Post by account_disabled on Mar 10, 2024 22:22:59 GMT -5
I don't know if BB-, to return to Star Wars, considered himself human, but I also don't think that LLMs are stochastic parrots , a difficult definition to attribute born from a group of critics of this software, who, like parrots, repeat and not they create. The criticism arises from the very widespread idea that LLMs generate texts "starting from the probability of distribution of words or sequences of words". However, as Aaron Brancotti, a developer very attentive to the philosophical and linguistic implications of AI, summarizes, "GPT- is not a gigantic T".
If it were, it would be much more similar to us than we tend to think, because we humans also generate texts starting from the probability of distribution of words or sequences of words , applying grammatical, syntactic, stylistic and sem Bosnia and Herzegovina WhatsApp Number antic rules. We too write what we know and if we have learned it from questionable sources we will repeat nonsense or a reduced and simplified version of reality. We too, if stimulated, improve the first draft, refine an idea, find new ones, change it.
Artificial intelligences are here for this: to help us do it better and better, starting from ideas and questions that no one else can ask. without information inside, aware, today, that a chatbot could unmask us by saying "I don't know, we don't know" much more quickly and comfortably than what happens after visiting several pages. Millennia ago, in the Phaedrus, this is how Plato made Socrates speak , regarding the written texts: "you might have the impression that they speak, almost as if they have the ability to think, but if you ask them about some of the concepts they expressed, with the intention of understand it, they give only one answer and it is always the same.
If it were, it would be much more similar to us than we tend to think, because we humans also generate texts starting from the probability of distribution of words or sequences of words , applying grammatical, syntactic, stylistic and sem Bosnia and Herzegovina WhatsApp Number antic rules. We too write what we know and if we have learned it from questionable sources we will repeat nonsense or a reduced and simplified version of reality. We too, if stimulated, improve the first draft, refine an idea, find new ones, change it.
Artificial intelligences are here for this: to help us do it better and better, starting from ideas and questions that no one else can ask. without information inside, aware, today, that a chatbot could unmask us by saying "I don't know, we don't know" much more quickly and comfortably than what happens after visiting several pages. Millennia ago, in the Phaedrus, this is how Plato made Socrates speak , regarding the written texts: "you might have the impression that they speak, almost as if they have the ability to think, but if you ask them about some of the concepts they expressed, with the intention of understand it, they give only one answer and it is always the same.