I am quite comfortable auditing and using LLM-generated code. I think I'm also comfortable auditing LLM-generated Indian food recipes. Here is one for a "creamy paneer dish with tomatoes." Someone should try to run it unmodified. (I did not.)
Ingredients:
- 1 pound paneer, cut into cubes
- 2 tablespoons ghee or vegetable oil
- 1 large onion, finely chopped
- 3 garlic cloves, minced
- 1 tablespoon grated fresh ginger
- 2 teaspoons ground cumin
- 2 teaspoons ground coriander
- 1 teaspoon turmeric
- 1/2 teaspoon cayenne pepper
- 1/2 teaspoon smoked paprika
- 1/2 teaspoon salt
- 1 can (14.5 ounces) diced tomatoes, undrained
- 1 cup heavy cream
- 1/4 cup chopped fresh cilantro
Instructions:
1. In a large skillet, heat the ghee or vegetable oil over medium heat. Add the paneer cubes and fry until lightly browned on all sides. Remove the paneer with a slotted spoon and set aside.
2. Add the chopped onion to the skillet and sauté until softened, about 5 minutes. Add the minced garlic and grated ginger, and sauté for another minute.
3. Add the ground cumin, ground coriander, turmeric, cayenne pepper, smoked paprika, and salt to the skillet, and stir to combine. Cook for 2-3 minutes, stirring constantly.
4. Add the undrained diced tomatoes to the skillet, and stir to combine. Bring to a simmer and cook for 5 minutes.
5. Add the heavy cream to the skillet, and stir to combine. Add the fried paneer cubes back to the skillet, and simmer for another 5 minutes, until the sauce has thickened and the paneer is heated through.
6. Garnish with chopped fresh cilantro before serving. Serve hot with rice or naan bread. Enjoy!
RT @StackOverflow:
AI giveth and AI taketh away. Our Head of Content @benpopper chats with Professor @emeryberger, who teaches computer science at the University of Massachusetts Amherst, to discuss recent advances in AI and potential ramifications of its use in education.
https://stackoverflow.blog/2023/02/10/does-your-professor-pass-the-turing-test-ep-537/
#TempleUniversity deactivated my health insurance and the insurance of all striking graduate workers today, effective immediately - a move no US university has EVER taken against striking workers.
We were not informed by the university this happened - workers found out when their insurance was rejected at doctor offices or when picking up prescriptions today.
Help us put #Temple on blast for this unethical and deeply harmful behaviour by sharing this post!
Interested in teaching CS at a small liberal arts school? Oberlin has two positions for two year visitors starting in Fall! https://jobs.oberlin.edu/postings/13534
New Techtonic episode: interview with Craig Silverman from @ProPublica about Google's ad business of "porn, privacy, and fraud" - and how this affects ALL of us.
Stream the show (interview starts at 5:31):
https://wfmu.org/archiveplayer/?show=123777&archive=229670
Playlist & listener comments:
https://www.wfmu.org/playlists/shows/123777
Download the podcast:
https://podcasts.apple.com/us/podcast/craig-silverman-and-googles-ad-business-of-porn/id1285537944?i=1000594801708
More Techtonic episodes: https://techtonic.fm
Wellesley is unbelievably lucky to have hired @avanhatt. They’ll get a top-notch compilers/systems/PL research talent, an incredibly strong teacher, and a natural leader all at once. 👏 https://discuss.systems/@avanhatt/109632057171449739
(reminder) APPLICATIONS DUE Tuesday 10 January 2023:
New #NSF-funded program in #AIEthics and #DataEthics at #NortheasternUniversity!
We'll bring in cohorts of approx. 10 #philosophy (or related) graduate students to #Boston for 10 weeks for each of the next 3 years (beginning this year) for training in #AIEthics and #DataEthics from our faculty and visiting scholars.
$12k stipend per student
More details+application info for summer 2023 at the program website.
https://cssh.northeastern.edu/ethics/summer-grad-training-program/
Phil Ting's #FreedomToWalk Act is now law in California. You can cross the street anywhere as long as you're not creating a hazard.
#Jaywalking, the fake "crime" created by automobile industry lobbying to blame victims of #TrafficViolence, no longer exists in California.
Enjoy your freedom to walk! https://ktla.com/news/california/new-law-allows-californians-to-legally-jaywalk/
I've done video exams for a few years now, and finally analyzed some (with Rachel Lim and Mia Minnes):
Post: https://jpolitz.github.io/notes/2022/12/27/video-exams.html
Full paper: https://jpolitz.github.io/docs/sigcse-2023-video-exams.pdf
We see misconceptions in video traces that are hard to see in static artifacts because they have to do with the *order* students did things in.
For example, the screenshot on this post has 3 different orders used to fill in values in traces of the same kind of loop.
Strategy B matches Java's semantics best, and there are hints that submissions of students in that group did better on the exam (and no, trace ordering wasn't on the rubric!).
From this analysis, I can think of loops to trace that distinguish these strategies. We can poke at if helping students pick a trace strategy helps with other learning, and I can highlight in class.
I wouldn't have noticed without these videos in the loop in their assessments.
Associate Professor of Computer Science at Northeastern University PRL.
Senior Scientist at Roblox Research.