Open-Source for AI Simulation and Modeling

Artificial intelligence (AI) is being incorporated into simulation solutions and new software options are entering the market. While open-source simulation tools can offer cost and deployment benefits, these tools can also pose challenges regarding implementation or integration.

Exploiting opportunities with open-source development has advantages and disadvantages. There are numerous ways to argue that open source, despite some challenges and shortcomings, could, in fact, be a game changer for best practices in AI simulation and modeling.

“There is no part of modern software that open source does not directly impact, so why would we expect a significant change in this next wave? Inarguably, the open collaboration and ability for the user to weigh in directly as to the usage and functionality of tools is a powerful benefit to the open-source model of software growth, and in the AI space in particular, is almost a necessity to ensure that we are not at the whim of a single entity’s models and assumptions,” says Chris Harrold, program director with Ansys.

“Certainly, you can make an argument for the burden of implementation being laid at the user’s feet in the case of open-source software (the price of ‘free’ as it were), but that masks the unlimited flexibility and power that comes from having access to the code itself,” Harrold says. “There is no possible scenario where an organization not already staffed with a deep level of AI expertise is going to be able to develop something from scratch better and faster than the AI open-source community has already done.”

Adnan Masood is the chief AI architect with UST, an IT services provider. He believes that using open-source software (OSS) is a viable and sound approach. Overall, an open-source approach is impactful in many ways to the future of AI’s role in simulation and modeling in the age of ChatGPT and large language models (LLMs).

“You can see how since the launch of ChatGPT, open-source large language models have already surpassed proprietary state-of-the-art models—now building upon this paradigm, in my professional assessment,” says Masood. “The integration of AI and agentic frameworks is indeed a game-changer for modeling and simulation.

“We would be leveraging natural language for model specification, validation, and interpretation, which would dramatically enhance accessibility and iterative development,” Masood adds. “The implementation of agentic frameworks using these LLMs would enable autonomous execution of simulations with nuance, incorporating diverse personas and complex constraints. This combination of AI and OSS is poised to create sophisticated system dynamics, agent-based modeling and discrete event simulation. The democratization of these tools, as seen with projects like Hugging Face’s open-source initiatives, could accelerate innovation in fields ranging from epidemiology to supply chain optimization.”

Rodrigo Domingues is the engineering director at CI&T, an information technology and software development company. Dominques says that it’s important to highlight that both AI and open-source initiatives are experiencing significant growth and that there is an increasing number of open-source models being released at an accelerating pace.

“Major organizations such as Microsoft and Meta have committed to advancing open-source AI models,” says Dominques. “This growing momentum and institutional backing could further strengthen the interplay between open-source initiatives, simulation solutions and cutting-edge AI models, making the connection between these three components even more powerful.”

Ankit Patel is senior director of developer marketing at NVIDIA. He acknowledges that simulation and modeling are becoming critical components of various industries, including healthcare, manufacturing, and climate science. But he says by using AI and OSS, developers can create sophisticated models that simulate complex systems, predict behavior, and optimize outcomes.

“By democratizing access, AI and [OSS] can unlock new possibilities for researchers, developers and entrepreneurs worldwide,” says Patel. “The Alliance for OpenUSD [AOUSD] is an open, nonprofit organization that aims to advance the standardization and development of OpenUSD, enabling interoperability of 3D content and large-scale simulation. The NVIDIA Omniverse platform built on top of OpenUSD, and the NVIDIA Modulus framework providing open-source tools for physics-informed neural networks, exemplify such collaborations. The ability for OpenUSD to be scaled and extended to meet real-world use cases that have different requirements than visual effects and animation is a key reason the technology has continued to grow in adoption across industries.”

Challenges of Open Source

An open-source approach poses challenges regarding implementation and integration. Collaboration with the open-source community is advantageous and may be a better practice than developing proprietary code.

“In my experience, open-source approaches enhance and democratize modeling and simulation efforts—and even though proprietary tools might have some feature advantages, they are limiting,” says Masood. “Using [platforms] like SimPy and OpenModelica are invaluable for collaborative development of complex simulations because their open nature allows rapid iteration for robust and innovative models. For instance, I’ve seen agent-based epidemiological models benefit greatly from community contributions that would be challenging with proprietary systems alone. Having said that, open-source modeling and simulation do pose some implementation and integration challenges, but I find the benefits of community collaboration, peer review and collective expertise, [that] help build adaptable solutions, far outweigh these issues.”

An essential part of ongoing success of an open-source approach is having the necessary tools and support solutions to make it possible, accessible and available for developers.

“The open-source approach is not without its challenges, particularly regarding implementation and integration,” says Jonathon Wright, chief technologist with Keysight Eggplant R&D Labs. “One significant issue is the lack of formal support and documentation, which can hinder the adoption of open-source tools in industry settings where reliability and robustness are critical. While the community can often provide assistance, this ad-hoc support is not a substitute for the dedicated, enterprise-grade support that typically accompanies proprietary software.”

Wright adds that integration can also pose significant hurdles, open-source simulation tools might lack standardization, leading to compatibility issues when integrating with other systems or software. He says this is particularly problematic in complex workflows where multiple tools must work seamlessly together. The variability in coding standards, documentation quality and software maturity across different open-source projects can exacerbate these integration challenges.

“The rapid pace of development in open-source projects, while generally a positive attribute, can lead to stability issues,” says Wright. “Frequent updates and changes can introduce bugs or alter functionality in ways that disrupt existing workflows,” says Wright. “In contrast, proprietary software often undergoes more rigorous testing and quality assurance processes before release, providing a more stable and predictable environment for users.”

The Democratization of Software Development

Open-source development has made its way into industries. The open-source development community continues to expand and grow. As it does, it’s likely that it will stimulate more free and open competition and hamper efforts for proprietary software.

“Open source has got to become a normalized way for companies to interact with their users, and for core technologies like AI [and machine learning] to be democratized,” says Ansys’ Harrold. “The fear of missed revenue on something that is not proprietary is, certainly in the case of AI, far outweighed by the practical and ethical concerns of a ‘cornered market’ that relies on a few providers for the majority of capability. Using commonly available, open-source tooling as a core for simulation and computer-aided engineering/modeling tools that leverage AI means that the playing field remains level for the user, who is the most important constituent in that ecosystem.”

More Ansys Coverage

More Keysight Technologies Coverage

More NVIDIA Coverage