Marinka Zitnik

I'm so so thrilled to see our first GothamQ paper formally published in my most favorite journal, PRX Quantum 🎉 (link at the end) I've published many papers as a grad student, postdoc, and now a group lead but this work is so unique to me. To be fully honest, for the first time I feel like we have done something really impactful! On the surface, this is a very simple work! We distributed entanglement in NYC. But in reality this paper is the result of 4 years of nonstop hard work of our physicists, engineers, and software developers. We had to invent and manufacture a new entanglement source that works at room temperature (hence actually useful for qu. networking) but has lots of advantages over all the other alternative options. We had to invent and manufacture the devices that very rapidly and with very low loss monitor the infrastructure for any imperfections. We had to write thousands of lines of codes so both devices work together automatically and without any need for manual optimization. And on top of all that we had to negotiate and build a quantum testbed in one of the world's most busy and chaotic cities. We did all that to prove one thing: Entanglement distribution is ready for prime time, beyond academic and research testbeds. The technology has reached a pivotal point where everything is robust and high quality for practical use cases by a much wider community than just quantum physicists. This shift is not only essential for us to be able to use entanglement for applications we already know of, but also to make it widely available for much more creative people than us to constantly think of use cases for quantum networks and entanglement links. I really hope if you are in this field, you enjoy reading this paper and as always don't hesitate to reach out to me for any questions. Goes without saying, this work is all thanks to our amazing team at Qunnect, and a huge congrats to all the authors Alexander Craddock, Anne Lazenby, Gabriel Bello Portmann, Rourke Sekelsky, Maël Flament https://lnkd.in/eteZNASt

246

11 Comments

OpenCoder: An Open-Source Code LLM Redefining Model Efficiency 🌟 The OpenCoder team has launched a game-changer in the world of code language models. Unlike many LLMs that focus on sheer size, OpenCoder brings something unique to the table: a data-focused training approach. By putting quality before volume, OpenCoder’s 1.5B and 8B models rival some of the biggest players, like DeepSeekCoder and QwenCoder, with only a fraction of their token count. The core strength of OpenCoder lies in its meticulous data processing pipeline and two-stage fine-tuning. With a training dataset composed of 2.5 trillion tokens, 90% raw code and 10% code-related web data, OpenCoder refines every aspect of its corpus. Advanced filtering and deduplication strategies, applied to 607 programming languages, optimize model efficiency and precision, ensuring high-quality data drives its training. Key Features: - Complete Training Transparency: Open-source datasets and training pipelines enable reproducibility and further research. - Instruction Fine-Tuning: Two-stage tuning sharpens both core CS concepts and real-world coding skills, prepping OpenCoder for production-level code generation. - Available on Ollama: Accessibility in deployment means more avenues for experimentation and improvement. OpenCoder's data-driven strategy showcases the potential of a leaner approach. By focusing on quality data rather than quantity alone, OpenCoder achieves file-level deduplication that outperforms traditional repo-level methods, making it both efficient and robust. 🚀 Read more here: https://lnkd.in/eEnG3yJA #OpenSource #MachineLearning #CodeLLM #AI #ModelEfficiency #DataQuality #Innovation #CodeGeneration #OpenCoder #TechNews #AIResearch #Programming #LargeLanguageModels #ArtificialIntelligence

1

Great to see the newly announced Bamba-9B, an inference-efficient Hybrid Mamba2 model 🐍 trained by IBM, Princeton, CMU, and UIUC on completely open data used PyTorch FSDP to train these novel architecture models and they are integrating inference with vLLM, which recently joined the PyTorch Ecosystem.

167

This surprising paper by Catarina G. Belem et al., titled "From Single to Multi: How LLMs Hallucinate in Multi-Document Summarization," points out the fundamental problems of Perplexity and related AI search platforms. https://lnkd.in/et_D_Axs - LLMs struggle to identify shared information. The study found that when explicitly instructed to summarize only information shared across multiple documents, LLMs actually hallucinated more frequently. (LLMs tend to make up rumours - when prompted to prevent them) - Hallucinations increase when summarizing non-existent information and LLMs frequently generated fabricated summaries. This indicates a concerning tendency to invent information rather than recognizing its absence. Notably, GPT-3.5-Turbo and GPT-4 exhibited this behavior more often than open-source models. (try it yourself!) - Simple mitigation strategies are only moderately effective. The researchers tested various post-processing techniques to reduce hallucinations, including filtering out unrelated or redundant information. While these methods showed some promise, the improvements were generally small. (this argument remains open. I am actively engaged in advancing this field and encourage those interested to reach out)

