Alison Smith

NEWS - Validation and Evaluation for Trustworthy Artificial Intelligence (VET AI) Act Introduced U.S. Senators John Hickenlooper and Shelley Moore Capito introduced the bipartisan Validation and Evaluation for Trustworthy Artificial Intelligence (VET AI) Act (details below). I'm excited about this bill because it builds on recommendations we--as the IEEE USA AI Policy committee--made in a position statement released last February (https://lnkd.in/e9sETwNM). It was a pleasure to co-lead this effort with Dr. Cari Miller. I'd also like to thank Sen. Hickenlooper's staff for being receptive to feedback that I, and colleagues, provided on this bill. Details: ➡ Directs NIST, in coordination with the Department of Energy and National Science Foundation, to develop voluntary specifications and guidelines for developers and deployers of AI systems to conduct internal assurance and work with third parties on external assurance regarding the verification and red-teaming of AI systems. ➡ Such specifications require considerations for data privacy protections, mitigations against potential harms to individuals from an AI system, dataset quality, and governance and communications processes of a developer or deployer throughout the AI systems’ development lifecycles. ➡ Establish a collaborative Advisory Committee to review and recommend criteria for individuals or organizations seeking to obtain certification of their ability to conduct internal or external assurance for AI systems. ➡ Require NIST to conduct a study examining various aspects of the ecosystem of AI assurance, including the current capabilities and methodologies used, facilities or resources needed, and overall market demand for internal and external AI assurance.

82

8 Comments

In microeconomics 101 (or 1001, for those of us whose school systems needed that extra digit) we learned that rational firms seek to maximize profit. And so us MMM providers tend to produce profit maximizing allocations as one our key prescriptive outputs. But the thing about "rational firms" is that they don't necessarily need to be maximizing profit _right now_. It is entirely rational to build a brand with a key audience in advance of a product being on sale, or to cut profit margins to under cut competitors in a category, or to focus on new sunscriver growth 'no matter the cost' if the strategic path to maximizing profit involves network effects or a known high re-subscription rate. Which is why when I advise on marketing allocation I prefer to directly align to goals the marketing team has bought into as part of the annual planning process -- instead of presuming to know what any given firm is trying to do, just ask them and then minimize the budget needed to accomplish that! #marketingscience

15

Call for Training Data Requests I'll be shortly returning to the US, and in the interim, during a period of career transition, would like to run an experiment/market demand test for high-quality, domain-specific training data for LLMs/other data-related applications. Americans are frequently surprised at the rates that you can hire extremely competent & hard-working folks in various disciplines remotely from the Philippines, eg $20k/year full-time fully loaded for a data analyst, $40k/year for a PhD/licensed expert in various fields. There’s a bit of work involved to navigate quality, compliance, & ensure strong outcomes, but it’s what I’ve been doing in various settings for the past 7 years. Please reach out if you are in need of any of the following: — Domain-specific expert data (eg labelled images of diagnostic biopsies from dermatopathologists, critiques from sociology PhDs, various implementations of the same function in Python, proofs in a certain field from Math PhDs, etc.) — High-quality labelling/annotation tasks of generic inputs (eg preference ratings & comments of chat traces on a Likert scale, human review of automated moderation decisions) — Any task related to data munging or visualization (eg creating dbt templates from scratch or modifying existing ones, assistance in validating & creating LookML, building complex dashboards in Tableau/PowerBI/etc.) If you have any of the above needs, I’d be able to quickly provide an implementation plan, timeline, and cost. A few other details: — Synthetic data creation/augmentation can be also performed by some pre-defined method, or you can do that yourself if you prefer. — If you are unhappy with your current training data provider or data munging process, please specify why

18

1 Comment

Another great post from Melinda Roy on discovering that data governance is already happening at your institution, just in less formalized ways. This article describes how you can match data governance roles to existing higher ed roles. #highereducation #datagovernance #irwatercooler educause

2

My holiday read. A fascinating perspective from data leaders as to why they left their previous environments. It seems like the challenges are similar across various industries.

9

1 Comment

Simplifying a model doesn't necessarily trivialize it. A former Capital One executive once challenged my data & analytics team to make predictive models relevant by rendering statistical models relatable. The concept was to reduce the overall quantity of independent variables and describe in plain-English language (and with simplistic charts) how each impacted the dependent variable. His thesis was that models are abstractions of reality and unless explainable managers would be reluctant to use them to make business decisions (and furthermore: the smaller quantity of variables reflected the fact managers control a limited extent of factors). Statisticians rightly complained that limiting the quantity of dependent variables impaired accuracy and sensitivity. But when carefully parsed we learned the models remained directionally correct (very strongly so) with tolerable loss of fidelity. The overall result was managers confidently understood the relationships and trade-offs between models’ “levers” and the effect on forecasted performance.

