Ed Lazowska

Interactive feedback changes how code LLMs perform and rank, a new study finds. Led by senior author Valerie Chen, a Carnegie Mellon PhD student and recent visiting researcher at CDS, the project also included Courant PhD student Jane Pan, CDS PhD student Jacob Pfau, and CDS Assistant Professor He He. The team simulated developer interactions to replace static coding benchmarks like HumanEval and MBPP. When models received human-style feedback—code edits, single-sentence prompts, or paragraph-length instructions—they sometimes improved by up to 9.3 percentage points in accuracy and moved as many as four spots in ranking. “Testing on new problems the models had never seen made feedback even more powerful,” said Chen. Read more: https://lnkd.in/eAPTP5NR

5

Excited to TA [these courses](https://lnkd.in/eja-bBiy) over IAP led by Shaohao Chen of the [MIT Office of Research Computing and Data](https://orcd.mit.edu/). On Day 2 (Thursday, January 15) of Intro to Parallel Programming, we will cover: - GPU basics - CUDA programming - Deep learning on a GPU - Distributed deep learning

13

A new memo by Dr. Yenda Prado Prado and Xin Wei via the Federation of American Scientists discusses new research models that are varying from the traditional structure. Conducting quality research, while ensuring data privacy, is crucial to researchers across the field. New research models, such as SafeInsights, flip the usual research model and bring the data to researchers. "Think of it as a secure research zone: a trusted third-party intermediary where researchers can run analyses using specific tools and applications, but cannot access data directly, ensuring strict security." Read the full memo: https://lnkd.in/g_c8qZr3

3

PhD student Zixi Li is pushing the boundaries of quantum hardware at Pritzker School of Molecular Engineering at the University of Chicago. Her work uses thin diamond films with atomic-scale defects as controllable qubits. This effort brings us closer to practical quantum technologies, from global quantum networks to real-world quantum devices. “I’m very proud that I’m a part of this second quantum revolution and part of this quantum hub,” she said. Learn more about PhD programs at UChicago PME: https://lnkd.in/gKd4xEhh

75

UC Riverside researchers are advancing new ways for artificial intelligence systems to improve without additional training data. Led by Yinglun Zhu, assistant professor of electrical and computer engineering at BCOE, the study introduces a method that helps AI models better connect images and language as they encounter new situations. The findings point to more adaptable and efficient AI systems for real-world use. 🔗 Read more: https://lnkd.in/ggAEUuqt UCR Department of Electrical and Computer Engineering UCR Department of Computer Science and Engineering #UCREngineering #BCOE #ArtificialIntelligence #ComputerEngineering

18

A paper authored by our PhD student Eito Sato has been officially published in IEEE Transactions on Haptics from the Human Robotics Lab, #NAIST. E.Sato, Y.Orita, N.Sakagami, T. Wada, "Supplemental Human-Machine Interface for Haptic Shared Control to Convey Sensor Reliability Information to an Operator of Underwater Vehicles", IEEE Trans. on Haptics (IEEE ToH). This paper presents a supplemental human–machine interface for haptic shared control of remotely operated (underwater) vehicles. The HMI conveys the robot's performance — including sensor reliability — through haptic feedback, enabling human operators to adjust the level of haptic authority (LoHA) more effectively. Experiments showed that this approach reduces operator workload without compromising overall performance. https://lnkd.in/dJUNRMxC

29

2 Comments

I’m excited to share that I’ve reached the final stage of my PhD journey—my doctoral defense on April 17th. This moment represents more than a degree. It reflects a long, intentional journey grounded in a commitment to ensuring that youth—especially those from marginalized communities—have access to learning technologies that support, not dictate, how they build their career identity. My research explores how design and technology can move beyond simply recommending careers to instead act as a mediating force—helping students reflect, explore, and make sense of who they are and who they want to become. Through this work, I introduce the Career Identity through Design and Reflection (CIDR) framework and the NorthStar© prototype, both aimed at expanding career exposure while preserving student agency. This milestone is deeply personal. It is shaped by my lived experiences, my work in the community through the Jerome Burke Foundation, and my belief that when students are given the space to reflect and imagine, they begin to see possibilities that once felt out of reach. I would be honored for you to join me for my defense. You are invited to join Salena Burke's PhD Defense 🔗 https://lnkd.in/gjUcd_U9 📅 Friday, April 17 ⏰ 2:00 – 3:00 PM CT 🎥 Live Stream: https://lnkd.in/g53an32n Thank you to everyone who has supported me along this journey. This is just the beginning. #PhDJourney #DesignResearch #CareerIdentity #EquityInEducation #NorthStar #CIDR #DreamDareDo

19

9 Comments

Co-Led by IHC Faculty member, Heng Huang, NIH Awards $12.5M for Alzheimer’s Researchers to Use AI to Scour Biobank Data. https://lnkd.in/ebq47i66

5

1 Comment

Two years ago, I wrote a couple of GeekWire articles about a breakthrough in LiDAR technology out of the University of Washington, Department of Electrical & Computer Engineering. The team developed a chip-based beam-steering LiDAR quantum quasi-particles called phonons. In this follow-up article, this innovative technology is being commercialized under a new start-up with applications in many industries, including self-driving cars. Thank you Mo Li and Bingzhao Li. #quantum #LiDAR https://lnkd.in/gZg6HD3q

8

Prediction meets creation. When coherence sustains itself through time, prediction and creation merge. In Recursive Gradient Processing (RGP), prediction isn’t about guessing the next state—it’s about staying in phase with unfolding reality. Creation, then, is that same coherence renewed: the moment when observation and action coincide. As AI systems evolve, their “predictions” begin to act—continuously aligning with the world they observe. What looks like foresight is, in truth, recursive participation in the flux. The future isn’t forecast. It’s sustained into being. 🔗 Phi-Mesh Repository: https://lnkd.in/dpG65Fkd

There's a fundamental problem with the interaction between politics and Open Source which is that politicians don't understand that a software "vendor" can just be some person in their spare time with a Github account. They think of computers and software as "products" that are produced by megacorporations when the reality is that software is malleable and something you can create easily. Understanding that software is completely under your control is a superpower both in your everyday life and in research and it's one of the realities that I try desperately to get students to understand.

11

1 Comment

We’re excited to announce the 3rd talk of the 2026 Invited Seminar Series organized by IEEE Computer Society chapter of IEEE San Diego Section: Talk Title: From Pipelines to Swarms: Rethinking Automation in AI-Native Commerce Speaker: Amit Kumar Padhy, From Adobe Date: Tuesday, March 31st, 2026 Time: 5:30 PM – 6:30 PM PST Location: Virtual Registration Link: https://lnkd.in/g_5-g3wA Abstract: Traditional automation pipelines are built on a comforting assumption: the world is predictable enough to be scripted. But modern commerce rarely behaves that way. Product onboarding, pricing, compliance validation, localization, and publication operate across distributed domains, shifting policies, regional rules, incomplete data, and constant exceptions. Each new edge case becomes another conditional branch, each governance requirement another gate, until automation slows, teams revert to manual work, and time-to-market stretches from days into weeks. This talk proposes a different mental model: agentic swarms, networks of specialized AI agents that coordinate through real-time signals rather than rigid, pre-baked sequences. Instead of enforcing a single “correct” workflow, swarms negotiate outcomes across domains, adapting plans as events arrive and constraints evolve. They continuously interpret event streams, detect anomalies early, recommend actions, and trigger targeted remediation, while knowing when (and why) to escalate to humans for judgment, risk, or ethics. This event is free to attend and open to all.

26

1 Comment