ChatGPT and other natural language processing (NLP) chatbots have democratized access to powerful Large Language Models (LLMs) and provided tools that enable more sophisticated investment techniques and scalability. This is changing the best way we take into consideration investing and reshaping roles within the investment career.
The report is aimed toward portfolio managers and analysts who need to learn more about alternative and unstructured data and learn how to apply machine learning (ML) techniques to their workflow.
“Staying current with technology trends, mastering programming languages for analyzing complex data sets, and being aware of the tools that augment our workflow are necessities that will move the industry forward in an increasingly technical area of investment,” says Pisaneschi.
“Unstructured Data and AI: Fine-tuning LLMs to Improve the Investment Process” covers a few of the nuances of an area that’s rapidly redefining modern investment processes – alternative and unstructured data. Alternative data is different from traditional data – like financial reports – and is commonly in unstructured form like PDFs or news articles, Pisaneschi explains.
In order to achieve insights from this data, more sophisticated algorithmic methods are required, he advises. NLP, the subset of ML that analyzes spoken and written language, is especially suited to coping with many different and unstructured data sets, he adds.
ESG case study shows the worth of LLMs
The combination of advances in NLP, an exponential increase in computing power, and a thriving open source community has fueled the emergence of generative artificial intelligence (GenAI) models. Crucially, unlike its predecessors, GenAI is capable of making latest data by extrapolating from the information it was trained on.
In his report, Pisaneschi demonstrates the worth of constructing LLMs by presenting a case study of environmental, social and governance (ESG) investing and its use in identifying material ESG disclosures from corporate social media feeds. He believes ESG is an area ripe for the adoption of AI and where alternative data will be used to use inefficiencies and generate investment returns.
The increasing capabilities of NLP and the growing insights gained from social media data motivated Pisaneschi to conduct the study. However, he regrets that because the study was carried out in 2022, a few of the social media data used is not any longer free. There is increasing recognition of the worth of the information that AI corporations must train their models, he explains.
Fine-tuning LLMs
LLMs have countless use cases on account of their ability to be adjusted in a process called fine-tuning. When fine-tuning, users create tailored solutions that take into consideration their very own preferences. Pisaneschi examines this process by first outlining the advances of NLP and the creation of frontier models similar to ChatGPT. It also provides a structure for starting the fine-tuning process.
The dynamics of fine-tuning smaller language models versus using frontier LLMs to perform classification tasks have modified because the introduction of ChatGPT. “This is because traditional fine-tuning requires significant amounts of human-labeled data, while frontier models can perform classification with just a few examples of the labeling task,” Pisaneschi explains.
Traditional fine-tuning of smaller language models can still be more efficient than using large boundary models when the duty requires a major amount of labeled data to grasp the nuances between classifications.
The power of different social media data
Pisaneschi’s research highlights the facility of ML techniques that analyze alternative social media data. He points out that ESG materiality could also be more worthwhile in small-cap corporations due to latest ability to glean information from social media publications closer to real-time information than from sustainability reports or investor conference calls. “It highlights the potential for inefficiencies in ESG data, particularly when applied to a smaller company.”
He adds: “The research shows fertile ground for using social media or other public information in real time.” But more importantly, it shows how, once now we have the information, we are able to easily adapt our research, by breaking down the information and on the lookout for patterns or discrepancies in performance.”
The study examines the difference in materiality by market capitalization, but Pisaneschi says other differences may be analyzed, similar to industry differences or a special weighting mechanism within the index, to search out other patterns.
“Or we could expand the labeling task to include additional materiality classes or focus on the nuances of disclosures. The possibilities are only limited by the creativity of the researcher,” he says.
The survey covers which libraries and programming languages are Most worthy for various parts of the investment skilled’s workflow related to unstructured data and provides beneficial alternative open source data resources sourced from survey respondents.
The way forward for the investment career is deeply rooted within the mutual collaboration of artificial and human intelligence and their complementary cognitive capabilities. The introduction of GenAI could herald a brand new phase AI plus HI (human intelligence) saying.
