Latin America was now the cradle of the famous literary worldwide and Musical speciesBasic foods such as potatoes and inspiration behind Known Happy meal. It can also become a new cradle of artificial intelligence.
The research institutions coalition works on what they call Latamgpt A tool that can take into account the linguistic differences in the region, cultural experiences and “privacy”.
The goal of this is to provide users by throwing a more clear peek at the Americas and the Caribbean region more than the large linguistic models (LLMS) that come from us mostly or Chinese Companies have been largely trained in English.
“We want to develop our capabilities, find local solutions based on artificial intelligence and create a better understanding of these tools in Latin America and Latin America,” said Rodrigo Duran Rojas, director of the National Center for AI in Chile, which coordinates.
Duran Rojas said that for general purposes, the project will be pressed hard to compete with “artistic models that contain budgets with millions of millions”, but “what our model can offer is that others are not richer and more representative of Latin America and the Caribbean,” their people and their outputs.
For example, Durán Rojas said that the initial test showed that Latamgpt has achieved much better results when inquiring about the history of South America, and that the same thing is expected when LLM is required, for example, writing a poem in the style of local authors or providing an overview of regional education policy.
There are more than 30 institutions involved in the development of LatamGPT from countries throughout the hemisphere, and includes Latin collaborators in the United States like Freddy Felsh MinsAssociate Professor of Spanish Studies at Louis & Clark College in Oregon. He said this in appreciation of how “Latin experiences and Latin America are a cultural fellowship that goes beyond geography.”
“There are elements of Latin America in Oregon, in California, in Texas,” said Felsh Minsis. “We want to make sure that this Latin experience is also combined.”
LatamGPT, which aims to launch the first version available to the public in about June, was announced last month in the wake of a regional commitment that was submitted during a summit On Uruguay’s artificial intelligence to focus on the “moral, comprehensive and useful” technological development “in order to” enhance and protect human rights “and explore the best possible public policies of artificial intelligence governance.
This motivation follows an increased absorption in the technological progress area such as the use of Drones to monitor the removal of forests In the Amazon rainforest, the development of applications to encourage more people to continue learning Native languagesCreation Algorithms To help search for Disappear People or adopt Blockchain Mechanisms to preserve the historical documents of the previous dictatorship procedures.
Some of these reserved documents are now used as sources of LatamGPT training, as well as the papers, records and records available by institutions such as national libraries and archives specifically for the project. Duran Rojas said this gives the model more more translated differences and expansion than the general internet data that tends to use other systems.
“Latamgpt will have more context than other typical languages, and therefore it should calm down much” when it comes to use cases, “Duran Rojas said. Halosa is what artificial intelligence researchers call when a model apparently makes an incorrect or wrong answer although it was presented as realistic.
To date, the project data collection contains more than 8 terabytes of information so that the model can work on about 55 billion teachers (the variables of LLM to remove prediction, such as neurons that coincide or communicate in the human brain). Duran Rojas said this is somewhat close to the first public version of Chatgpt when it launched Openai in the fall of 2022.
Various dialects challenges and complex rules
ChatGPT and other models such as Google Gemini in recent years have sought to include a wider range of data to provide programs in languages other than English and “Emiratization” – such as LLM with the knowledge of the response in the metropolitan system when they are relevant or to understand expressions.
These companies recognize the importance of expanding this offer. “It is a dedicated experience” that can be “necessary for cultural and allergic replacement.”.
But they also realize that it is a particularly complex endeavor, because most of the training data available to them is in English. Choi said: “The complications of different languages can constitute a major obstacle for all artificial intelligence models … Languages with complex grammatical rules, various dialects or limited digital resources may be more difficult in training.”
Latamgpt, through its institutional networks with libraries and archives, has somewhat somewhat – but not completely. Duran Rojas said they are still struggling to integrate the original languages that millions speak in the region because the written documents are not widely available.
But they are still aiming to try because she constantly mastered its model – although it stresses the importance of cooperation.
“The quality and features of the results we can get dependent on us as Latin America joins to contribute as much as we can,” said Vilches Meneses, Professor Louis & Clark.
Currently, with the initial launch date of June, LatamGPT still receives data as collaborators regularly with specific questions to measure it compared to other models available.
Among the questions they test about inquiries in many different names and terms used in the region for a specific word such as “car”, or a request from GPT to provide a comparative scheme to how the countries of the region respond to collective migration from Places like Venezuela.
There is a great goal of Latamgpt is to identify these technological developments so that they can be included in public policies and regulations, according to Rojas Rorat.
Therefore, the creation of the Transcontinental network to help develop the project is essential, and Durán Rojas is likely to remain.
“The most important aspect, the greatest legacy, is this bond that we have found to promote and develop solutions based on artificial intelligence,” he says. “The model, I mean, it’s great to make it, but cooperation – this is what will affect how to build things to move forward.”
However, there is an increasing opportunity to make more contributions with a Latin touch.
“On its base, this commonly creates something from Latin America to Latin America and to obtain the world, as evidence of ourselves and others, we can also produce high technology,” said Vilches Meneses, “We can contribute to knowing artificial intelligence while we are still using social and cultural intelligence.”
A previous version of this story was published for the first time by Noticias Telemundo.
This article was originally published on NBCNEWS.com
