Chris E.

Will AI increase the rate of progress in science or not? A new paper from Sayash Kapoor and Arvind Narayanan (Could AI slow science?) says it probably won't AI is already used in science (See figure) “to uncover trends in data using sophisticated pattern-matching algorithms; hand-written machine learning models specified based on expert knowledge; or even generative AI to write the code that researchers previously wrote.” But “AI has already led to widespread errors. Even before generative AI, traditional machine learning led to errors in over 600 papers across 30 scientific fields. In many cases, the affected papers constituted the majority of the surveyed papers, raising the possibility that in many fields, the majority of AI-enabled research is flawed.” More importantly, AI modeling doesn’t lead to an improved understanding of scientific phenomena. “Solving problems and writing papers about them can be seen as a ritual that leads to the real prize, human understanding, without which there can be no scientific progress.” One Fields Medal-winning mathematician “emphasizes that the point of mathematics is not simply to figure out the truth value for mathematical facts, but rather the accompanying human understanding.” The same holds for AI. If AI solves “open research problems without leading to the accompanying understanding, AI could erode these useful byproducts by reducing incentives to build understanding.” Another reason that AI will likely not improve the conduct of science is that the AI-for-science suppliers want flashy “AI discovers X!” headlines “so that they can sustain the narrative that AI will solve humanity’s problems, which buys them favorable policy treatment.” He says “we should be skeptical of #AI-for-science news headlines. Many of them are greatly exaggerated. The results may fail to reproduce, or AI may be framed as the main character when it was in fact one tool among many.” If you are developing tools, “How about tools for finding errors in scientific code or other forms of quality control? Or “mathematicians have repeatedly said that tools for improving human understanding are much more exciting than trying to automate theorem-proving, which they view as missing the point.” The way we evaluate these tools should also change. For a lit review tool: 1) Does it save a researcher time and produce results of comparable quality to existing tools? 2) How does the use of the tool impact the researcher’s understanding of the literature compared to traditional search? 3) What will the collective impacts on the community be if the tool were widely adopted”? Currently, only the first question is considered thus guaranteeing that evaluations are “guaranteed to provide a highly incomplete and biased picture.” In conclusion, “The impact of AI on science as an institution, rather than individual scientists, is a different question that demands a different kind of analysis.” #technology #artificialintelligence

122

56 Comments

🧬 The concept of Neural DNA (nDNA), initially proposed by Dr. Amitava Das, opens up a new avenue for interpreting LLM behavior. nDNA treats an LLM as something we can sequence (just like DNA sequencing) by reading its internal signals -- its lineage and layer-level changes. It captures a model’s lineage, tuning-induced mutations, and functional expression so we can compare LLM families, spot drift, and explain why similar benchmarks behave differently across a range of LLMs. In practice, nDNA looks for signals such as: (i) layer shifts, (ii) MoE routing patterns, and (iii) controlled probes. It lets us audit change over time and pick models for tasks with evidence. Collaborating with Dr. Amitava Das, and his Pragya group (https://pragyaai.github.io) at the Department of CSIS BITS Pilani Goa Campus, we've been building and testing the concept of nDNA to understand LLM behavior across major job families: - Llama tends to refine mid–to-late layers while early layers remain stable; - Mistral localizes expertise while the early layers remain steady; - Gemma aligns with light-touch parameter changes; - Qwen rewrites more in its middle layers, especially in multilingual scenarios; - DeepSeek adds dialogue capability in mid layers, while preserving base geometry. One of the other distinct advantages of nDNA is the ability to identify model collapse. On the surface you might see a slow loss of creativity and nuance in a model’s behavior. Inside, through the nDNA lens, the model’s internal landscape flattens -- fewer distinct patterns light up -- much like reduced neural diversity in biology. With nDNA, that shift becomes measurable over time, which makes collapse diagnosable and monitorable. Dr. Amitava Das's novel thought process and insistence on clear hypotheses has led to the inception of nDNA as a practical lens for AI -- one that detects drift before deployment, keeps model lineages auditable, and matches models to tasks based on evidence rather than guesswork. Taken together, these nDNA lenses reframe how we interpret LLM behavior. Our other recent works, DPO-Kernels (ACL 2025; https://lnkd.in/gXGQvXwX), Yin-Yang Align (ACL 2025; https://lnkd.in/gWVCxG8B), QuickSilver (https://lnkd.in/gpZQKMmP), and the Counter Turing Test (EMNLP 2023 Outstanding Paper Award; https://lnkd.in/gaVHeXgJ), share a similar first-principle approach to LLM refinement/understanding. If nDNA is something you'd like to explore further -- we'll be launching a website for nDNA soon, stay tuned! #nDNA #AI #FoundationModels #ModelInterpretability

