Writer's New AI Models Excel in Healthcare and Finance Tasks

San Francisco-based AI company Writer has launched two specialized large language models (LLMs) designed specifically for the healthcare and financial services sectors. The models, Palmyra-Med-70b and Palmyra-Fin-70b, are positioned to transform how these heavily regulated industries adopt artificial intelligence.

Available as open-source offerings on major AI platforms like Nvidia, Baseten, and Hugging Face, these models reportedly outperform larger, generalized AI models, such as GPT-4, in domain-specific tasks.

"In general, there are not enough domain-specific models because they are difficult to build," said Waseem Alshikh, CTO and Co-Founder of Writer. "These models require not just engineering, but also specialized data and expertise. You need industry experts to help develop them."

Specialized AI Models Outperform General-Purpose Counterparts

The introduction of these models comes as healthcare and finance grapple with navigating complex regulatory environments while leveraging AI's capabilities. Writer’s healthcare model, Palmyra-Med-70b, boasts an average accuracy of 85.9% across all medical benchmarks in zero-shot testing, outperforming Google’s Med-PaLM-2, which achieved 84% accuracy after multiple attempts.

"We have a healthcare specialist, Dr. Ziad, who helped us develop this model," Alshikh explained. "He ran tests that facilitated direct interaction with the model, creating a reward system for its learning process."

Pioneering Finance: Palmyra-Fin-70b Passes CFA Level III Exam

In finance, Writer's Palmyra-Fin-70b stands out as the first AI model to pass the CFA Level III exam, known for its difficulty in the investment management field.

“In financial use cases, our model excels, especially with commercial debt,” Alshikh noted. "The model engages like a financial expert, asking questions and providing customized financial plans."

Open-Source Strategy: Ensuring Transparency and Safety

Writer’s decision to make both models open-source marks a significant shift from the common closed-source approaches of many AI companies. This initiative is part of a broader movement in the AI sector, following similar efforts by tech giants like Meta and Mistral.

"We believe open source will allow our customers to scale effectively," Alshikh stated. "Relying on black boxes for critical operations is a precarious proposition."

This move towards domain-specific AI could alleviate ongoing concerns around AI safety and regulatory compliance. "To effectively address AI safety, transparency through open source is imperative," Alshikh added.

Looking ahead, Writer aims to develop more domain-specific models, including potential future projects in creative domains and enhancing existing models with multimodal capabilities.

Writer's focus on specialized AI for regulated sectors positions it competitively against established tech giants and emerging AI startups. As the industry evolves, Writer’s launch signifies a pivotal moment that may redefine how enterprises in highly regulated industries approach AI adoption. By merging domain-specific expertise with a transparent open-source strategy, Writer is paving the way for a future where specialized, trustworthy AI solutions become the standard in enterprise applications.