beyond initial pilots to deliver concrete end results for big enterprises. As the AI race heats up, Alembic's “hallucination-free” approach could grow to be a serious selling point – or a cautionary tale concerning the gap between research breakthroughs and real-world impact.