15

2 Comments

Check out the Artificial Intelligence (AI) News Roundup for the Week of December 23 presented by the Mississippi Artificial Intelligence Network (MAIN)—the nation's first and only statewide AI initiative! One of the objectives of MAIN is to raise awareness about AI. MAIN helps facilitate this awareness by staying up-to-date on AI news. As part of this effort, MAIN shares daily AI news and compiles a weekly roundup featuring a variety of AI-related updates. Stay connected with MAIN and learn about FREE AI resources available to anyone in Mississippi: https://mainms.org #MAIN #Mississippi #MississippiAI #MississippiArtificialIntelligenceNetwork #AI #News #LatestNews #NewsUpdate #ArtificialIntelligence #AINews #Workforce #WorkforceDevelopment #GenerativeAI #GenAI

16

Check it out! Next Wednesday (July 17) my colleagues Kevin Lobo and Joshua Johnston are going to be hanging out talking about what it takes to have a truly great data team. Whether you are starting from scratch (I hope you are not) or planning for the future, they have a ton of experience to help your data team level up! #data #datateam #it #informationtechnology #datastrategy #bi #datamanagement #dataleaders https://lnkd.in/emf2xP-Y

3

My team at West Monroe has been building an Agentic AI workflow tool to tackle some of the biggest / hardest problems in the data engineering and analytics space. We're creating solutions where we have found GenAI is the most effective: > converting and generating code > generating data models > analyzing and assessing legacy code > overall enabling data engineers and data scientists Check out our latest video showing an Agentic conversion workflow, that helps migrate legacy code from tools like MS SQL, SSIS, SAS, Snowflake, Redshift, Teradata, ... -> to Databricks! *News Alert* Forbes (via Gartner) just spotlighted Agentic AI as the #1 imperative for companies in 2025. Link in comments. What we've learned on this journey building Agentic AI is that this is a skill, a muscle that your teams build. It's all about creativity and problem solving. I can confidently say that West Monroe has more people working on building these skills and applying these techniques than anyone I've run into in the market (by % of team). I'm VERY excited where we're headed next year! Link to our latest video: https://lnkd.in/gvh7UjYi Team credit: Thomas Zhang, Julia Waggoner, Lindsay Michelini, Meghan Patel, Zafeerah Siddiqua, Doug MacWilliams, Robin van Overbeek, Patrick Miller, Sam Atkinson, Brad Ptasienski, Scott Rozen-Levy, David Choi, Lana Buseman, Cory Chaplin, Eric Johnson (and so many others who have helped move this initiative along)! And partner/advisor credit to our friends at Databricks too: Franco Patano, Jeffrey Lipkowitz, Kaylee Harrison, Swaminathan Venkatesh, PhD - appreciate your guidance and support on this journey!

92

2 Comments

Very humbled and blown away to see Statistical Tableau featured as a book review from the Data Visualization Society. Huge thanks to Emmie Mercer, DBA, MSIS for the thorough review. How are you using #Statistics in your analysis? Let me know in the comments below and check out her review.

41

3 Comments

Want to know something about the data Treasury use? - There's *ALOT* of it. Nerd-out incoming, of novel LLM functionality we're starting to explore. One of the cooler things we're working on is how to best help policy professionals take advantage of Generative AI. We've already built an architecture that let's Policy teams upload documents, index those explore different ways to more effectively search and retrieve intelligence. It's practically old hat now - and the RAG space is developing all the time. But we're sat atop a lot of quickly arriving intelligence, and now it's all indexed ready to be used by the LLM, the next question is how can we help teams benefit from a shared intelligence, to help signpost intel that team a that team b could benefit from. Sure, teams talk - but is there an opportunity to be even better? Enter... the KRAKEN! (not the real name). We're building a fetcher of fetchers, an LLM powered pipeline that builds mesh of metadata, facts, concepts and entities from documents indexed that day. It records facts gleaned from documents, then tracks those in future documents, allowing us to specifically look for things we care about, and observe new concepts or entities emerge over time. We think that that this will both help us ground specific LLM experiences, as well as form a foundation of an organisational central intelligence pool, and we'll keep you posted on how it's going. LLMs prove a really interesting dimension to generating structured data and I like sharing where we're looking to exploit tools to benefit. Keen to hear from others working in a similar space? Peace out - have a great weekend! #generativeai #centralgov #hmtreasury #llm #rag #data #pipelines

