To receive industry-leading AI updates and exclusive content, sign up for our daily and weekly newsletters. Learn more
Artificial intelligence startup Alembic announced today that it has developed a new AI system that it claims can completely eliminate the generation of misinformation, a problem known as “hallucination” that plagues other AI technologies. In an exclusive interview with VentureBeat, Alembic co-founder and CEO Tomás Puig revealed that the company will unveil the new AI during a keynote address at the Forrester B2B Summit today, and again next week at the Gartner CMO Symposium in London.
The key breakthrough, according to Puig, is the startup’s ability to use AI to identify causation, not just correlation, across huge corporate datasets over time. “We’ve essentially immunized GenAI to not hallucinating,” Puig told VentureBeat. “It’s a deterministic output. It can actually talk about cause and effect.”
Solving the hallucination problem
Hallucinations have been a major obstacle to enterprise adoption of AI systems like chatbots and virtual assistants. While leading AI models can generate lifelike text, they often produce false or gibberish, making them risky to deploy in business-critical applications. By eliminating hallucinations, Alembic aims to make AI safe and trustworthy enough for enterprises to use for a wide range of data analysis, prediction, and decision support needs.
According to a diagram provided by the company, Alembic’s new AI system ingests data from various sources, processes it with an “observe and classify” module and a geometric data component, and then feeds the results into a causal graph neural network (GNN) to generate deterministic predictions and strategic recommendations. (Image courtesy of Alembic)
To achieve this feat, Alembic built its own supercomputing infrastructure and developed new mathematical techniques to represent enterprise data as a time-aware graph neural network. “Every time we look at one of these chain reactions or levers, we understand all of the fundamental components of our client’s business,” explains Puig. “They all become like little mini-neurons and are plugged into a giant graph neural network — but it’s a graph neural network that’s causally aware and time-aware.”
Causal inference engine drives deterministic AI
At the heart of Alembic’s breakthrough is a new type of graph neural network that acts as a causal inference engine: This AI brain ingests data from a wide range of enterprise systems, from sales databases and marketing platforms to analytics tools and even TV and radio, and organizes it into a complex web of nodes and connections to capture how different events and data points relate to each other over time.
“It’s like a 3D representation of your enterprise,” Puig told VentureBeat. “Imagine being able to see every interaction between every customer and every part of your business, and how those interactions ripple through your organization to drive outcomes.”
Importantly, Alembic’s AI doesn’t just learn patterns and correlations from this data, it identifies the causal relationships that actually drive business outcomes. By understanding the “why” behind past outcomes, the system can also predict the impact of future actions with a high degree of confidence and recommend optimal interventions to achieve desired goals.
A video showcasing Alembic’s technology shows how the company’s AI analyzes complex data and generates specific strategic recommendations, such as increasing investment in metaverse marketing campaigns based on strong interest indicators like video plays or download form submissions. (Video credit: Alembic)
The underlying neural network can extrapolate and generate predictions as new data points are added, simulating potential future impacts. “When you insert a new node, it predicts and generates what the chain reaction will be,” Puig says. This generation capability is built on Alembic’s custom “foundation models,” setting the company’s approach apart from the “mix of experts” employed by other enterprise AI vendors, which Puig dismissed as “just microservices.”
Strong interest from Fortune 500 companies and analysts
Puig said interest in Alembic’s AI breakthroughs has been so strong that the company has already signed deals with 9% of the Fortune 500 companies and has received private briefings and endorsements from Nvidia PhD graduates and other undisclosed large customers. [Forrester and Gartner]”They basically lost it. I’ve never seen anything like it. They had me go and meet with 26 analysts so far, both in IT and marketing communications.”
Alembic’s technology marks a key milestone in enterprise AI adoption: According to IDC, spending on AI technologies is expected to exceed $500 billion by 2024, yet trust remains a major barrier. If Alembic can truly deliver enterprise AI that business leaders can trust without fear of embarrassment or costly hallucinations, it could accelerate AI adoption across industries, from finance to marketing to manufacturing.
With strong interest from early customers and endorsements from leading analyst firms such as Gartner and Forrester, Alembic seems positioned to shake up the competitive enterprise AI market. But the company still faces the challenge of proving its technology can scale beyond early pilots and deliver tangible benefits to large enterprises. As the AI race heats up, Alembic’s “no hallucinations” approach could be a key selling point, but it could also serve as a cautionary tale about the gap between research breakthroughs and real-world impact.
VB Daily
Stay up to date! Get the latest news every day by email
Thanks for subscribing! Check out other VB newsletters here.
An error occurred.