Home Artificial Intelligence ECB explores ChatGPT-style large-language models in ‘careful embrace’ of AI

ECB explores ChatGPT-style large-language models in ‘careful embrace’ of AI

Eurozone prices: AI is being used by the ECB to help ‘deepen understanding of price-setting behaviour and inflation dynamics’ | Credit: Joshua Miranda (Pixabay)

The European Central Bank (ECB) is exploring the use of ChatGPT-style large-language models (LLMs) as it gives the booming field of artificial intelligence (AI) a ‘careful embrace’.

The ECB’s main purpose is to handle monetary policy and maintain price stability across the 20 European Union (EU) member states that use the euro. It also supervises major banks through the Single Supervisory Mechanism (SSM).

In a newly published blogpost, the Frankfurt-based authority’s chief services officer Myriam Moufakkir provides a trio of examples of how the ECB is currently using AI across its functions. She also notes areas under exploration, including LLMs – a type of AI algorithm that uses deep-learning techniques and processes very large datasets to generate ‘human-like’ text.

Moufakkir titles her blogpost ‘Careful embrace: AI and the ECB’. The ECB is, she writes, working through the possibilities and challenges of ‘rapidly developing’ AI alongside other central banks in the European System of Central Banks (ESCB – which comprises the ECB and central banks of all 27 EU member states) and national competent authorities in the SSM, as well as through the Bank for International Settlements’ Innovation Hub.

The three AI uses mentioned are: to help ‘deepen understanding of price-setting behaviour and inflation dynamics’ in the EU – a particularly high-profile and sensitive area given relatively high current inflation rates; the use of machine-learning (ML) techniques to enable the automation of data classification; and for banking supervision.

RELATED ARTICLE ECB aims for AI-fired SupTech ‘cockpits’ in ‘smartphone for supervisors’ – a news story (3 October 2023) on a speech by ECB Supervisory Board member Elizabeth McCaul at the ECB-hosted ‘Supervision Innovators Conference 2023’

Gathering real-time product price data

In terms of helping to get a handle on eurozone inflation – which, although easing slightly in recent months, remains at a level more than double its two per cent target – Moufakkir writes that through applying web-scraping and ML, the ECB is ‘able to assemble a huge amount of real-time data on individual product prices.’

However, data collected are ‘largely unstructured and not directly suitable’ for calculating inflation, she states in the blogpost (which was published on 28 September). ‘Together with economists and researchers at the other euro area central banks – via the Price-setting Microdata Analysis network ­– we are therefore exploring how AI can help us structure these data to improve the accuracy of our analyses,’ she continues. The Price-setting Microdata Analysis network, also known as PRISMA, was set up in 2018 by the European System of Central Banks (ESCB).

In terms of using AI for data classification, Moufakkir explains that the ECB’s statisticians collect, prepare and disseminate data from more than ten million legal entities in Europe, which are classified by institutional sector (for example, financial institutions, non-financial corporations or the public sector). ‘We need these classifications to have the right data to support our decision-making. Doing this manually, however, is very time-consuming,’ she explains, stating that ML techniques ‘allow us to automate the classification process, meaning that our staff can focus on assessing and interpreting these data’.

On the topic of banking supervision, Moufakkir states that its teams ‘analyse a broad range of relevant text documents (for example, news articles, supervisory assessments and banks’ own documents)’ and that to consolidate this information, the ECB has created platform – ‘Athena’ – that helps supervisors ‘find, extract and compare this information’.

‘Using natural-language processing models trained with supervisory feedback, the platform supports supervisors with topic classification, sentiment analysis, dynamic topic modelling and entity recognition,’ she writes. ‘Supervisors can now collate these kinds of enriched texts within seconds, so they can more quickly understand the relevant information – instead of spending time searching for it.’

RELATED ARTICLE ECB looks to ‘Project Olympus’ to transform banking supervision using SupTech – a news story (20 September 2022) on a speech by McCaul one year ago (at the 2022 event)

‘A few possible uses’ of LLMs

Moufakkir, who joined the ECB earlier this year from the private sector, rounds off her blogpost by mentioning areas being explored and stating that the authority wants to quicken AI’s adoption.

‘Large-language models (of which ChatGPT is the best known) are another area which we are exploring,’ she writes, adding that the ECB has ‘identified a few possible uses for them’. ChatGPT, developed by US-headquartered company OpenAI, soared in profile earlier this year as user numbers rocketed.

‘They could be used to write initial drafts of code for experts for use in analysis, or to test software more quickly and thoroughly,’ states Moufakkir (of LLMs). ‘These models can also analyse, summarise and compare the documents prepared by the banks we supervise. This supports the work of our supervisory teams.’

‘The technology is also capable of helping to more quickly prepare summaries and draft briefings, which can assist colleagues across the bank in policy and decision-making activities,’ she continues. ‘A large language model can also help improve texts being written by staff members, making the ECB’s communication easier to understand for the public. Relatedly, we have used neural network machine translations for a while now to help us communicate with European citizens in their mother tongues.’

She closes by concluding that the ECB is ‘cautious about the use of AI and conscious of the risks it entails’ and states that the authority is ‘looking at key questions in the fields of data privacy, legal constraints and ethical considerations (such as fairness, transparency and accountability).’ But that nonetheless, the ECB wants to ‘pull together the various strands of our work on AI and accelerate its adoption across our organisation’.

RELATED ARTICLE SupTech on the rise: financial supervisors explore innovative technology – write-up of SupTech session at the Global Government Fintech Lab 2023 (the session featured representatives from Norway, Croatia and Bermuda describing how their authorities are using supervisory technology and barriers to greater uptake)

Central banks’ mounting AI interest

Moufakkir’s blogpost is published (28 September) as central banks worldwide, as well as private sector organisations, similarly wrestle with the challenges and opportunities created by AI.

The Bank of England (BoE), for example, published a discussion paper – ‘Artificial Intelligence and Machine Learning’ – almost a year ago covering how the UK central bank is using, and considering using, AI.

The BoE, the paper stated, is ‘us[ing] AI for predictive analytics, the study of non-linear interactions between variables, and analysis of larger and richer datasets, which can potentially help forecast GDP growth, bank distress and financial crises prediction.’ It added that the BoE was also exploring how AI-enabled text analysis of newspapers could help improve economic forecasting and ‘how AI could create “faster indicators”, which may enable real-time economic analysis’.

BoE staff have recently published articles on the BoE’s staff blogging site ‘Bank Underground’ on AI/ML applications, specifically ‘Can data science capture key insights in news articles?’ (on 27 September) and ‘Dissecting UK service inflation via a neural network Phillips curve (on 10 July). Its archive of staff working papers on ML techniques stretches back to 2017.

Also in the UK the chief executive of the Financial Conduct Authority (FCA), Nikhil Rathi, delivered a speech (‘Our emerging regulatory approach to Big Tech and Artificial Intelligence’) in July in which he said the authority was ‘using AI methods for firm segmentation, the monitoring of portfolios and to identify risky behaviours.’

The UK House of Lords communications and digital committee launched an inquiry in July examining LLMs and what needs to happen over the next one to three years to ensure the UK can respond to their opportunities and risks’.