283

5 Comments

📝 Announcing our paper that proposes a unified cognitive and computational framework for Artificial General Intelligence (AGI) -- going beyond token-level predictions -- one that emphasizes modular reasoning, memory, agentic behavior, and ethical alignment 🔹 𝐓𝐡𝐢𝐧𝐤𝐢𝐧𝐠 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧𝐬: 𝐅𝐫𝐨𝐦 𝐁𝐫𝐚𝐢𝐧‑𝐈𝐧𝐬𝐩𝐢𝐫𝐞𝐝 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 𝐭𝐨 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧𝐬 𝐟𝐨𝐫 𝐀𝐫𝐭𝐢𝐟𝐢𝐜𝐢𝐚𝐥 𝐆𝐞𝐧𝐞𝐫𝐚𝐥 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 𝐚𝐧𝐝 𝐢𝐭𝐬 𝐒𝐨𝐜𝐢𝐞𝐭𝐚𝐥 𝐈𝐦𝐩𝐚𝐜𝐭 🔹 In collaboration with University of Central Florida, Cornell University, UT MD Anderson Cancer Center UTHealth Houston Graduate School of Biomedical Sciences, Toronto Metropolitan University, University of Oxford, Torrens University Australia, Obuda University, Amazon others. 🔹 Paper: https://lnkd.in/gqKUV4Mr ✍🏼 Authors: Rizwan Qureshi, Ranjan Sapkota, Abbas Shah, Amgad Muneer, Anas Zafar, Ashmal Vayani, Maged Shoman, PhD, Abdelrahman Eldaly, Kai Zhang, Ferhat Sadak, Shaina Raza, PhD, Xinqi Fan, Ravid Shwartz Ziv, Hong Yang, Vinija Jain, Aman Chadha, Manoj Karkee, @Jia Wu, Philip Torr, FREng, FRS, Seyedali Mirjalili ➡️ 𝐊𝐞𝐲 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬 𝐨𝐟 𝐓𝐡𝐢𝐧𝐤𝐢𝐧𝐠 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧𝐬' 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞‑𝐂𝐨𝐦𝐩𝐮𝐭𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐀𝐆𝐈 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤: 🧠 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤: Integrates cognitive neuroscience, psychology, and AI to define AGI via modular reasoning, persistent memory, agentic behavior, vision-language grounding, and embodied interaction. 🔗 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧‑𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧: Critiques token-level models like GPT-4.5 and Claude 3.5, advocating for test-time adaptation, dynamic planning, and training-free grounding through retrieval-augmented agentic systems. 🚀 𝐑𝐨𝐚𝐝𝐦𝐚𝐩 𝐚𝐧𝐝 𝐂𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧𝐬: Proposes a roadmap for AGI through neuro-symbolic learning, value alignment, multimodal cognition, and cognitive scaffolding for transparent, socially integrated systems.

221

9 Comments

From UI frameworks to data dashboards to cutting-edge NLP, at New Math Data, we’ve been busy exploring the cutting-edge technology that moves teams and organizations forward. ✅ Angular 20 is here, and Sriprasad's article touts it as the most developer-friendly release yet. With standalone components, zoneless change detection, and a smarter build pipeline, this version doesn’t just modernize, but it simplifies and accelerates. This article has an overview of the new features with code examples to boot! 📊 QuickSight QuickStart: Chris K. shows how to go from a blank AWS account to a fully deployed BI environment in one command. No click-heavy setup. Just dashboards, data, and fast insights. 🧠 RAG Meets Computational Linguistics: Here, I discuss computational linguistics concepts and how Retrieval-augmented generation becomes far more powerful when paired with knowledge of and clever implementation of the fundamentals. This isn’t just theory. It’s how you build smarter RAG systems. We're always aiming to ship ideas that matter and push us forward, as well as systems that scale. Dive into the details or reach out if you're solving similar challenges! https://lnkd.in/gTzHibvk https://lnkd.in/gawTbraM #Angular #AWSQuickSight #NLP #RAG #DataEngineering #SoftwareDevelopment #MachineLearning #AI #CloudComputing #NewMathData

16

2 Comments

xLSTM for Aspect-based Sentiment Analysis: https://lnkd.in/dQWhwQwv Another success story of xLSTM. “We propose xLSTM with Multihead Exponential Gated Fusion (MEGA)” “Experimental results on three benchmark datasets demonstrate that MEGA outperforms state-of-the-art baselines, achieving superior accuracy and efficiency in ABSA tasks.” “MEGA efficiently extracts global features and enables dynamic cross information exchange,  chieving good accuracy with a lightweight architecture, surpassing attention-based, syntax-based, and integrative baselines on public ABSA datasets.”  

