Goldman Sachs Deploys Its First Generative AI Tool Across the Firm

🌈 Abstract

The article discusses Goldman Sachs' rollout of its first generative artificial intelligence (AI) tool for code generation, which will be made available to thousands of developers across the company by the end of the month. The article also covers Goldman's approach to centralized management of proprietary generative AI use, the benefits and challenges of this approach, and the broader adoption of generative AI in the financial services industry.

🙋 Q&A

[01] Goldman Sachs' Approach to Generative AI

1. What is Goldman Sachs' approach to using generative AI?

• Goldman Sachs is taking a centralized approach to using generative AI, with all proprietary uses of the technology managed on an internal platform.
• This approach has pros and cons - it may have slowed down the initial rollout, but has allowed the company to gain more velocity afterwards.
• The centralized platform allows Goldman to fine-tune the AI models with its own internal data in a safe and compliant way.
• The platform also enables developers to build custom applications on top of the models more quickly, as they can leverage existing safety guardrails and functionality.

2. What are some of the key benefits of Goldman's centralized approach?

• Ability to fine-tune the AI models with internal data in a safe and compliant manner
• Developers can build custom applications faster by leveraging the existing functionality and safety controls on the platform
• Allows the company to maintain control and oversight over the use of generative AI

3. What were some of the challenges or pushback that Goldman faced with this approach?

• Some employees wanted to move faster and adopt generative AI tools like OpenAI's ChatGPT more quickly
• The company had to push back against those who wanted to move faster, in order to ensure safety and compliance

[02] Generative AI Adoption in Financial Services

1. What are some of the key factors influencing the speed of generative AI adoption in financial services?

• The need to safeguard data and remain compliant with existing and upcoming data/AI regulations
• Financial services is one of the most regulated industries, which slows down the pace of AI adoption

2. What are some common starting points for generative AI use cases in financial services?

• Code generation is a common starting point, as it is text-centric and provides clear efficiency gains

3. How does Goldman's use of generative AI for code generation compare to other applications?

• Goldman's code generation tool, based on Microsoft's GitHub Copilot, is currently its most scaled-out use of generative AI
• The company is also exploring other applications like document translation and research summarization, but these are in earlier stages

4. What is the current scale of generative AI investment at Goldman Sachs?

• Generative AI remains a relatively small part of Goldman's overall technology budget, with most resources still dedicated to running the core banking operations and ensuring safety and compliance