Google introduced its new family of artificial intelligence models focused on the medical domain recently: Google Med-Gemini AI models.
One of the notable features of this AI model is its long-context abilities that allow it to process and analyse health records and research papers. Google claims that the AI models surpass GPT-4 models in benchmark testing.
For the Google Med-Gemini AI models, they are built on top of medical large language model Gemini 1.0 and Gemini 1.5 LLM. There are a total of four models: Med-Gemini-S 1.0, Med-Gemini-M 1.0, Med-Gemini-L 1.0, and Med-Gemini-M 1.5. All the models have been trained on billions of medical data.
Unlike general large language models, which take a more generic approach to the data ingestion process, a finance LLM will include hundreds of millions, if not billions, of parameters that define the resources that it should favour, the communications it should undertake, and the ‘on the job’ learning that it should conduct.
Like Google and its medical LLM, we are not far from a financial-gemini or a financial-GPT, that is, a financial LLM.
Applying LLM in finance is different than in healthtech. It is more challenging because financial documents often include dense numerical information and domain-specific jargon requiring advanced numerical processing and reasoning capabilities.
This means that financial LLM models need extensive domain knowledge before they can capture the nuanced implications of accounting and financial measures, economic indicators, and market trends.
This is further compounded by the speed at which financial markets operate, where real-time analysis is crucial but challenging to achieve.
However, models such as Mistral, Llama, or OpenAI are already being used across worldwide fintech to overcome those challenges and are soon to be further developed, lowering the barriers to adoption in the next 12 months.
And the great thing with financial LLMs is that they are immune to the usual critiques of hallucination.
The ‘hallucination’ factor is well known to those who work with language models, referring to the generation of content that appears coherent but lacks grounding in factual or contextual reality.
Financial AIs are far less prone to this, since their data ingestion range is far narrower in scope than a generalist AI.
Financial AI should be up, running, and widely available at a fractional cost within the next 12 months. Those financial AI will be capable of:
- Finance operations: draft onerous activities or have little analysis needed, for example, creating contracts, supporting credit reviews, editing preliminary drafts to help streamline the process.
- Investor relations: help with nearly all the quarterly earnings calls activities.
- Accounting and financial reporting: provide views on the first presentation for repeat amendments or edit the audit trails for reclassification memos during the month-end closures.
- Finance planning and performance management: view ad-hoc variance analysis of business structured or unstructured data sets for the firm, for example, comparing current years to plan, making reports for business partners to write narratives explaining their unit’s financial performance.
AI is set to shake up how finance companies deal with their customers. By personalising interactions and giving tailored advice, AI will make managing money a whole lot easier and more reliable.
Going forward, financial companies need to understand that AI is here for the long term and will significantly change the way businesses function across the industry. Embracing the changes sooner than later will allow financial companies to make the most out of the AI in financial services.
