Andy Berres

🌟 New Blog Alert! 🌟 Dive into our latest post where we sit down with Daniel Tran, an ETH Zürich grad and former MATLAB Student Ambassador. Discover how his academic pursuits in Materials Science paved the way to a promising career at Besi Switzerland. Don't miss his insights! #CareerJourney #BESI #MaterialsScience #StudentSuccess #MATLAB

3

Dive into the world of "Mathematical Modeling in Biology" 📚Explore how mathematics intersects with biology through ecological applications and expand your research skills across sciences. With practical MATLAB examples, this book is your gateway to experiencing the thrill of scientific research. #MathBiology #MATLAB #STEMeducation

15

We recently had a Research Away-Day for the Department of #Meteorology at University of Reading. Here's a blog article that I wrote about it. UniofReading undefined @UniRdg_Met #UofReading #ResearchAwayDay

5

1 Comment

"A Robot Walks into a Bar" - I heard this title so much I made it into a socio-technical systems and human-computer interaction study at Google DeepMind, with my fantastic co-authors Juliette Love, Kory W. Mathewson and Shakir Mohamed. Fresh out of press on arXiv and just presented at ACM FAccT. "A Robot Walks into a Bar: Can Language Models Serve as Creativity Support Tools for Comedy? An Evaluation of LLMs' Humour Alignment with Comedians" We interviewed 20 comedians who use AI in their artistic process, as part of 3-hour workshops on "AI x Comedy", conducted at the Edinburgh Festival Fringe in August 2023 (and online): * Comedy writing session with large language models (LLMs) * a human-computer interaction questionnaire to assess the Creativity Support Index of AI as a writing tool * and a focus group interrogating the comedians' processes of using AI * as well as their ethical concerns about bias and cultural appropriation, censorship and cultural erasure, and copyright. Participants noted that existing moderation strategies used in safety filtering and instruction-tuned LLMs reinforced hegemonic viewpoints, by erasing minority groups and their perspectives - a form of censorship. At the same time, most participants felt the LLMs did not succeed as a creativity support tool, by producing bland and biased comedy tropes, akin to "cruise ship comedy material from the 1950s, but a bit less racist". Our work extends scholarship about the subtle difference between, one the one hand, harmful speech, and on the other hand, "offensive" language as a practice of resistance, satire and "punching up". We also interrogate the global value alignment behind such language models, and discuss the importance of community-based value alignment and data ownership to build AI tools that better suit artists' needs. https://lnkd.in/ejg8e_Xf

183

41 Comments

👉Anthropic has succeeded in interpreting the internal workings of AI (LLM)! Anthropic has succeeded in interpreting the internal workings of the Claude 3 Sonnet AI model, addressing concerns about AI being a black box. This breakthrough marks the first concrete interpretation of a commercial large-scale model. By mapping neuron activation patterns into millions of human-interpretable concepts, they created a visual conceptual map. This allows for a better understanding of the model's predictions and behaviors, enhancing safety by identifying potential harmful actions in advance.

1

Part-Time Researcher to build A Multi-Modal Foundation Model POC for NASA’s Solar Dynamics Observatory (SDO) Trillium Technologies is excited to be working with NASA on SDO-FM: A Multi-Modal Foundation Model POC for NASA’s Solar Dynamics Observatory (SDO). We are looking for researchers and engineers with experience in building foundation models to join us on this journey. About the project The Solar Dynamics Observatory (SDO) is a NASA mission to understand the influence of the Sun on the Earth and near-Earth space by studying the solar atmosphere in extreme detail. SDO hosts three scientific experiments that combine to offer an unprecedented view of our closest star over a broad range of wavelengths. SDO’s combination of multiple instruments and a large temporal baseline make it an ideal candidate for a multi-modal foundation model - a new category of machine learning models trained with large-scale data so that it is generalizable for a large range of use-cases. An SDO foundation model would streamline many investigations, improve access to large datasets, and potentially unlock yet unidentified physical relationships. The SDO foundation model can be a powerful tool for the Heliophysics community where the integration of AIA, HMI and EVE would create an ML-enabled Digital Twin of the Sun’s complex physical interactions. The project will build an end-to-end SDO fFoundation model and adaptors which will be a trailblazer example for the safe application of AI to NASA data. Open Roles: ● Foundation Model Expert We are in search of talented and highly motivated researchers with backgrounds in generative AI and foundation models who can join the current team members who have machine learning and heliophysics expertise. In particular we are looking for researchers who have practical experience designing, building and validating foundation models as well as cutting-edge approaches for multimodal foundation model pre-training. The research team will follow an interdisciplinary research structure that optimizes the collaboration between technical know-how and domain expertise. The project has been segmented into three distinct stages; (Phase : Proof of Concept1) Data collection, engineering and model choice, (Phase 2) Pre-training and validation, and (Phase 3) Fine-tuning and release. This call for researchers is for the initial Phase 1: Proof of Concept only. Phase 1: Timeline April 2024 - July 2024 Responsibilities Work part time (10-20) hrs a week to complete the work packages (WP) during the project timeline ● WP1: Data Collection, Pilot task alignment and Preprocessing (10 weeks) ● WP2: Design of Validation (4 weeks) ● WP3: Model choice and evaluation (6 weeks) ● WP4: Science Definition Meeting (1 week) June ‘24 Link to apply: https://lnkd.in/emrDKwkQ

5

I was on a panel at the National Science Board yesterday with Eric Horvitz, Bart Selman, and Lynne Parker about AI and how to grow it best. You can see the video here: https://lnkd.in/gaBvM2Kg. #AI #NSF #NSB