2

Massachusetts Institute of Technology let's do an earthshot! The institute plans to clarify it's decarbonization plans for MIT’s Campus by the end of Fall 2024 and I challenge them to be ambitious! Background: Last week a few students got together to write a letter to MIT to be inclusive and ambitious in their goals and I 100% support them. In all my time at MIT folks always an attitude of pushing the envelope, wholly knowing they may not be successful, but that the juice was worth the squeeze. On this issue, MIT has the opportunity to be a leader. The letter, linked below, calls on MIT to aim for an “Earthshot” goal of 2035 for campus decarbonization, to set intermediate milestones, and to be more inclusive of students in climate initiatives. Please consider reading through and signing on if you’d support these goals and being part of the decarbonization efforts on campus! Please read the full letter here: https://lnkd.in/gmRiC9iw If you're affiliated with MIT, consider signing the letter here: https://lnkd.in/gm6Ppc5w

13

My dissertation is now available on arXiv. You can read it here: https://lnkd.in/eseT6-Xb Rethinking Video with a Universal Event-Based Representation Abstract: Traditionally, video is structured as a sequence of discrete image frames. Recently, however, a novel video sensing paradigm has emerged which eschews video frames entirely. These “event” sensors aim to mimic the human vision system with asynchronous sensing, where each pixel has an independent, sparse data stream. While these cameras enable high-speed and high-dynamic-range sensing, researchers often revert to a framed representation of the event data for existing applications, or build bespoke applications for a particular camera’s event data type. At the same time, classical video systems have significant computational redundancy at the application layer, since pixel samples are repeated across frames in the uncompressed domain. To address the shortcomings of existing systems, I introduce Address, Decimation, ∆t Event Representation (AD∆ER, pronounced “adder”), a novel intermediate video representation and system framework. The framework transcodes a variety of framed and event camera sources into a single event-based representation, which supports source-modeled lossy compression and backward compatibility with traditional frame-based applications. I demonstrate that AD∆ER achieves state-of-the-art application speed and compression performance for scenes with high temporal redundancy. Crucially, I describe how AD∆ER unlocks an entirely new control mechanism for computer vision: application speed can correlate with both the scene content and the level of lossy compression. Finally, I discuss the implications for event-based video on large-scale video surveillance and resource-constrained sensing.

15

This is a fascinating paper evaluating DeepMind’s AlphaFold 3 and its approach to predicting ligand docking: https://lnkd.in/gieM2nW5 The authors used a clever method to evaluate AlphaFold, finding that it can make glaring mistakes when placing molecules in binding sites because its model is likely only fitted to the protein’s backbone geometry instead of actually learning the underlying physical interactions. Not that AlphaFold isn’t a breakthrough technology for drug discovery, but this paper highlights that even the most cutting-edge applications of AI are perhaps still best used to generate candidates, but humans still need to perform robust tests to validate them. We’ve been hearing a similar sentiment from nearly all of our users when bringing AI into cloud infrastructure management. Flightcrew has been read-only from the start, using AI to generate insights but requiring a human to sign off before any changes are applied to their infrastructure. The goal is to speed up existing workflows, not create a new point of failure. Drug discovery and managing Kubernetes are completely different applications but we’re all similarly wary of trusting full AI automation and at this rate, likely will be for some time.

19

I thought it was very interesting to hear Andrej Karpathy talk at GPU MODE IRL 2024 about llm.c and the conversion of PyTorch code into C/CUDA, as well as the potential implications of this work for future LLM progress. The idea of converting high-level deep learning frameworks like PyTorch into low-level, highly efficient C/CUDA code could drastically improve model performance and execution speed, especially in specialized environments. It will be exciting to see how software engineering (SWE) practices evolve over the coming years as this becomes more widely adopted. One of the most intriguing aspects is the potential for a model to generate a specific kernel for a use case at someone's company in code that has no dependencies and could be fully optimized for that specific task. This capability could not only make models more efficient by reducing overhead but also streamline the workflow for developers, allowing them to focus on higher-level design and problem-solving. Essentially, it introduces a whole new paradigm for how we think about integrating AI into SWE roles. Instead of relying on pre-built libraries or packages, engineers might one day leverage AI models to dynamically create specialized, hyper-optimized code that meets their unique needs, reducing both the time and complexity of development. This shift could also lead to a greater level of automation, where AI models help developers by taking over tasks like performance optimization or low-level code generation, potentially reshaping the way we approach software development as a whole. You can listen to the keynote here: https://lnkd.in/eqJYS8Gb #AI #MachineLearning #DeepLearning #PyTorch #CUDA #SoftwareEngineering #C #LLM

2

🚀 While many studies highlight LLMs' success in answering medical questions, research from Yanjun Gao, PhD, and colleagues takes it a step further. They evaluated LLMs on real-world EHR datasets for clinical prediction, comparing them directly against classical machine learning models—widely regarded as state-of-the-art in this field. By examining tasks like mortality prediction and hospital length-of-stay prediction, which greatly influence clinical workflows, the findings demonstrate that LLMs have significant potential to advance clinical predictive modeling.

7

Very interesting that the “promise” of synthetic data has really taken a downturn over the last few years. LLM’s just being the latest case. There is a need for a different approach to data away from ‘big’ and ‘fast’ (or stolen/scraped). We need a “Slow Data” approach - carefully curated and annotated (meta-data) data that values context (including cultural) and semantics and diversity.

9

1 Comment

Here is our new paper accepted in Neurocomputing journal. SITU: Stochastic input encoding and weight update thresholding for efficient memristive neural network in-situ training The Analog-to-Digital Converter (ADC) sensing and conductance update are the most power-demanding processes in the in-situ training of memristive neural networks. In this work, we propose a new thresholded weight update method in conjunction with stochastic input encoding to reduce the ADC sensing requirement and weight updates. This leads to better power and area efficiency for simple neural network models developed in situ on memristive crossbars. This is an interesting research work by Xuening Dong and collaboration between University of Toronto, James Cook University , and York University. https://lnkd.in/gY4-GgtC

57

1 Comment

In this new paper we explore the use of classical programming techniques in implementing the quantum lattice Boltzmann method in the Intel Quantum SDK. https://lnkd.in/gpDSFe6b

14

Google Unveils Willow: The Future of Quantum Computing Google’s new quantum device, Willow, is a game-changer. To put it simply: What takes a classical computer 10,000 years, Willow does in seconds. Here's how it could impact different areas in simple terms: Faster Medicines: Design new drugs in days, speeding up cures for diseases. Better Weather Predictions: Accurately forecast storms and combat climate change. Safer Internet: Build stronger security to keep your data safe. Smarter AI: Create AI that learns and evolves much faster, improving everything from personal assistants to self-driving cars. Traffic-Free Cities: Optimize city traffic for smoother travel and fewer jams. Cleaner Energy: Design better energy solutions like advanced solar panels and nuclear fusion. Cheaper Flights and Deliveries: Optimize routes to save time, money, and resources. And here's the big one: Understanding the Universe: Simulate the origins of the universe, like the Big Bang, or explore the mysteries of black holes, helping us understand how everything began. Willow isn’t just a beginning of faster chip solving problems—it’s helping us unlock the secrets of life and the universe itself! quantum computing is here.. Dr. Rajeev Tiwari Dr.Taruna Gautam Nihar Amoncar PhD PUNEET SHARMA #QuantumComputing #GoogleWillow #Innovation #FutureTech https://lnkd.in/gY3UFWpm

23

2 Comments

Very interesting paper, on how to inject insights from domain knowledge in inferring from data, using DNNs. Thanks Aleksander Molak for sharing.

4

Do "musical lectures" work? An experiment in my class... My Fall24 class midterm exam had 5 questions, each worth 20 pts. For 4 of the 5 questions, the median score < 20. Except 1 question, whose class-wide median was 20 (i.e., > 50% students got a perfect score!) That q was on Lamport+vector timestamps...the only topic in the class with a "song" lecture.

8

1 Comment

LLMs CANNOT generalize outside their domain distribution. You could argue that they can generalize within the distribution of the domain where they have seen 70% of the data. Even with 70%, they fail terribly. Here is a simple thought experiment. Suppose LLMs were expected to behave like the mathematician Girolamo Cardano in 1501. Could they have independently come up with the imaginary number “i,” the square root of -1? They would likely react with rejection. Logic does not allow you to conjure a number like “i,” because the square of no number can yield a negative. This requires intuition or imagination, a conviction so strong that one accepts the existence of such a number and is willing to discover its properties, leading to the development of complex numbers. This leap laid the groundwork for the discovery of quantum mechanics. This is not simply a case of out of domain generalization; it represents a mysterious leap into the improbable. It is unlikely that this would emerge as a behavior of even a 1,000 trillion parameter model, as it remains constrained by the distribution coefficients of its domain. We need to rethink AI models from ground-up if we need to mimic this unique human trait. #AI #Humans #Math

8

1 Comment

Joint work with Aby Philip, Soorya Rethinasamy, and Vincent Russo now published in Quantum - the open journal for quantum science: https://lnkd.in/giwDBREV Title: Schrödinger as a Quantum Programmer: Estimating Entanglement via Steering Popular Summary: Entanglement is a unique feature of quantum mechanics, initially brought to light by Einstein, Podolsky, and Rosen. Determining whether a general quantum state is entangled or not, known as the separability problem, is of fundamental interest and relevant to various fields of physics: condensed matter, quantum gravity, quantum optics, and quantum information science. Motivated by the rapid development of quantum computers, we provide a variational quantum steering algorithm (VQSA) to address the separability problem. The quantum steering effect, discovered by Schrödinger, is central to the design of our algorithm, which is implementable on quantum computers that are available today. We provide benchmarks and simulations of our algorithm and establish a meaningful connection between steering, entanglement, quantum algorithms, and quantum computational complexity theory.

77

4 Comments

The Human-or-Machine Issue In the June 2024 issue of Communications of the ACM, David Harel reflects on Alan Turing's "imitation game" and introduces the broader "human-or-machine" (H-or-M) issue. As machines increasingly mimic humans, Harel explores new social, psychological, and technical questions. Will humans care whether their interacting agent is human or machine? How will behaviors and emotional responses differ in human-machine interactions compared to human-human ones? #HumanComputerInteraction #ImitationGame #Interactions #HumanBehavior

3

2 Comments