A Generative Artificial Intelligence (GAI) service has just started operating at Unicamp, on an experimental basis. The tool, adapted to accept a variety of types of training data, will initially be offered to professors and technical-administrative staff, with the aim of allowing the university community not only to gain experience with this type of resource but also, by learning about its advantages and its defects, to develop a greater critical sense regarding the ethical and responsible use of this type of resource.
The platform is expected to gain acceptance in various areas of activity at Unicamp, such as teaching, research, and extension, as well as in the administrative, welfare, and legal sectors. IAG will assist users in carrying out tasks such as text production, translation, data analysis, and creation of standards. It will also assist in the preparation of preliminary technical studies and the creation of terms of reference, risk maps, etc., that is, in any bureaucratic, repetitive process that has already established and consolidated protocols and procedures, due to its ability to learn from large volumes of data, to which it can respond by producing individualized results.
Information and Communication Technology Director Ricardo Dahab: IAG will assist users in carrying out tasks such as text production, translation, data analysis and creation of standards
Unicamp’s Director of Information and Communication Technology, Ricardo Dahab, says that the policy adopted to govern the new service is being drafted together with the vice-chancellors. “We already have a document under discussion that basically deals with the use of this resource in teaching, with recommendations and warnings, which will be released soon,” he explains.
Hours
Delgado Filho pointed out that the creation of a process, whether bidding or technical references, follows this same logic. “I already have the rules that govern that procedure, with legislation that is also already consolidated and the models used at the University. The virtual assistant will be responsible for combining the rules and producing a document that complies with those rules,” he explained. “Of course, the user will read it and check for inaccuracies. The responsibility for the authorship of the document will always lie with the user, never with the tool.” According to the Detic advisor, the system “is in full compliance with Unicamp’s information security guidelines.”
Software Solutions Coordinator Edmilson Chiavegatto Detic’s office advisor Adauto Delgado Filho
The coordinator of the Software Solutions Division of the board, Edmilson Bellini Chiavegatto, responsible for leading the work of developing the system, emphasizes that the user can add files in order to improve the selection. “You can provide files and ask the system to prepare a preliminary technical study based on specifications contained in these files. It will do this, but you will have to evaluate whether [the result] is correct or not”, he states.
According to Chiavegatto, once the service is consolidated, the plan is to create AI hubs to serve specific sectors. “What we are implementing now is a more general tool. But we intend to create specialized assistants for the different sectors, such as a legal assistant, a personalized assistant for the Attorney General’s Office, one for the DGA [General Administrative Directorate] and others. This way, you will be able to train the AI [artificial intelligence] more deeply for that specific subject”, he explains.
Chiavegatto reports that the documents added to the queries will remain on the platform, a measure that should further speed up the demand and response process. “This will greatly increase productivity in these sectors,” he says.
