Jeff Tharsen 康森傑
@tharsen.bsky.social
doting father, friend & ally; hyperpolyglot computational philologist & sinologist;
currently teaching AI, deep learning + multilingual NLP/NLU + HPC + humanities data science @UChicago, creating new methods for multilingual intertextuality、古聲韻學、文字學等等
currently teaching AI, deep learning + multilingual NLP/NLU + HPC + humanities data science @UChicago, creating new methods for multilingual intertextuality、古聲韻學、文字學等等
We've had wonderful success teaching humanists (and other non-CS students) the transformer architecture, incl. things like encoders vs. decoders, stochastic gradient descent, ReLU etc.
It's not the material, it's that most curricula are designed for CS folks and not for a more general audience.
It's not the material, it's that most curricula are designed for CS folks and not for a more general audience.
December 14, 2025 at 4:13 PM
We've had wonderful success teaching humanists (and other non-CS students) the transformer architecture, incl. things like encoders vs. decoders, stochastic gradient descent, ReLU etc.
It's not the material, it's that most curricula are designed for CS folks and not for a more general audience.
It's not the material, it's that most curricula are designed for CS folks and not for a more general audience.
Reposted by Jeff Tharsen 康森傑
We are actively trying to spread the word globally beyond English-speaking institutions. We have been attending conferences and other events outside the US and speaking with researchers from South America, Asia, and Europe. If you can help us get the word out, it would be appreciated.
December 13, 2025 at 8:56 PM
We are actively trying to spread the word globally beyond English-speaking institutions. We have been attending conferences and other events outside the US and speaking with researchers from South America, Asia, and Europe. If you can help us get the word out, it would be appreciated.
Reposted by Jeff Tharsen 康森傑
Here are the top 10 most popular post-1945 American authors at the Seattle Public Library over the last 20 years.
December 11, 2025 at 7:00 PM
Here are the top 10 most popular post-1945 American authors at the Seattle Public Library over the last 20 years.
I find the LLMs generally produce about a C paper; sometimes a bit better, sometimes a bit worse.
My tactic has been to tell them this out loud in class and tell them to get a decent grade they'll want to write it using their own brains, using the machines for the thinking part doesn't really work.
My tactic has been to tell them this out loud in class and tell them to get a decent grade they'll want to write it using their own brains, using the machines for the thinking part doesn't really work.
December 11, 2025 at 3:31 AM
I find the LLMs generally produce about a C paper; sometimes a bit better, sometimes a bit worse.
My tactic has been to tell them this out loud in class and tell them to get a decent grade they'll want to write it using their own brains, using the machines for the thinking part doesn't really work.
My tactic has been to tell them this out loud in class and tell them to get a decent grade they'll want to write it using their own brains, using the machines for the thinking part doesn't really work.
Reposted by Jeff Tharsen 康森傑
My colleague points out here that possibly one in seven requests to our library now may be assisted by LLMs.
When those point to fictional documents, that's a huge staff time drain. There's often no quick way to check if an undigitized primary source doc exists beyond physically pulling some boxes.
When those point to fictional documents, that's a huge staff time drain. There's often no quick way to check if an undigitized primary source doc exists beyond physically pulling some boxes.
December 9, 2025 at 2:06 AM
My colleague points out here that possibly one in seven requests to our library now may be assisted by LLMs.
When those point to fictional documents, that's a huge staff time drain. There's often no quick way to check if an undigitized primary source doc exists beyond physically pulling some boxes.
When those point to fictional documents, that's a huge staff time drain. There's often no quick way to check if an undigitized primary source doc exists beyond physically pulling some boxes.