The Reserve Bank of Australia’s (RBA) recent foray into AI was motivated in part by the need to enhance their data collection methods, specifically on companies’ price-setting behaviour and flow-on effects on the economy. This approach utilised companies’ earnings call transcripts with Meta’s large language models and AI software to “listen” and compile company commentary on labour costs, consumer demand and supply shortages. The language model employed by the RBA is one of many available models, for example:
- OpenAI’s GPT (generative pre-trained transformer) model
- Google AI’s PaLM (Pathways Language Model)
- Meta’s LLaMA (Large Language Model Meta AI)
Using the gathered data for back-testing in inflation prediction models, the RBA found it matched up well with data collected through existing research methods, such as the bank’s liaison program (physical meetings with companies) and business surveys (e.g. National Australia Bank’s monthly survey). The timeliness of the data was surprising given the time lag and backwards-looking nature of earnings calls (calls usually reflect conditions from previous six months).
Australia’s National Electricity Market (NEM) is another area where timely data becomes more pertinent as the NEM undergoes a period of profound yet volatile transition, whereby centralised large fossil fuel generators are gradually replaced or complemented with a myriad of renewable energy generators, grid-scale batteries and demand response. This transition comes with a set of challenges unique to the given energy mix and technology deployed. Hence, timelier and more relevant data becomes crucial to better inform regulators, market participants and consumers. These issues have led to initiatives, such as one by the Australian Energy Regulator to introduce a Power-BI network performance dashboard to enable rapid and wider exposure to network data. Timely data can also help with coordinating demand response (voluntary reduction or shift in electricity use), providing stability and flexibility to the grid.
Combating inflation and the transition towards sustainable energy are but 2 of many pressing challenges facing our society. However, as timeliness of data shape our understanding of these challenges, so too will this help with a timely resolution.
Reference: