https://thenewstack.io/how-should-we-define-open-ai/ TNS OK SUBSCRIBE Join our community of software engineering leaders and aspirational developers. Always stay in-the-know by getting the most important news and exclusive content delivered fresh to your inbox to learn more about at-scale software development. EMAIL ADDRESS REQUIRED [ ] SUBSCRIBE RESUBSCRIPTION REQUIRED It seems that you've previously unsubscribed from our newsletter in the past. Click the button below to open the re-subscribe form in a new tab. When you're done, simply close that tab and continue with this form to complete your subscription. RE-SUBSCRIBE The New Stack does not sell your information or share it with unaffiliated third parties. By continuing, you agree to our Terms of Use and Privacy Policy. Welcome and thank you for joining The New Stack community! Please answer a few simple questions to help us deliver the news and resources you are interested in. FIRST NAME REQUIRED [ ] LAST NAME REQUIRED [ ] COMPANY NAME REQUIRED [ ] COUNTRY REQUIRED [Select ... ] ZIPCODE REQUIRED [ ] Great to meet you! Tell us a bit about your job so we can cover the topics you find most relevant. What is your job level? REQUIRED [Select ... ] [ ] Which of these most closely describes your job role? REQUIRED [Select ... ] [ ] How many employees are in the organization you work with? REQUIRED [Select ... ] What option best describes the type of organization you work for? REQUIRED [Select ... ] [ ] Which of the following best describes your organization's primary industry? REQUIRED [Select ... ] [ ] LINKEDIN PROFILE URL REQUIRED [ ] Welcome! We're so glad you're here. You can expect all the best TNS content to arrive Monday through Friday to keep you on top of the news and at the top of your game. What's next? Check your inbox for a confirmation email where you can adjust your preferences and even join additional groups. Follow TNS on your favorite social media networks. Become a TNS follower on LinkedIn. Check out the latest featured and trending stories while you wait for your first TNS newsletter. PREV 1 of 2 NEXT VOXPOP What Is our P(doom) Level? P(doom) is Silicon Valley's latest hot metric, an approximation of how worried you are that AI, once fully cognizant, will destroy humanity. Inversely proportional to e/accs (the "effective accelerationists"). 0-25: AI will usher in a new age of utopianism. 0% 26-50: What, me worry? 0% 51-75: We'll probably be OK with some government oversight. 0% 76-100: I'm stocking up on batteries and canned food now. 0% Thanks for your opinion! Subscribe below to get the final results, published exclusively in our TNS Update newsletter: [ ] SUBMIT [ ] Search More Results ARCHITECTURE Cloud Native Ecosystem Containers Edge Computing Microservices Networking Serverless Storage ENGINEERING AI Large Language Models Frontend Development Software Development API Management Python JavaScript TypeScript WebAssembly Cloud Services Data Security OPERATIONS Platform Engineering Operations CI/CD Tech Careers Tech Culture DevOps Kubernetes Observability Service Mesh CHANNELS Podcasts Ebooks Events Newsletter TNS RSS Feeds THE NEW STACK About / Contact Sponsors Sponsorship Contributions PODCASTS EBOOKS EVENTS NEWSLETTER ARCHITECTURE ENGINEERING OPERATIONS Cloud Native Ecosystem Containers Edge Computing Microservices Networking Serverless Storage Using an Internal Developer Portal for FinOps Visibility Apr 3rd 2024 9:05am, by Aidan O'Connor KubeCon24: Mirantis Doubles Down on Open Source Mar 28th 2024 8:03am, by Steven J. Vaughan-Nichols KubeCon24: Red Hat OpenShift 4.15 Accelerates App Delivery Mar 21st 2024 9:40am, by Steven J. Vaughan-Nichols KubeCon24: VictoriaMetrics' Simpler Alternative to Prometheus Mar 20th 2024 1:14pm, by Joab Jackson An Open Source Journey to Greener Cloud Native