Jennifer Chien

Gone are the days when you would Google important things/life advice like "How to transition from SWE to ML Engineer" without running into overly optimized trite SEO Optimized articles. I find good and important information on Reddit/Twitter, but then, I feel the information is scattered and unorganized again. Introducing TextData, a new tool made by Kevin Ros on which I am contributing these days. It allows you to save your notes, bookmark those hundreds of tabs scattered across your desktop, and, most importantly, build public communities and creators (or organizers? haha) can come up with curated content compilations. The vision is to create a community marketplace where creators could come up and develop roadmaps, and organize gated knowledge/personal notes that eventually would benefit members. Do give it a spin! #BuildInPublic! What are your thoughts on the platform?

28

1 Comment

I’m a social sciences student who has never taken a single CS class before. With nothing but a few ideas, some encouragement, and a whole lot of trust in YouTube tutorials, I applied to #Hackthe6ix with zero coding experience on my resume and prayed my enthusiasm for new challenges would reflect in my words. Somehow, I was accepted. We were a team of mostly new hackers hoping to build a pencil-and-paper game we invented and played for hours the night before. We worked tirelessly; brainstorming, coding, and problem-solving late into the night alongside nearly 350 participants working on 85 different projects. 🏆 Against all odds, I’m thrilled to announce that our game, WikiChess, won Best Game Hack @ Hack the 6ix! 🏆 If I, a social science student with zero tech experience can win in a hackathon, anything is possible—you’ll never know until you take that leap of faith! Thank you to Hack the 6ix, Major League Hacking, Department of Computer Science, University of Toronto, Computer Science Student Union - University of Toronto for hosting this incredible event. A special thanks to Best Buy Canada for sponsoring our prize and all the sponsors for making this event possible. A huge thank you goes to my incredible teammates: Ray Bao, Chantal Zhang, and Torrey Liu; I am beyond grateful to have the best teammates in the world! You can learn more about WikiChess here: https://lnkd.in/eZDZXFdb. #Hackthe6ix #Hackathon #HackathonSuccess #MLH #CSSU #UofT #HackathonSuccess #GameDesign

42

8 Comments

For the past few weeks I've been writing and speaking about why I find AI grading problematic. If you've been playing along at home, you'll have seen my main arguments: - LLMs are technically incapable of understanding, evaluating, or making justifiable judgements on student work - Numerical grading, letter based A-F, and criteria scoring are all essentially random - Even with research showing that AI grading is "more accurate" than humans, there are valid ethical reasons to avoid But despite my opinions, I also know that AI grading WILL happen. In fact, it already is. In Texas, for example, educators were "surprised" by the announcement that "ChatGPT style" technology would be used to provide initial evaluations of written answers to the STARR exams. Well established tech and edtech platforms are scrambling to incorporate AI grading features. Whatever I think, AI grading is already here. But, it's not just me. The vast majority of the comments on my posts and others' articles in the past few weeks have shown that the sentiment towards AI grading from teachers is largely negative. Educators know that these technologies are problematic. Students want grades from humans, not AI. Keep pushing back against uses of AI (generative or otherwise) which you don't want in your classrooms. #AI #aiwriting

66

48 Comments

I’ve had a few encounters lately with Uber drivers asking me what I do for work, and when I tell them I work in the field of high-performance computing, they tell me they’ve never heard of that before. I realized that much of my LinkedIn network might be in the same boat — and that’s not to shame, HPC is a tech area that isn’t in mainstream news very often and isn’t taught at many colleges. So I’d like to take a moment to answer 3 questions that will hopefully spark questions and follow-up posts 1. How is a supercomputer different from a regular computer? 2. What is Oak Ridge National Laboratory? 3. What do I do? — 1. How is a supercomputer different from a regular computer? A supercomputer is basically a bunch of very powerful computers (think a desktop or gaming PC) that can talk to each other very quickly (10-1000x faster than accessing the internet in most cases). It isn’t just one super powerful desktop PC. Each one of these sub-computers is called a “node”. Depending on who you talk to, a supercomputer can mean 10 nodes or 10,000 nodes. So a supercomputer draws most of its capability from nodes communicating with each other — instead of 1 node doing 500 units of work, 500 nodes can each do 1 unit of work, then share the result to each other. — 2. What is Oak Ridge National Laboratory? ORNL is a Department of Energy research laboratory dating back to the Manhattan Project (World War 2). Today, ORNL has 8 core science areas (see www.ornl.gov/what-we-do for a full view) and 8“user facilities”. These facilities offer tools for scientists to investigate the most challenging research problems facing the world. Relevant to supercomputing is the Oak Ridge Leadership Computing Facility, OLCF, which was established in 2005 to specifically serve scientists addressing the largest and most challenging computational problems. OLCF currently houses 2 of the top 10 supercomputers in the world, as measured by a linear algebra-based benchmark called high-performance Linpack (HPL). — 3. What do I do? My primary role is 2-fold. The first is to test these systems when they are deployed, before we open them to users. With this large of an investment ($600m for our latest, Frontier), we make sure every detail of the contract is met. This is extended to also testing the supercomputer periodically to make sure it is still working as it should. The second role is to support the scientists on the system. When they report issues, I am part of a team of people who investigate and resolve the issue. In a nutshell, my job is to make sure scientists can do their work on our machines. Hopefully that helps anyone who didn’t know of HPC before now! If you have any questions, drop them in a comment or send them directly to me!

45

2 Comments

The gravitational force has long been understood through spacetime curvature, but what if there's a more fundamental explanation? I've provided a brief overview of that possibility in this blog. #QuantumMechanics #Spacetime #QuantumInformation #BlackHoles #TheoryOfRelativity #QuantumBits #ScientificExploration #FutureOfPhysics

1

I played around with OpenAI 's o1-Preview model (and exhausted my weekly tokens). Thoughts - The list of text based problems that humans can solve but LLMs can't is growing ever so small. For example, I only have an edge in programming and very hard logic/physics problems vs this model. This is the biggest jump in capability I've seen since gpt 3. But it is still just a tool, a productivity multiplier. Importantly, I believe we are a few iterations away from super-intelligence, which I define as a model that beats the leading human expert at a task 99.99% of the time. Human performance is only an arbitrary benchmark for ML models. This would have profound implications. With super-intelligence we could potentially solve problems like biological aging, make better medical equipment and massively increase the quality and quantity of our goods. The sky is the limit.

6