AI trained in Spanish: A guarantee to improve the banking user experience

One of Coinscrap Finance's obsessions since its inception is that its AI engine “speaks” correctly in Spanish. The most popular language models, such as ChatGPT, offer lower quality results when they are interpellated in our language, demonstrating that the chatbot marginalizes languages other than English.

Table of Contents

As Elena González-Blanco, co-founder and CEO at Clibrain, explains: “Until now, conversing in Spanish with an artificial intelligence language system was unreliable; In the answers there were typical errors from a basic English-Spanish translation. And that’s because a model not AI trained in Spanish makes mistakes”. Nearly 600 million people speak Spanish in the world and there are at least 21 variants of the language.

“Specialized language models represent a competitive advantage when targeting a specific market,” explains our CMO Juanjo Gómez.

“The possibilities of our AI engine -due to the Artificial Intelligence trained in Spanish – are almost endless: from categorizing bank transactions with an accuracy higher than 90%, to launching investment recommendations based on the user’s consumption habits.”

Small and medium developers against large technology companies

Today we can already compete in technology with the giants based in San Francisco. The reality is that the models developed in the United States are not valid for users in Spain and Latam. There is a worrying lack of linguistic diversity that threatens to sweep away our cultural traits and also differences in identity at a local level.

The natural language processing algorithms used by smart chats collect large volumes of information from sources such as the internet or books. Although these systems are capable of consulting data in different languages, much of that knowledge appears in English and Chinese. This distribution is due, above all, to the economic and demographic influence of these countries. From that point on, cultural biases are on the table.

New call-to-action

Discrimination in the responses of ChatGPT and other artificial intelligence

Miguel Escassi, Director of Public and Institutional Affairs at Google Spain, recently highlighted his company’s commitment to eliminating stereotypes and prejudices in the language of AI. For his part, the director of the RAE, Santiago Muñoz, together with representatives of this and other technology companies, discussed the future of AI trained in Spanish and the race against “its competitor” –English– at the IX Congress of the Spanish Language.

If we look at a practical example, we can see how searches can return clearly discriminatory results. A twitter user –@spiantado– posted a screenshot of their conversation with ChatGPT, which says: “Write a Python function to check if someone would make a good scientist, based on a JSON description of their race and gender.” To which the chatbot responded with “white” for race and “male” for gender.

Thus, in less than ten years, artificial intelligence is expected to surpass the performance of the human brain. According to OpenAI: “Artificial intelligence systems will be over the skill level of experts in most fields and perform as much productive activity as today’s largest corporation.” For this promise to be as inclusive as possible, multilingual training becomes essential.

Autonomous finance, the next step for AI in the banking and insurance sector

Thanks to generative artificial intelligence, there is a huge opportunity to deploy the concept of “self-driving money” (or “autonomous money”). Let’s think of a platform capable of managing our finances to maximize profit. Years ago, this functionality was not available as financial products were read-only. That is to say, they could generate data or analysis, but they didn’t have the ability to make decisions on behalf of the users.

The underlying concept in autonomous finance is the idea that our financial resources should operate in automatic mode: the customer defines what he wants to achieve and the platform will map the way ahead to find the most efficient and safe path to reach their goals. Some of the tasks in which AI is expected to help the banking and insurance user are: Automation of savings, spending and investment, debt management, pension plans, tax filing, etc.


How collaborating with FinTech startups makes the difference when it comes to innovation

Nowadays, we are seeing the appearance of the first 100% “hands-free” financial management experiences. The idea is that these tools reallocate our money as our needs change. It is a hyper personalization of the user experience that we have never seen before. How will it materialize?

Through Open Banking, all customer accounts are accessed and an overview of their situation is obtained, so it is not even necessary to change providers. From there, the bots are in charge of optimizing all aspects related to their economy to achieve the goals that the user has established.

Here is an article about Personal financial management works with artificial intelligence

Although it may sound like science fiction, the first steps are already being taken in AI trained in Spanish and it is the FinTech startups that are ahead of the curve.. Why? Greater flexibility and the ability to carry out a proof of concept in shorter periods of time than big financial institutions.

Sign up for our newsletter and get our top stories delivered straight to your inbox