5:19 p.m.
Dear all,
“Programming to learn” is the idea of writing a program in order to better understand
something else. We have been working hard on a short and sweet “programming to learn”
activity that helps to better understand LLMs. Interested? Read on!
# Study over the holidays: Thank you & vouchers
Many of you participated in our study on errors and provided deeply insightful comments
and responses. Thank you for your efforts! We are working on getting those CHF 150
vouchers for SBB-CFF-FFS sent to you (sorry that this is taking a while). And we will let
you know about the results of the study in a future newsletter once we publish them.
Together with the paper, we will also publish the entire set of questions.
# Static checks and educational materials
The PyTamaro Web platform now supports live static checks. Those of you who participated
in the study already saw what this is all about. If you didn’t manage to take part, or if
you want to have another look, it is still possible to access the educational activity we
created for the study. The activity features an interactive discussion of fundamental
ideas such as syntax, static semantics, and dynamic semantics; it connects those ideas to
the errors (such as name and type errors) that your students see every day and to the new
optional checks available on the platform. You can find the activity here:
https://pytamaro.si.usi.ch/activities/luce/learn-errors/en/v1
Feel free to copy illustrations and examples if you find them useful for your teaching.
And even if you don’t use PyTamaro in your teaching but you find the static checks
pedagogically useful, you can always demonstrate the checks in your lectures on the
playground: https://pytamaro.si.usi.ch/playground
# Let your students program their own LLM (from scratch) in Python
Many students are constantly using ChatGPT & Co. – but do they know how those
“magical” products actually work? We built a short sequence of surprisingly simple
activities you could adopt in your classroom to provide your students with a basic
understanding of how a Little Language Model (LLM, ha!) works.
The activities accomplish two goals at once: students get to learn about LLMs, and they
get a little bit more programming practice in doing so. The activities don’t require
anything but some basic programming in Python, so they’re suitable for the mandatory
informatics course. No complex libraries, no neural networks, and yet we experience ideas
such as generative textual models, training & inference, model parameters, bias,
ethical and copyright issues with training data, and much more.
We have already piloted these activities with different audiences, and they were quite
well received. Would you be interested to join us for a workshop in which we can
experience these activities and discuss pedagogies to integrate them in courses?
We would like to understand how to best plan this so that the workshop can fit the
availability of those of you who are interested. The workshop is free and we will provide
food, beverages, and a contribution towards your train trip.
Could you take 30 seconds to help us out by filling in this form?
https://usi.qualtrics.com/jfe/form/SV_8GNYLUquNrz1LV4
Your answers will help us find a good date. We will let you know in the next issue of the
newsletter. Thanks!
That’s it for today. Now enjoy the Winter Olympics – and if you travel through Lugano to
attend any of the Olympic events, consider dropping by!
Greetings from Lugano!
Matthias and the LuCE team
PS: You can find this and the previous message in our mailing list archive:
https://lists.usi.ch/mailman3/hyperkitty/list/luce-news@lists.usi.ch/
If you would like to unsubscribe from our list, or if you would like to update your email
address, simply let me (Matthias.Hauswirth(a)usi.ch) know. You also can do it yourself
here:
https://lists.usi.ch/mailman3/postorius/lists/luce-news.lists.usi.ch/