Open-Source AI and the Future of Financial Institutions



At Asia Risk Congress yesterday, I shared my prediction that open-source AI will redefine the future of financial institutions in Asia. Going forward, institutions will shift towards open-source models they can fine-tune, manage, and scale internallyโ€”giving them a competitive edge in a rapidly evolving landscape.

At its core, AI is built on three key pillars: Data, Algorithms, and Compute. Here’s why I see open-source AI become a driving force, propelled by Metaโ€™s singular strengths:
1๏ธโƒฃ ๐——๐—ฎ๐˜๐—ฎ: Meta has unparalleled access to massive volumes of (your ๐Ÿ˜) text, images, and video data, vital for training the next generation of multimodal AI models.
2๏ธโƒฃ ๐—”๐—น๐—ด๐—ผ๐—ฟ๐—ถ๐˜๐—ต๐—บ๐˜€: Metaโ€™s strong engineering culture continuously pushes the boundaries of AI capabilities.
3๏ธโƒฃ ๐—–๐—ผ๐—บ๐—ฝ๐˜‚๐˜๐—ฒ: With immense financial resources from its core business, Meta has the infrastructure to support AI at scale.

๐—ช๐—ต๐—ฎ๐˜ ๐—ฑ๐—ผ๐—ฒ๐˜€ ๐˜๐—ต๐—ถ๐˜€ ๐—บ๐—ฒ๐—ฎ๐—ป ๐—ณ๐—ผ๐—ฟ ๐—ณ๐—ถ๐—ป๐—ฎ๐—ป๐—ฐ๐—ถ๐—ฎ๐—น ๐—ถ๐—ป๐˜€๐˜๐—ถ๐˜๐˜‚๐˜๐—ถ๐—ผ๐—ป๐˜€?
Just like with cloud computing a decade ago, financial institutions initially embraced it, but quickly realized that adoption was complex and costlyโ€”especially in a regulated environment. As a result, on-premises and hybrid infrastructures still dominate today.
I see a similar path for AI: the future will be ๐—ผ๐—ฝ๐—ฒ๐—ป-๐˜€๐—ผ๐˜‚๐—ฟ๐—ฐ๐—ฒ ๐—น๐—ผ๐—ฐ๐—ฎ๐—น ๐—”๐—œโ€”AI that institutions can control and manage themselves.

Metaโ€™s release of Llama 3.2 just two days ago coincidentally perfectly illustrates this. Llama isnโ€™t just another modelโ€”itโ€™s a comprehensive AI ecosystem that allows institutions to:
โžก๏ธ Download models directly from Metaโ€™s website.
โžก๏ธ Fine-tune models with torchtune.
โžก๏ธ Deploy AI systems using frameworks like ExecuTorch, TorchChat, and OLLaMA.

Ecosystems like LlamaStack are paving the way for financial institutions to start customizing and managing AI models in-house, driving deeper AI adoption.

Much like ImageNet accelerated computer vision and PyTorch democratized deep learning, LlamaStack will make open-source AI more accessible and customizable, further strengthening Meta’s moat.

Big tech will dominate in algorithms and compute power. So, if financial institutions want to compete, they need to focus on Data.
In Asia’s context of diverse languages and cultures, the real winners will beย those that have strong data foundations, upskill their teams to work with open-source models, and develop the capacity to leverage local, adaptable AI at scale.

Leave a Comment

Your email address will not be published. Required fields are marked *