Very interesting indeed! I shared it on the social networked formerly known as Twitter and Blusky. I’m just curious: which is the reference for the Catalan book? I couldn’t find it in the Library’s catalogue. Catalonia was a major consumer and trader in Brazilian sugar and tobacco until the war of Spanish Succession and did two weeks of research there earlier this year about that.
Thanks for sharing and reading Thiago. Here is the catalogue entry - the book is a translation of Nicolas Lemery's "Traité universel des drogues" into Catalan, with unfinished illustrations and a new prologue (and I think some other changes or commentaries). https://wellcomecollection.org/works/t5raapw8
Very interesting. I am a very amateur historian working for my town historical society. Do your tools, or others that you know about, have any potential for getting historical insights from social media posts?
Thanks Benjamin. It seems Chatai is not that smart because it cannot keep track of my password updates. Chatai complains " this is not the password you originally signed up with !"
Very interesting article. Btw on the redaction test any type of transformer based NLP should be promising for these type of tasks because ‘cloze’ methods or masking is one of the primary ways they are trained. You might need to do some ‘fine tuning’ - that is supplement the basic GPT with an external data set to supplement the specific knowledge in less well known areas. Also some of the other models like BERT or T5 might be able to do better because of slightly different ‘cloze’ implementation. Limited training data may also be the challenge in identifying the 7 people in the photo. Possibly some of the figures may have been too obscure to have had enough training examples and some sort of augmentation could assist.
Very interesting indeed! I shared it on the social networked formerly known as Twitter and Blusky. I’m just curious: which is the reference for the Catalan book? I couldn’t find it in the Library’s catalogue. Catalonia was a major consumer and trader in Brazilian sugar and tobacco until the war of Spanish Succession and did two weeks of research there earlier this year about that.
Thanks for sharing and reading Thiago. Here is the catalogue entry - the book is a translation of Nicolas Lemery's "Traité universel des drogues" into Catalan, with unfinished illustrations and a new prologue (and I think some other changes or commentaries). https://wellcomecollection.org/works/t5raapw8
Phenomenal. I wish I’d had this tool when I did my PhD in History.
marvellous stuff, thank you!
Very interesting. I am a very amateur historian working for my town historical society. Do your tools, or others that you know about, have any potential for getting historical insights from social media posts?
Thanks Benjamin. It seems Chatai is not that smart because it cannot keep track of my password updates. Chatai complains " this is not the password you originally signed up with !"
Cannot access historians friend. Keeps giving me a Bitfender nag ad.
OpenAI changed their URL, which seems to have broken the old link. This one should work: https://chatgpt.com/g/g-u5cMhl7RY-the-historian-s-friend
Very interesting article. Btw on the redaction test any type of transformer based NLP should be promising for these type of tasks because ‘cloze’ methods or masking is one of the primary ways they are trained. You might need to do some ‘fine tuning’ - that is supplement the basic GPT with an external data set to supplement the specific knowledge in less well known areas. Also some of the other models like BERT or T5 might be able to do better because of slightly different ‘cloze’ implementation. Limited training data may also be the challenge in identifying the 7 people in the photo. Possibly some of the figures may have been too obscure to have had enough training examples and some sort of augmentation could assist.