Beyond LLMs: How SandboxAQ’s big models can grow the AI ​​industry

While the major linguistic models (LLMs) are generative AI have dominated the AI ​​industry conversation for the past year, there are other ways businesses can benefit from AI.

Another option is large quantity models (LQMs). These models are trained to achieve specific goals and parameters related to the company or application, such as physical assets or financial risk metrics. This is in contrast to language comprehension and LLM production activities. Among the leading directors and dealers of LQMs are SandboxAQwhich today announced that it has raised $300 million in new funding. The company was originally part of Alphabet and was it was created as a separate business in 2022.

The money is proof of the company’s success, and above all, its prospects for future growth as it appears to be resolved. business AI users. SandboxAQ has partnered with major consulting firms including Accenture, Deloitte and EY to share its business solutions. A major advantage of LQMs is their ability to solve complex, domain-specific problems in industries where physical foundations and quantitative relationships are important.

“It’s all about building companies that use our AI,” SandboxAQ CEO Jack Hidary told VentureBeat. “And if you want to develop drugs, diseases, new products or you want to manage risks in a big bank, that’s where the multi-dimensional models appear.”

Why LQMs are needed for business AI

LQMs have different objectives and work differently than LLMs. In contrast LLMs that use online dataLQMs generate their data from mathematical formulas and physical facts. Its purpose is to solve the problems that the business may face.

“We generate data and get data from many sources,” Hidary said.

This method helps to make progress in areas where traditional methods have stopped. For example, in battery development, where lithium-ion technology has dominated for 45 years, LQMs can accommodate millions of different chemical compounds without physical properties.

Similarly, in drug development, where traditional methods face the risk of failure in clinical trials, LQMs can analyze molecular structure and electron interactions. In economics, meanwhile, LQMs overcome the limitations of traditional methods.

“Monte Carlo simulations are no longer sufficient to deal with the complexities of structured materials,” Hidary said.

Monte Carlo simulation is an advanced type of computational algorithm that uses random samples to obtain results. With the SandboxAQ LQM method, a financial company can grow in a way that Monte Carlo simulations cannot. Hidary noted that some financial portfolios can be very complex with all kinds of instruments and options.

“If I have a profile and I want to know what the tail risk is given the change in the profile,” said Hidary. “What I want to do is I want to make 300 to 500 million copies of that profile and change it a little bit, and then I want to look at the risk of the tail.”

How SandboxAQ is using LQMs to improve cyber security

LQM’s Sandbox AQ technology is focused on helping businesses create new products, tools and solutions, rather than just complementing existing ones.

One of the business areas the company has been developing is cybersecurity. In 2023, the company released its first Sandwich cryptography management technology. This is also enhanced by the company’s AQtive Guard solution.

The software can analyze business files, applications and network traffic to identify the encryption algorithms being used. This includes detecting the use of old or broken algorithms like MD5 and SHA-1. SandboxAQ provides this information into a management system that can inform the chief information security officer (CISO) and stakeholder groups about potential issues.

When a An LLM can be used for the same purposeLQM offers a different approach. LLMs are taught on a broad, unstructured network, which can include a lot of cryptographic and complex algorithms. In contrast, Sandbox AQ’s LQMs are built using analytical, quantitative algorithms for encryption, their properties and known problems. LQMs use this well-structured data to create models and knowledge graphs specifically for cryptographic research, rather than relying on language understanding.

Looking ahead, Sandbox AQ is also working on future updates that can automatically update and implement updates in use.

Quantum measurements without quantum computers or transformers

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

Hidary and his team quickly realized that real quantum computers would not be easy to come by or powerful enough in the short term. SandboxAQ uses a number of rules that are implemented through GPU development. Through the partnership, SandboxAQ has expanded Nvidia’s CUDA capabilities to run more processes.

SandboxAQ no longer uses variables, which are the basis of almost all LLMs.

“The models we train are neural networks and knowledge graphs, but they are not transformative,” Hidary said. “You can build from equations, but you can also have a lot of data from sensors or other types of sources and networks.”

Although the LQM is different from the LLM, Hidary does not see it as a business problem.

“Use the LLMs for what they’re good at, then bring in the LQMs for what they’re good at,” he said.