labarba.bsky.social
Engineering professor, computational scientist, jazz buff, techie, academic writer & font geek. Faculty director of the GW Open Source Program Office.
https://lorenabarba.com
www.linkedin.com/in/lorenabarba
94 posts
3,613 followers
385 following
Regular Contributor
Active Commenter
comment in response to
post
Here's a practical intro for anyone who values simplicity, openness, and wants to build their own online presence the #OpenSource way.
Full tutorial (~40 min): youtu.be/j-tXer7dIes
comment in response to
post
You can read all about my #GenAI adventures (and misadventures) here: doi.org/10.6084/m9.f...
Let's chat – what are YOU seeing in your classrooms? #teaching #python #highered
comment in response to
post
Key takeaway: We have to adapt. AI isn't going away. We need to find the line between helpful tool and crutch. And share our failures, not just our successes. It requires courage!
#genAI #learninginpublic
comment in response to
post
Tried to adjust the exam format, and the students revolted. 😬 It highlighted a key issue: are our assessments valid in the age of AI?
What now? I've scrapped homework and exams (yes I did!) and moved to in-class, collaborative exercises. It's better, but still a WiP…
comment in response to
post
The "illusion of competence"—that's when learners think they understand, but the knowledge isn't really sticking (like what happens after a "clear lecture"). AI gave them quick answers, but they didn't actually get the concepts. Sound familiar?
comment in response to
post
I really thought providing an AI tool trained on course materials (via RAG) would be a game-changer. Instead, students started using it to bypass actual learning. Attendance tanked. Class dynamics broke down as I tried to remedy learning loss…
comment in response to
post
Sorry to hear your grant was terminated 😕
comment in response to
post
The WH text plagiarized you?
comment in response to
post
The "Source code Harmonization And Reuse in Information Technology Act or the SHARE IT Act" requires agency contracts for custom-development of software to acquire and exercise rights sufficient to allow government-wide access, sharing, use, and modification #GWOSCON
www.congress.gov/bill/118th-c...
comment in response to
post
Amazing keynote showing how open source software can help deliver better health outcomes for America!
#GWOSCON
comment in response to
post
#OpenSource software is more secure! In 2024/25, the CMS OSPO partnered with CISA to ran a bug bounty: 200 vulnerabilities were submitted and 18 critical vulnerabilities identified and cleared. If exploited they would have cost millions of dollars!
comment in response to
post
Day 2 of the #GWOSCON launches with keynote speakers Andrea Fletcher and Remy DeCausemaker, leaders of the Centers for Medicare & Medicaid Services OSPO
comment in response to
post
Paul Barrett began advocating for the use of #Python in astronomy while working at Goddard Space Flight Center. In his talk, "The Early Years of Scientific Python," he notes that key to scientific Python was the slicing or indexing syntax, enabling multi-dimensional arrays! 🚀
#GWOSCON
comment in response to
post
Inessa Pawson from Open Teams highlights the barriers for participation in open source software within academia
comment in response to
post
Open source software used by NASA
#GWOSCON
comment in response to
post
I have given a talk about this paper at a workshop yesterday! Here are the presentation slides:
— Barba, Lorena A. (2025). Verification and Validation for Trustworthy Scientific Machine Learning. figshare. Presentation. doi.org/10.6084/m9.f...
comment in response to
post
The first talk "Leveraging machine learning techniques to support planning and emergency response management for storm surge risk" by Prof. Alex Taflanidis (Univ. of Notre Dame) was a revelation. Surrogate models for the win!
Side note: this important work was supported by NOAA, whose data is key.
comment in response to
post
You are too kind!
It's complemented by a post on Medium from around the same time: medium.com/code-like-a-...
comment in response to
post
youtu.be/_7Kp33xuzW8
Part 2, Linear Regression in Python with Jupyter—includes how to compute mean values using both custom functions and NumPy's built-in functions, obtaining regression coefficients, and how to calculate and visualize the best-fit line
#EngineersCode
comment in response to
post
youtu.be/zQtU8WIaYzg
Part 1 of a lesson on simple linear #regression with real data of earth temperature over time.