86

14 Comments

The new #Planner has the ability to #Baseline! I'm jumping up and down seeing this new feature that PMs have been asking for! 👏🥂🎉 Reminder - Learn more about the new Planner with me April 30 at the Generative AI Day for PMs conference - https://lnkd.in/eMsC_ART

48

2 Comments

Check out the #AI News Roundup for the Week of April 29 from the Mississippi Artificial Intelligence Network (MAIN)! One of the objectives of #MAIN is to raise awareness about Artificial Intelligence (AI). MAIN helps facilitate this awareness by staying up-to-date on AI news. As part of this effort, MAIN compiles a variety of AI news every week and creates a news roundup. #MAIN #MississippiArtificialIntelligenceNetwork #AINews #News #AIAwareness #NewsRoundup #Mississippi #ArtificialIntelligence #GenAI #GenerativeAI

5

Exciting News: My new book "Charting Made Easy: 109 Ideas for Simplifying Complex Data" is (finally) available! 📚 I'm thrilled to announce the release of my book, "Charting Made Easy: 109 Ideas for Simplifying Complex Data," now available on Amazon. This book is a culmination of my passion for data visualization and my desire to help others make sense of complex data. What’s Inside? - 𝟭𝟬 𝗞𝗣𝗜 𝗗𝗲𝘀𝗶𝗴𝗻𝘀: Learn how to design KPIs that highlight when action is needed. - 𝟭𝟬 𝗪𝗮𝘆𝘀 𝘁𝗼 𝗗𝗶𝘀𝗽𝗹𝗮𝘆 𝗩𝗮𝗿𝗶𝗮𝗻𝗰𝗲 𝘄𝗶𝘁𝗵 𝗕𝗮𝗿 𝗖𝗵𝗮𝗿𝘁𝘀: Discover effective methods to showcase differences in your data. - 𝟭𝟭 𝗠𝗲𝘁𝗵𝗼𝗱𝘀 𝗳𝗼𝗿 𝗖𝗼𝗺𝗽𝗮𝗿𝗶𝗻𝗴 𝗧𝘄𝗼 𝗠𝗲𝗮𝘀𝘂𝗿𝗲𝘀: Compare and contrast measures with clarity. - 𝟭𝟴 𝗕𝗮𝗿 𝗖𝗵𝗮𝗿𝘁 𝗩𝗶𝘀𝘂𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻𝘀: Explore innovative ways to represent data using bar charts. - 𝟲𝟬 𝗧𝗲𝗰𝗵𝗻𝗶𝗾𝘂𝗲𝘀 𝗳𝗼𝗿 𝗩𝗶𝘀𝘂𝗮𝗹𝗶𝘇𝗶𝗻𝗴 𝗧𝗶𝗺𝗲: Master the art of time-based data visualization. Each page presents a unique visualization alongside a detailed explanation of when and how to use it. I would love for you to use this book is your go-to reference for turning complex data into clear, actionable insights. Why I Wrote This Book? I wrote this book to share practical, easy-to-implement visualization techniques that can help anyone, regardless of their data expertise. The goal is to simplify the way we present and understand data, making it more accessible and actionable for everyone. Grab Your Copy: I don't make any money from this book; it's a project born from my dedication to the field of data visualization. You can find it on Amazon here: https://amzn.eu/d/0aQn4Tny Available in hardback and paperback. Thank you for your support, and I hope you find "Charting Made Easy" as insightful and useful as I intended it to be.

107

21 Comments

Those interested in interoperability, including laboratory interoperability should comment through August 23rd on the expanded Data Usability Implementation Guide. The guide focuses on health data exchange across EHR, HIE, Public Health, and laboratories, leveraging published standards to lift everyone up together and help retain vital information in exchanges for downstream users.

6

2 Comments

We just published the second installment in the Decision Maker's Guide to the Modern Data Stack, this one on cloud storage. Another great piece if you're thinking about how to structure your data stack.

17

Brilliant piece by Nick Hart, Ph.D. and Suzette Kent on the importance of a national data strategy that includes the state CDOs. Kaptivate has argued that better data refined through cross sector collaboration can transform workforce development. As the piece notes, “The workforce implications are equally compelling. Better data integration between education providers, employers and workforce agencies could help match people to opportunities more effectively.” It’s time to commit ourselves to better data. “The foundations are in place. The technology exists. The return on investment is clear. It’s time to move from strategy to action and build the data infrastructure America needs to own its future” Let’s go! 🚀 #PivotandThrive Dominique Harrison Aluko, Ph.D. Natalie Evans Harris Jennifer "Jenn" Stein Katherine Kempe, MPA Andres M. Baquero Kathryn de Wit Amanda Ponzar Melissa Newman Kathryn Newcomer Francella Ochillo https://lnkd.in/effQq_Cf

21

4 Comments