Eric Kramer

Would be interesting to see how this shapes up and speeds up data engineering et al. Governance will be mich more important! #databricks #AI #datamanagement

3

LLMs are getting commoditized. The pace of core model breakthroughs has clearly slowed. We are no longer seeing step-function jumps from “just a bigger model”. For most real systems today: • architecture matters more than the base LLM, • planning, orchestration, memory, and verification dominate quality, • the model itself is increasingly a replaceable component. In my own code factory, I can swap GPT, Claude, or Gemini with minimal impact. There are differences in latency and edge cases, but they no longer define the system. The next real breakthroughs will not come from GPT-6. They will come from new architectures built around models, not inside them. LLMs are becoming infrastructure. Systems are becoming the product.

1

Exciting insights from the Nvidia partners event at JPM led by Jensen Huang (reposting now because original post was mysteriously deleted): - Patrick Collison, The visionary Founder and CEO of Stripe and co-founder of Arc Insitute (they have developed Evo (1, 2) genomic foundation models) - Christina Zorn, Chief Administrator Officer of Mayo Clinic, aims to enhance patient care by harnessing their vast repository pf 20 Billion images and clinical data sets - Jacob Thaysen, PhD, CEO of Illumina, is steering the company from their current status as a sequencing leader to a pioneering digital biology platform - Ari Bousbib, CEO of IQVIA, wants to develop AI agents to extract insights from a remarkable 64 petabytes of exclusive healthcare data Lots of great promise for AI in improving lives of patients! #NvidiaEvent #healthcareinnovation #AIfor healthcare

115

6 Comments

Carlos Guzman’s #OpenMed walkthrough is a fantastic proof point: bring a healthcare-tuned model to your data, run it at scale, and only pay while it’s running. 1M clinical patient notes, 4 GPU’s, 5m30s, ~$2–3. This is how we vault healthcare forward with AI —by unlocking real use cases at low cost. #Snowflake #CortexAI

12

Comparing models.. Finding that a small 1B #llm hallucinates less than a 4B for semantic to structural matching. Small llms for targeted tasks for the win. The new Meta llama3.2 1B is better than expected and cheap on token cost.

3

Document-centric tasks sit at the core of many enterprise, business, and government workflows. Search is important, but the real challenge is going beyond retrieval, enabling systems to reason across documents, verify facts, and handle multi-step information tasks. Great to see a new model from Databricks moving in this direction. I expect we’ll see many more models, including open-weight ones, designed specifically for document-driven workflows, an area with huge potential across enterprise and government use cases.

23

1 Comment

This looks useful for folks in research or scanning what's coming out on the frontier: Lumi from Google's PAIR research team, a tool for better understanding papers on arxiv. Reading smarter, not harder -- that tagline brings me back to Mortimer Adler's classic 'How to Read a Book'! From their GH repo: Lumi uses AI to help you quickly read and understand arXiv papers. Features include: ✏️ AI-augmented annotations - read summaries at multiple granularities 🔖 Smart highlights - highlight text + ask questions 🖼️ Figure explanations - ask Lumi about images in the paper Demo: https://lnkd.in/ex7erTBb GH repo: https://lnkd.in/emz_Ghcs Medium article: https://lnkd.in/enky8dKC GitHub discussions: https://lnkd.in/efAmiCcp

19

I’m curious -- is anyone actually doing end-to-end data analysis with LLMs and trusting the final result? Not bioinformatics pipelines, but the last-mile exploratory stuff in Jupyter/R. Beyond just asking for a quick snippet, how are you auditing the logic? When you use an LLM for software, you have a built-in safety net: observability. If the code is wrong, the button doesn't click, the API throws a 500, etc. You see the error, you fix it, you move on. But with data analysis, errors are silent. That's terrifying. A few ways this goes sideways without throwing a single error: ❌ The Label Flip: The model arbitrarily decides the "positive" class is the first label it sees. Your results are technically perfect, just 100% inverted. ❌ Silent Swaps: Label misalignments or bad joins that don't break the code but render the insight useless. ❌ Missing Context: The LLM has no idea your plate map was rotated or that "Control B" was an outlier. Worse, these errors compound. You might have an accidental label swap in Cell 5, then an off-by-one error in Cell 15, then use the wrong (but "almost correct") column in Cell 16. That's why many of us inspect, tweak, and re-run each Jupyter cell dozens of times before moving on. We are convincing ourselves that the logic and the output are correct. In bioinformatics, the pipelines are generally standardized with well-defined QC criteria. In general data work? It’s much easier to accidentally trust a result that looks right but is fundamentally broken under the hood. How are you all handling this? Is anybody actually letting LLMs do their data analysis? #DataScience #Python #RStats #MachineLearning #LLM #AI

6

7 Comments