170

12 Comments

📊 Can quantum-powered optimization beat generative AI in designing new drug candidates? That’s the question tackled in a new paper by Kendall Byler and Shahar Keinan from PolarisQB (Thank you for bringing this to my attention, Shahar! Link in the comments 👇). Rather than comparing quantum computing and AI as broad technologies, the study puts two structure-based drug design (SBDD) platforms head-to-head: ⚙️ QuADD, which uses quantum annealing to optimize fragment-based molecular design ⚙️ BInD, a deep generative diffusion model that builds molecules and their protein interactions from scratch using AI Both methods were tested on the same drug target, thrombin, a key enzyme involved in blood clotting. The idea was to see which platform produces more viable drug-like molecules in terms of binding affinity, drug-likeness, interaction fidelity, and synthetic feasibility. 💡 After generating and analyzing over 3,600 candidate molecules from each platform, the results showed clear performance differences: - QuADD molecules had consistently stronger predicted binding affinities and better replication of key hydrogen bonds (notably with Ser214 and Gly216) - BInD outputs were more structurally diverse, but often fell short in drug-likeness (e.g., low TPSA, high synthetic complexity) - QuADD ran 80× faster (~30 minutes vs. 40 hours), producing drug-like candidates with lower synthetic complexity scores, making them more realistic for follow-up synthesis In short, while BInD explores a broader chemical space, QuADD delivers more viable, synthesizable, and pharmacologically relevant candidates, so I think the results highlight the promise of quantum-optimized search in early-stage drug discovery. But also, I think there is clear complementarity between methods... What do you think?

154

46 Comments

Much of deep learning still feels more like alchemy than science, but optimizing model performance doesn't have to - even at huge scale! DeepMind researchers shared an excellent 12-part series on how to scale LLMs. This series (link in comments) aims to demystify the science of scaling language models: how TPUs (and GPUs) work and how they communicate with each other, how LLMs run on real hardware, and how to parallelize your models during training and inference so they run efficiently at massive scale. If you've ever wondered "how expensive should this LLM be to train?" or "how much memory do I need to serve this model myself?" or "what's an AllGather?", this series will be useful. This will help researchers designing new architectures and engineers working to make current LLMs run faster. Promising model architectures routinely fail because they can't run efficiently at scale or because no one puts in the work to make them do so. The series also provides useful perspective on the evolving distinction between scientists and engineers in AI. While I believe there's still a meaningful distinction, this work illustrates that scaling considerations are engineering topics that ML researchers now need to master - something that wasn't as critical 4+ years ago. Here is the outline: Part 1: Preliminaries Chapter 1: A Brief Intro to Roofline Analysis Chapter 2: How to Think About TPUs Chapter 3: Sharded Matrices and How to Multiply Them Part 2: Transformers Chapter 4: All the Transformer Math You Need to Know Chapter 5: How to Parallelize a Transformer for Training Chapter 6: Training LLaMA 3 on TPUs Chapter 7: All About Transformer Inference Chapter 8: Serving LLaMA 3 on TPUs Part 3: Practical Tutorials Chapter 9: How to Profile TPU Code Chapter 10: Programming TPUs in JAX. Part 4: Conclusions and Bonus Content Chapter 11: Conclusions and Further Reading Chapter 12: How to Think About GPUs Have fun! Excited to hear what you think of the series. #ai #genai #llm

174

13 Comments

A paper by Cornell University argues that SLMs are the future of agentic AI: https://lnkd.in/g688UxTJ Check out #Couchbase's Mohan Varthakavi's POV on when to use SLMs over LLMs when designing AI apps: https://lnkd.in/gnVP2n_5

13

Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models (Rice University, March 2025) A call to #humans and #LLM’s a like;) 😂 😂 Paper: https://lnkd.in/gX3xkZ47 Abstract: "Large Language Models (LLMs) have demonstrated remarkable capabilities in complex tasks. Recent advancements in Large Reasoning Models (LRMs), such as OpenAI o1 and DeepSeek-R1, have further improved performance in System-2 reasoning domains like mathematics and programming by harnessing supervised fine-tuning (SFT) and reinforcement learning (RL) techniques to enhance the Chain-of-Thought (CoT) reasoning. However, while longer CoT reasoning sequences improve performance, they also introduce significant computational overhead due to verbose and redundant outputs, known as the "overthinking phenomenon". In this paper, we provide the first structured survey to systematically investigate and explore the current progress toward achieving efficient reasoning in LLMs. Overall, relying on the inherent mechanism of LLMs, we categorize existing works into several key directions: (1) model-based efficient reasoning, which considers optimizing full-length reasoning models into more concise reasoning models or directly training efficient reasoning models; (2) reasoning output-based efficient reasoning, which aims to dynamically reduce reasoning steps and length during inference; (3) input prompts-based efficient reasoning, which seeks to enhance reasoning efficiency based on input prompt properties such as difficulty or length control. Additionally, we introduce the use of efficient data for training reasoning models, explore the reasoning capabilities of small language models, and discuss evaluation methods and benchmarking." Project Website: https://lnkd.in/gbr46XNy Lohith Srikanth Pentapalli Kelly Cohen Scott Phillips Tri Nguyen Summer Crenshaw Zachary Huhn Monica Turner

21