Reflections on Machine Translation // Douglas R. Hofstadter

Douglas R. Hofstadter, College of Arts and Sciences Distinguished Professor of Cognitive Science and Computer Science, Adjunct Professor in Compara- tive Literature und Direktor des Center for Research on Concepts and Cognition an der Indiana University Bloomington versteht sich selbst als einen Grenzgänger zwischen Natur- und Geisteswissenschaften, zwischen Physik, Mathematik und Computer Science auf der einen und Musik-, Sprach- und Literaturwissenschaft auf der anderen Seite. Im Mittelpunkt seines Interesses als Physiker und Kognitionswissenschaftler steht «the human mind» in all ihren Facetten.
Weitere Infos unter: amp.phil-fak.uni-koeln.de/

[@wyeth1023]

Thank you for uploading!

[@zhuyuanxjtu]

His translation of 鹿柴 is brilliant👍

[@haraldwolte3745]

Has anyone found Hofstadter comments on ChatGPT?

[@amiran5273]

In a recent interview his former student David Chalmers (perhaps even more famous than DH) said that DH confessed his scepticism might have been mistaken. That was around the time of GPT-2. kzbin.info/www/bejne/noG0e4iGapeJaqM

[@RamakrishnaAppicharla]

Amazing..🤩

[@wyeth1023]

I don't know who you all are but if you upload Douglas; I sub and hit the bell, I'll read about you tomorrow xD have a nice day ;)

[@AimlessWander32]

Academic courage!

[@roccococolombo2044]

DeepL does a much better job nowadays, here is DeepL translation into french. Besides the Jean/John problem, it is quite good. Petit Jean cherchait sa boîte à jouets. Il l'a finalement trouvé. La boîte était dans l'enclos. John était très heureux.

[@user-wd8wx5md5z]

It is very interesting and funny, but there is a problem: the language model has never been trained on classical chinese language. but only on modern putonghua, which is completely a different language. But anyway the translations in French and German are also ridiculous. And the point of the professor is still valid. The NLP models cannot "understand" the meaning of the words and only translate according to the context. Adding punctuation is futile. This kind of problem is however beind addressed with the use of ontologies and reasoning.

[@girottomatheus]

Hosftadter said the same thing about computer playing chess. Eventually we will prove him wrong on translation also.

[@heater5979]

You missed the part where Douglas explains that he is talking about the state of MT art today. Not some supposed future.

[@girottomatheus]

Heater Don’t get me wrong, I love the guy, but he doesn’t understand of APPLIED computer science. And also it has been 4 months that I saw this lecture, but I remember him saying ML is hopeless. I actually remember him talking about some guy who worked on ML in the 60-70s that gave up and said it was impossible. Please look at all video.

[@chuchaichu]

HOW DID HE MANAGE TO GET THE CHINESE TONES RIGHT? ??!!!!???!??!?!?! But the poem is not supposed to rhyme in Mandarin... It rhymes better in the 700 AD Chinese.

[@excitedaboutlearning1639]

His German accent sounded pretty good too.

[@veep5712]

A master of forgery, that’s why he is so obsessed with semantics but actually fairly superficial interpretations. He obviously has a lot of researchers but he doesn’t seem fully engaged or understand material, he presents stolen ideas.

[@veep5712]

Imposter syndrome is a facet of actually being an imposter who has ripped off others academic work.