Open-source artificial intelligence (AI) models are rapidly closing the gap with proprietary systems, a shift that could dramatically reshape the business landscape by giving companies of all sizes access to powerful, affordable AI tools.
The AI industry is experiencing a transformation as open-source models narrow the gap with proprietary systems. The Allen Institute for Artificial Intelligence (AI2) recently announced its new Molmo family of multimodal models, marking another step in this trend.
Molmo reportedly represents AI2’s most advanced effort in creating open-source AI models. The multimodal system can process and generate different data types, including text and images, making it applicable across various industries and use cases. However, the path to competitive open-source AI models faces substantial hurdles.
“It’s important to note that training AI is very expensive. Thus, staying open-source and being able to pay the bills is quite challenging,” Simona Vasytė, CEO of Perfection42, told PYMNTS.
Data quality poses another challenge in developing these models.
“Open-source developers are not able to train their models on billions of data points,” Vasytė said. “For example, Molmo only used 600,000 data points, a fraction of a percent of what OpenAI did.”
This disparity in data access can impact the performance of open-source models compared to their proprietary counterparts.
The development of Molmo and similar open-source projects is part of a broader movement toward democratizing AI technology. These sophisticated models offer businesses access to capabilities once reserved for tech giants with vast resources.
The open-source AI ecosystem has seen a surge in new models, with notable releases like Meta’s LLaMA 2 and Stability AI’s Stable Diffusion XL. This proliferation of open-source options has accelerated innovation and democratized access to advanced AI capabilities, challenging the dominance of proprietary models from tech giants like OpenAI and Google.
Vasytė highlighted the critical role of open-source AI in fostering innovation.
“Having open-source alternatives is vital for accessibility and healthy development of AI systems,” she said. “Without open-source AI models, access to the most powerful solutions and the competitive advantage that comes with it would be only reserved to those with the biggest budgets.”
For the commercial sector, this shift carries significant implications. The availability of powerful, open-source AI models could level the playing field, allowing smaller companies to compete more effectively with larger corporations in AI-driven innovation.
“Open-source models make AI solutions accessible to everyone, meaning businesses can build solutions for themselves without having to pay a fortune to the big tech companies,” Vasytė said.
“Open-source models offer transparency that doesn’t exist with closed-source solutions,” Michael Berthold, CEO of KNIME, told PYMNTS.
“For instance, if you look in the context of data science, the transparency that comes with open-source software offers everybody the ability to understand how a data science system processes data and generates responses,” he said. “One of the big problems with many GenAI tools out there is that we don’t know how they are trained and optimized and have very limited understanding of the viability of performance claims. Open-source software in data science offers the ability to validate their inner workings.”
However, adopting open-source AI models comes with its share of challenges.
“The recently released details from the Office of the National Cyber Director around several projects looking to study open-source software (OSS) highlight the growing concern around its role in critical infrastructure,” Tony Baker, chief product safety and security officer at Rockwell Automation, told PYMNTS. “Though it has undeniably supported digital transformation, consumers of OSS content are realizing that additional effort and investment are required to successfully sustain that transformation.”
Baker highlighted the unique ecosystem of open-source software and the need for specific risk-management strategies.
“Unlike typical software providers, OSS works in a distinct ecosystem that demands particular risk management strategies for consumers,” he said. “To properly leverage OSS, both consumers and downstream users must take a more active role.”
Industries as diverse as eCommerce and healthcare may find new ways to leverage these technologies to improve their products, services and operations. Vasytė warned that “the performance can only be challenged if the quality of data is high. Only then developers can build high-performing models that provide less inaccuracies in their generations.”
The increasing capability of open-source AI models presents both opportunities and strategic considerations for businesses. Companies must carefully evaluate whether to invest in proprietary AI solutions or leverage open-source alternatives. This decision could have significant implications for their future competitiveness and innovation capabilities.
Looking ahead, Vasytė predicted: “Open-source models will help keep Big Tech companies in check — as long as developers have access to open-source models, companies like OpenAI or Google will be forced to keep not only their models competitive, but their prices as well.”
We’re always on the lookout for opportunities to partner with innovators and disruptors.
Learn More