Bonus: how to create beautiful equations in #Jupyter with LaTeX and Mathpix
#EngineersCode
comment in response to
post
youtu.be/2M5kSM3PVug
#Python functions: basic concepts and syntax, use of the 'def' keyword, importance of docstrings; difference between local and global variables and how variable scope can affect your code. Practical examples & tips to avoid common pitfalls for beginners.
comment in response to
post
We have no idea how much the educational process needs to change. I feel left alone to face this without any institutional support. Just the word count of student emails has me overwhelmed: five paragraph emails to ask for late submission!
comment in response to
post
Our systems of peer review & publishing are crumbling.
Predatory publishers, exorbitant costs, unchecked bias, and AI are eroding trust.
It’s now more important than ever for scientists to be able to affordably and easily share their findings.
2/
comment in response to
post
I'm losing sleep over this quagmire we're in. Students are using AI to complete their work without thinking, and it gives them the *illusion of competence* (just like "very clear lectures" do). It's really urgent we find a way to teach them to work with AI effectively for learning support.
comment in response to
post
Oh, there’s a whole catalog of such papers. I went down that rabbit hole—read, “wormhole optimizer algorithm”—a few years ago…
x.com/lorenaabarba...
comment in response to
post
Students used the autograder combined with AI to "solve" coursework essentially by trial-and-error without thinking. Copy-pasted output from one-shot prompts was rampant despite my insistent pleas. Attendance cratered as students prioritized courses with closed-book paper exams. They learned little.
comment in response to
post
Our partners are IBL.ai, an edtech company that is building a solid product for generative AI in the education vertical—an open-source-software stack but not an open-source product. Can be source-available on-prem for clients who want that and pay for it.
Longer white paper: doi.org/10.6084/m9.f...
comment in response to
post
We can try via the system prompt to induce a response style that is more like a tutor and doesn't directly give answers for easy copy-paste, but if you lock down the behavior too much, students just use ChatGPT instead.
#ai4edu
comment in response to
post
It is an epic challenge finding a method to both embrace AI helping students develop skills to use it effectively, and avoid the harms of over-reliance or uncritical use. I made a course chatbot, using my materials via RAG—I don't regret it, as I learned a lot, but it failed
doi.org/10.6084/m9.f...
comment in response to
post
Indeed, many exams I reviewed had signs of copy-paste from an LLM with zero understanding of what is going on, leading to multiple errors. This semester was a total bust: students just don't know how to use AI as a "copilot" and one-shot-prompt their way through each task without thinking.
comment in response to
post
We face a similar challenge in all courses that have a coding component. On the one hand, we need to critically revise learning objectives when LLMs can code so well. On the other, redeveloping all assessments has huge labor costs! I’ve been losing sleep over this.
comment in response to
post
Generated the audio overview with #NotebookLM, using as source my lesson written on a Jupyter notebook (rendered in nbsanity and added via URL). Added the downloaded audio to a project in Descript, edited bits of the script out, added captions and video elements.
New one here! 👉 youtu.be/bnTguyi-aoE
comment in response to
post
I see. Thanks. I'll have to think about it. The images are in a folder of the same repo as the notebook, so in principle could be added via URL—yet for a user who clones the whole repo, not sure it makes sense for images to load remotely.
comment in response to
post
One thing I discovered is that nbviewer links don't work on NotebookLM, but those from nbsanity do work!
comment in response to
post
Does nbsanity render images embedded in Markdown cells via an <img> tag? It didn't work in this case: nbsanity.com/static/295a9...
comment in response to
post
I cannot imagine completing my engineering degree in Chile without reams of photocopying, nor developing my musical cultivation without bucketfuls of taping. Public-benefit arguments sound quite appealing to me 😬
(Apologies for commenting without reading the full 89 pages of your article…)