Meta’s New Open-Source Llama 3.1 and LLM’s for Fintech

AI | Jul 24, 2024

Image: Freepik/storyset

Llama 3.1: The Latest and Most Powerful Open-Source AI Model

Meta AI just announced the release of Llama 3.1. It's the most recent and powerful open-source AI technology, with 405 billion parameters that rival, if not outperform, the best proprietary models in some categories.

• Llama 3.1's key improvements include expanded context duration of up to 128K, multilingual support, and sophisticated tool use, making it ideal for difficult tasks like synthetic data generation and model distillation.
• Strategically, Meta has released Llama 3.1 as open-source tech which allows developers to fine-tune the model with their own data while maintaining complete control over its deployment, making development more flexibility and cost effective.
• Llama 3.1 includes new safety and security features such as Llama Guard 3 and Prompt Guard, which are intended to improve the responsible use of AI.
• The model is being introduced with the backing of over 25 partners, including major technology companies such as AWS, NVIDIA, and Google Cloud, making it available on a variety of platforms.
• Meta sees Llama 3.1 as a foundation for future AI work, promoting a strong open-source ecosystem.

Open-Source vs. Centralized AI Models

The release of Llama 3.1 highlights the ongoing debate about open-source versus centralized AI models, particularly in the financial sector where data security, compliance, and operational expenses are significant issues.

See:  Meta’s AI-Powered Metaverse to Kickstart New Era in Gaming?

Open-Source Advantages

• Open-source models can be openly accessed, modified, and deployed, allowing enterprises to tailor them to individual requirements. This is in contrast to proprietary solutions, which can come with costly licensing and restricted access.
• Open-source models offer full transparency, which is critical for ensuring trust and compliance in financial applications. The ability to inspect code and understand how the model processes data and makes decisions improves security and data integrity.
• Meta says that Llama 3.1 is up to 50% cheaper than proprietary models like GPT-4o, making it a cost-effective alternative for real-time and batch processing jobs.
• An open-source approach is akin to community-driven innovation and creates a collaborative atmosphere where updates and innovations can be quickly discussed and adopted, resulting in more robust and flexible models.

Centralized Advantages

• Proprietary models like GPT-4o frequently provide comprehensive support and are optimized for excellent performance, supported by significant resources and infrastructure. Alpha Architect suggests that they are better suited for large-scale deployments requiring stability and performance guarantees.

See:  Beyond QR Codes. Enhance Fintech Security with GenAI

• Proprietary models can provide superior security features and compliance assistance, which is critical for financial organizations that handle sensitive data.
• These models frequently include a suite of integrated tools and services, offering full stack solutions for a variety of business requirements.

Image: Freepik/DC Studio

Which Models Are Well Suited to Fintech Innovation?

Large language models (LLMs) are being used in fintech for a variety of applications such as financial analysis, sentiment analysis, fraud detection, and customer support. The choice of LLM is heavily influenced by the financial institution's specific demands and the type of data it handles.

See:  PwC Partners with OpenAI to Resell ChatGPT Enterprise

Here's a look at some LLMs that are appropriate for the fintech sector (there are more, be sure to do your research!).

1. FinGPT

FinGPT is an open-source LLM designed primarily for financial applications. It has been fine-tuned with financial datasets, making it particularly adept at tasks such as financial sentiment analysis, stock price prediction, and financial text processing. This model excels at financial activities, outperforming general-purpose models like GPT-4 in specific benchmarks.

2. BloombergGPT

BloombergGPT is another LLM tailored exclusively for the finance sector. Bloomberg created it using a vast dataset that includes both financial papers and general-purpose materials. This approach excels at activities including market analysis, risk assessment, and customer relationship management. FinPile, a unique dataset with over 700 billion tokens, provides accurate and contextually relevant insights.

3. FinBERT

FinBERT is a domain-specific modification of the popular BERT model that has been fine-tuned using a collection of financial communication data. It performs particularly well in sentiment analysis, entity recognition, and financial document classification. This tool excels in analyzing earnings reports and market sentiment due to its specialized training data as discussed at ar5iv.

4. Llama 3.1

While not specifically developed for fintech, Meta's Llama 3.1 provides extensive capabilities that may be tailored to financial applications. Its open-source nature enables modification and fine-tuning, making it a versatile alternative for enterprises wishing to employ AI solutions in finance without incurring significant licensing costs.

Considerations for Choosing an LLM for Fintech

When choosing a large language model (LLM) for fintech applications, below is an overview of various considerations.

• FinGPT and BloombergGPT models provide specialized features for financial tasks, including market analysis and sentiment identification, and have been trained on domain-specific knowledge.  More general yet capable models may require additional 'fine tuning' to achieve the same level of fintech expertise to handle nuances and complexities of financial content to optimize analysis and decision making.

See:  Responsible AI in Finance: CFTC’s New Framework

• Of course, Founders need to consider total cost and infrastructure.  While Llama 3.1 lowers initial costs by eliminating license fees, it necessitates extensive infrastructure for deployment and tuning. This requires both computational resources and qualified staff to manage and optimize the model.
• ]There is rising concern about the environmental impact of large-scale model deployment. Sustainable approaches, such deploying energy-efficient technology and optimizing model training, are becoming more significant.
• Data privacy, security, and compliance with regulatory norms is critical in the fintech industry. Open-source models enable more autonomy over data management, whereas commercial models often offer extensive compliance solutions.

The "best" LLM for fintech is not universal and is highly dependent on the specific application, data availability, and desired balance of model specialization and adaptability. Each of these models has distinct advantages that can be used based on the financial institution's specific requirements.

Conclusion

Llama 3.1, with its feature rich capabilities and open-source accessibility, offers a flexible and cost-effective solution. It demonstrates how open-source methods may encourage innovation while providing the openness and flexibility that proprietary solutions frequently lack. As the industry continues to unfold, open-source AI usage is expected to accelerate, establishing new benchmarks for accessibility and collaborative development in the financial sector and elsewhere.

The National Crowdfunding & Fintech Association (NCFA Canada) is a financial innovation ecosystem that provides education, market intelligence, industry stewardship, networking and funding opportunities and services to thousands of community members and works closely with industry, government, partners and affiliates to create a vibrant and innovative fintech and funding industry in Canada. Decentralized and distributed, NCFA is engaged with global stakeholders and helps incubate projects and investment in fintech, alternative finance, crowdfunding, peer-to-peer finance, payments, digital assets and tokens, artificial intelligence, blockchain, cryptocurrency, regtech, and insurtech sectors. Join Canada's Fintech & Funding Community today FREE! Or become a contributing member and get perks. For more information, please visit: www.ncfacanada.org

Canadian Public Markets Study Seeks Industry Input

Digital Wallet Control Points for Banks and Fintechs

OSFI Flags Frontier AI Risks Across Canadian Finance

FNZ Turnaround Tests La Caisse and Canadian Pension Capital

Senso’s Path From Fintech AI to Zeta Acquisition

National Security Is Now Part of Financial Supervision

How Canadians Pay Is Changing Before Real-Time Rail

Canadians Cut Holiday Spend but Still Pay More for Canadian

Ondo Launches Tokenized Pre IPO Private Markets

Snaplii Puts AI Agent Spending in a Separate Wallet