Organisations have been astonished by the ability of ChatGPT and other large language models (LLMs) to express complex ideas in articulate language.
However, as most LLMs are primarily trained on internet-based information their practical application is limited when it comes to enterprise models.
While there are many challenging issues involved in building and using generative AI systems trained on a company’s knowledge content, this article suggests the following three paths as the necessary foundations to begin leveraging LLMs within an enterprise
1 Start from Scratch – Bloomberg, recently announced that it had created BloombergGPT for finance-specific content and a natural-language interface with its data terminal.
Bloomberg has over 40 years’ worth of financial data, news, and documents, which it combined with a large volume of text from financial filings and internet data. Few companies have those resources available.
2 Fine Tuning an Existing LLM – Google, used fine-tune training on its Med-PaLM2 (second version) model for medical knowledge starting with Google’s general PaLM2 LLM and retrained it on carefully curated medical knowledge from a variety of public medical datasets.
3 Prompt Tuning an Existing LLM – fastest path to production When an organization is struggling to hire AI talent or is new to a data-first culture. Morgan Stanley used prompt tuning to train OpenAI’s GPT-4 model to provide the company’s financial advisors with accurate and easily accessible knowledge on key issues they encounter in their roles advising clients. The prompt-trained system is operated in a private cloud that is only accessible to Morgan Stanley employees.
While all three have pros and cons, prompt tuning appears to give the most time and cost-effective path to enable any employee — and customers as well — to easily access important knowledge within and outside of an organisation
Source: Harvard Business Review
Leave a Reply