Borrar la información errónea: el nuevo gran reto de la Inteligencia Artificial - Diario Primicia
The incident involving Australian mayor Brian Hood illustrates the critical difficulty of correcting errors in artificial intelligence—a challenge that goes beyond simply removing data. The ability of AI to “unlearn” incorrect information is neither automatic nor trivial, exposing the fragility of current models when faced with requests for correction and highlighting the urgent need to develop new technical tools capable of making algorithms “forget” problematic records. This case underscores the direct relevance of open data and privacy in the AI era, particularly in light of the impending European data protection regulations. The scientific community is exploring “machine unlearning” to enable companies to delete specific information from their training datasets. However, experts warn that there is no magic solution; significant uncertainties remain regarding the internal workings of these systems and the ethical management of data collection, leaving both technical and legal issues unresolved. The situation demonstrates that transparency about the origin of data and the ability to audit and correct information are fundamental pillars of public trust. Although the spread of false information in this case was partially addressed through media exposure, this does not substitute for the need for robust, automated mechanisms. For the open data ecosystem, this phenomenon highlights that information integrity depends not only on its availability but also on the technical feasibility of ensuring its accuracy and the right to be forgotten.
Source: primicia.com.vePublished on 2023-12-12