The founder of Geo: Human validation to improve the reliability of AI answers remains to be verified

In an interview with Crypto.News, the founder of Geo, Yaniv Tal, pointed out that online information had four weaknesses that made AI's answer unreliable: loss of source, authority flattening, hiding differences and repeating model creation errors. He stated that AI was not a question of truth, but of the Internet. The information lost its key context during its capture and re-issuing on the website, making statements received by the model impossible to trace to the original source. In addition, Tal mentioned that language models often combined competitive positions into one answer, covering up real disputes among qualified experts. Geo's solution is to separate the declaration from its author and supporting evidence and to record the relationship between the entries in order to rank the arguments according to the strength of the assessment while retaining the competing material。