Beyond LLMs: How SandboxAQ’s large quantitative models could optimize enterprise AI

While large language models (LLMs) and generative AI Although they have dominated business AI discussions over the past year, there are other ways businesses can benefit from AI.

An alternative are large quantitative models (LQMs). These models are trained to optimize specific objectives and parameters relevant to the industry or application, such as material properties or financial risk metrics. This is in contrast to the more general language comprehension and language generation tasks of LLMs. One of the leading proponents and commercial providers of LQMs is SandboxAQwhich announced today that it has raised $300 million in a new funding round. The company was originally part of Alphabet and was spun off as an independent company in 2022.

The funding is a testament to the company’s success and, more importantly, its future growth prospects as it seeks a solution Enterprise AI use cases. SandboxAQ has established partnerships with major consulting firms such as Accenture, Deloitte and EY to distribute its enterprise solutions. The main advantage of LQMs is their ability to address complex, domain-specific problems in industries where underlying physics and quantitative relationships are crucial.

“It’s all about core product development at the companies that use our AI,” Jack Hidary, CEO of SandboxAQ, told VentureBeat. “So if you want to develop a drug, a diagnostic, a new material, or do risk management at a large bank, then quantitative models are for you.”

Why LQMs are important for enterprise AI

LQMs have different goals and function differently than LLMs. As opposed to LLMs that process text data obtained from the InternetLQMs generate their own data from mathematical equations and physical principles. The aim is to address quantitative challenges that a company might face.

“We generate data and obtain data from quantitative sources,” Hidary explained.

This approach enables breakthroughs in areas where traditional methods have faltered. In battery development, for example, where lithium-ion technology has dominated for 45 years, LQMs can simulate millions of possible chemical combinations without the need to create physical prototypes.

Similarly, in pharmaceutical development, where traditional approaches face high failure rates in clinical trials, LQMs can analyze molecular structures and interactions at the electron level. In the financial services sector, however, LQMs eliminate the limitations of traditional modeling approaches.

“Monte Carlo simulation is no longer sufficient to handle the complexity of structured instruments,” said Hidary.

A Monte Carlo simulation is a classic form of computing algorithm that uses random sampling to obtain results. The SandboxAQ LQM approach allows a financial services company to scale in ways that Monte Carlo simulation cannot. Hidary noted that some financial portfolios can be extremely complex with all sorts of structured instruments and options.

“If I have a portfolio and I want to know what the tail risk is for changes in that portfolio,” Hidary said. “I want to create 300 to 500 million versions of this portfolio with minor changes to it and then look at tail risk.”

How SandboxAQ uses LQMs to improve cybersecurity

Sandbox AQ’s LQM technology focuses on enabling companies to develop new products, materials and solutions rather than just optimizing existing processes.

One of the areas in which the company has innovated is cybersecurity. In 2023, the company first released its Sandwich cryptography management technology. This has since been further expanded with the company’s AQtive Guard enterprise solution.

The software can analyze an organization’s files, applications, and network traffic to identify the encryption algorithms used. This includes detecting the use of outdated or faulty encryption algorithms such as MD5 and SHA-1. SandboxAQ feeds this information into a management model that can alert the Chief Information Security Officer (CISO) and compliance teams to potential vulnerabilities.

While a LLM could be used for the same purposeThe LQM offers a different approach. LLMs are trained on large, unstructured internet data, which can contain information about encryption algorithms and vulnerabilities. In contrast, Sandbox AQ’s LQMs are created based on targeted, quantitative data about encryption algorithms, their properties, and known vulnerabilities. The LQMs use this structured data to build models and knowledge graphs specifically for encryption analysis, rather than relying on general language understanding.

Looking forward, Sandbox AQ is also working on a future remediation module that can automatically suggest and implement updates to the encryption in use.

Quantum dimensions without quantum computers or transformers

The original idea behind SandboxAQ was to combine AI techniques with quantum computing.

Hidary and his team recognized early on that real quantum computers would not be easy to obtain in the short term and would not be powerful enough. SandboxAQ leverages quantum principles implemented through improved GPU infrastructure. Through a partnership, SandboxAQ has expanded Nvidia’s CUDA capabilities to include quantum technologies.

SandboxAQ also does not use transformers, which form the basis of almost all LLMs.

“The models we train are neural network models and knowledge graphs, but they are not transformers,” Hidary said. “You can generate from equations, but you can also get quantitative data from sensors or other types of sources and networks.”

Although LQMs are different from LLMs, Hidary does not see it as an either/or situation for companies.

“Use LLMs for what they are good at and then bring in LQMs for what they are good at,” he said.