36

4 Comments

If you are interested in presenting "work in progress" that intersects computing and society, consider submitting a talk to ACM SIGCAS Works In Progress (WIP). These are online discussions where you can present your work, discuss it with the SIGCAS community, and gain insightful feedback. You can submit a short talk (10 minutes), long talk (30 minutes) or full talk (45 minutes). To submit a talk, complete the following form: https://lnkd.in/euGZRjZp We encourage researchers, professionals, and students to participate! #computing #computerscience ACM, Association for Computing Machinery

6

If you didn't get a chance to attend the NSF CISE Career Workshop the slides from all of ours talks are posted on the materials page: https://lnkd.in/g3dRcxEx And Josiah made a great Overleaf template: https://lnkd.in/gNdUG2kY Some key takeaways from our CAREER Awardee Talks: 1. The first page should tell your whole story - this is the first impression for the reviewer. You don't want them to have to look any further into the proposal to be sold on your idea. 2. We all had different degrees of preliminary work. Unfortunately, I don't think there is a hard metric for this - and I think it further emphasizes that the CAREER is more about your vision than anything else. 3. Have a baby? Sleep deprivation may lead to transformative research ideas - caveat: this is based on a small sample size (3 out of 4) so maybe not a strong recommendation d=

42

𝗠𝗮𝗻𝗮𝗴𝗶𝗻𝗴 𝗲𝘅𝘁𝗿𝗲𝗺𝗲 𝗔𝗜 𝗿𝗶𝘀𝗸𝘀 𝗮𝗺𝗶𝗱 𝗿𝗮𝗽𝗶𝗱 𝗽𝗿𝗼𝗴𝗿𝗲𝘀𝘀 Released yesterday, this will most likely remain baseline thinking for near future. "Drawing on lessons learned from other safety-critical technologies, authors layout comprehensive plan that combines technical research and development (R&D) with proactive, adaptive governance mechanisms for a more commensurate preparation." Authors - YOSHUA BENGIO, GEOFFREY HINTON, ANDREW YAO, PIETER ABBEEL et al,. #ml #machinelearning #ai #research #rai #artificialintelligence #generativeai (Link in comments)

5

1 Comment

A Nature Communication paper I co-first authored: https://lnkd.in/eCp5TxGE A droplet bounces back from what we call superhydrophobic, but it may fail so, if you leave a hollow in its heart. Wow, look at the capillary waves riding the thin and thick, yet all mystery is solved by a model, that can't be simplified.

18

2 Comments

New ODE (ordinary differential equation) solvers in #MATLAB from the SUNDIALS suite, provides the ability to perform parameter sensitivity analysis. Learn more here 🔗

7

🌟 New Blog Alert! 🌟 Dive into our latest post where we sit down with Daniel Tran, an ETH Zürich grad and former MATLAB Student Ambassador. Discover how his academic pursuits in Materials Science paved the way to a promising career at Besi Switzerland. Don't miss his insights! #CareerJourney #BESI #MaterialsScience #StudentSuccess #MATLAB

3

It's an active day on the arXiv for the QSI team! Two papers emerging from the DARPA Quantum Benchmarking program (https://lnkd.in/guubMFvf and https://lnkd.in/gs9KgdT3) and another (https://lnkd.in/gJC6wbRR) by Zixin Huang a fellow at Macquarie and long-term visitor to our Centre. In "Quantum computing for corrosion-resistant materials and anti-corrosive coatings design" our team worked with collaborators at HRL, Boeing, Wisconsin, and MIT LL to investigate how quantum computers could be used in existing workflows to discover new anti-corrosives. This has been a really fun project where we've been able to learn a lot from subject matter experts in materials design and modelling to understand how they are currently attacking this difficult, but highly valuable, problem. We look at two different types of corrosion, potential materials, and corresponding workflows and estimate the cost of using QPE as a key subroutine in these workflows using a new package - pyLIQTR https://lnkd.in/g8WrCySs and a host of other tools to set up the instances. Not surprisingly, we find that the costs are high! We intentionally focussed on improvements to existing workflows and techniques so that we could get an understanding of the costs of heuristic methods without dependency of new breakthroughs in quantum algorithms in this space. I personally believe that there is a lot of opportunity to bring down the cost of such computations through improved workflows, algorithms, and analysis. This is a great challenge problem that we hope will motivate to improvements in quantum algorithms and software!

33

1 Comment

It's a pleasure to teach Physics to Audio Engineering students at Belmont University, especially when it comes to class projects. Here are two worthwhile take-aways from student researchers in this semester's Physics for Audio Engineering (PHY2010) class: 1. Cheap $6 knock-offs of the classic Shure SM57 & SM58 microphones match the frequency characteristics of their original counterparts to strikingly high degree of accuracy. However, experienced listeners were tipped off by differences in the transience responses, and the cheaper mics' intermittent tendency to emit a 16kHz tone. ---See comment though! 2. Bouncing in a DAW: Fast vs. Real-Time: They're identical. There was longstanding folklore/hesitancy to bounce in anything other than real time for fear of artifacts. Maybe that was once true -- and likely still is true if your bounce includes passing through outboard analog gear -- but nowadays all major DAWs (Pro Tools, Logic, Ableton) showed *zero* difference (i.e. -infinity dB on a sample level) between the fast/slow bouncing (when everything is in-the-box).

44

3 Comments

Check out our work on interpretable motion prediction at #ICRA2024!

11

This week, we're teaching #GenerativeAI at UC Civil and Natural Resources Engineering. Undergrads and postgrads in ENCN404 will be learning to apply #GenAI in #civilengineering situations. We'll also be teaching output verification, the risks of GenAI, and how these are minimised through ethical practices. Thanks to Daniel van der Walt for helping me develop this material. UC Engineering Saurabh Sinha

42