es basicamente eso: las llm son modelos probabilísticos y la traducción fijo que tira por "por el contexto que tengo en este momento cuál es la palabra que es más probable que aparezca". También es posible que el traductor no haya dado el contexto necesario, porque dar contexto cuesta tokens.
![But Silke Schutz, 63, a baker who called herself a “staunch AfD voter”, described the Stolpersteine as “a waste of money, regardless of who pays for them”.
“There are so many victims no one cares about,” she said, citing people killed in traffic accidents. “We had nothing to do with the war and at some point it [remembrance] has got to come to an end. But when I express my opinion, I’m insulted as a Nazi.”](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:lgdgw47psj6lxfmkeggwlspp/bafkreifg2wdmfk7twt2g7wftf2tdlyqnkmqxqh6l5ayupjltd5y4tkgyq4)