Environments Mar 20th 2024 6:35am, by Graziano Casto Chainguard: Outdated Containers Accumulate Vulnerabilities Mar 29th 2024 3:00am, by Joab Jackson Evolve Manual and Templated Dockerfiles with Automation Mar 26th 2024 10:33am, by Rak Siva Virtualization and Containers: Better Together Mar 19th 2024 6:18am, by Alex Handy Docker and Chainguard Join Forces to Deliver Secure Containers Mar 18th 2024 1:14pm, by Steven J. Vaughan-Nichols Simplify Linux and Docker Command Lines with Bash Completion Mar 9th 2024 6:00am, by Jack Wallen The Future of AI: Hybrid Edge Deployments Are Indispensable Mar 22nd 2024 10:00am, by Luis Ceze How RapidAI Uses Edge, Kubernetes and AI to Boost Stroke Care Mar 15th 2024 10:30am, by Charles Humble Trusted Boot: What to Know About Securing Devices at the Edge Mar 14th 2024 7:15am, by Ettore di Giacinto Architecting for Industrial IoT Workloads: A Blueprint Jan 31st 2024 7:34am, by Dunith Danushka Enabling AI in IoT Apps with a Cloud-to-Edge Database Jan 31st 2024 6:11am, by Mark Gamble API Design Is Pretty Bad -- Here's How to Fix It Apr 3rd 2024 6:01am, by Lebin Cheng Will Spotify Open Source its Microservices Framework? Apr 2nd 2024 7:45am, by Loraine Lawson Enhancing Business Security and Compliance with Service Mesh Apr 1st 2024 10:00am, by Ninad Desai 10 Ways Kubernetes Observability Boosts Productivity, Cuts Costs Mar 27th 2024 6:08am, by Eric Schabell Evolve Manual and Templated Dockerfiles with Automation Mar 26th 2024 10:33am, by Rak Siva How Observability Is Different for Web3 Apps Mar 15th 2024 12:00pm, by Sarah Morgan Simplify Kubernetes Hosted Control Planes with K0smotron Mar 11th 2024 10:00am, by Jussi Nummelin The Black Hole That Is the Kubernetes Network Mar 4th 2024 6:30am, by Mitch Connors Netflix Releases bpftop: An eBPF-Based Application Monitor Feb 27th 2024 8:53am, by Steven J. Vaughan-Nichols Samba Network Shares for RHEL-Based Linux Distributions Feb 25th 2024 6:00am, by Jack Wallen Meet DBOS: A Database Alternative to Kubernetes Mar 12th 2024 4:00am, by Joab Jackson Pulumi Templates for GenAI Stacks: Pinecone, LangChain First Feb 21st 2024 9:00am, by Joab Jackson CNCF CloudEvents: A Li'l Message Envelope That Travels Far Jan 31st 2024 4:00am, by Joab Jackson Bringing the AWS Serverless Strategy to Azure Jan 19th 2024 6:00am, by Rak Siva Serverless Computing In 2024: GenAI Influence, Security, 5G Jan 4th 2024 5:00am, by Chris J. Preimesberger From Postgres to ScyllaDB NoSQL, with a 349x Speed Boost Apr 1st 2024 11:27am, by Cynthia Dunlop The Architect's Guide: A Modern Data Lake Reference Architecture Mar 26th 2024 9:17am, by Keith Pijanowski Cloud Data Migration or Cloud Data Tiering? Mar 25th 2024 10:00am, by Kumar Goswami KubeCon24: MinIO Object Store Equipped with Enterprise Features Mar 19th 2024 2:00pm, by Joab Jackson Architect's Guide to a Reference Architecture for an AI/ML Data Lake Mar 19th 2024 12:09pm, by Keith Pijanowski AI Large Language Models Frontend Development Software Development API Management Python JavaScript TypeScript WebAssembly Cloud Services Data Security Breaking the Big Tech Grip: An AI Playing Field for Everyone's Benefit Apr 3rd 2024 10:00am, by Tim Davis RAG without OpenAI: BentoML, OctoAI and Milvus Apr 3rd 2024 7:49am, by Yujian Tang Beyond Git: A New Collaboration Model for AI/ML Development Apr 2nd 2024 10:00am, by Gorkem Ercan AI Code Generation: 6 FAQs for Developers Apr 2nd 2024 8:57am, by Manish Kapur Why Vector Size Matters Apr 1st 2024 8:23am, by Jonathan Ellis RAG without OpenAI: BentoML, OctoAI and Milvus Apr 3rd 2024 7:49am, by Yujian Tang Why Vector Size Matters Apr 1st 2024 8:23am, by Jonathan Ellis Unlock Hyper-Personalization with AI-Driven Adaptive Apps Mar 29th 2024 8:36am, by Tim Rottach How to Build a RAG-Powered LLM Chat App with ChromaDB and Python Mar 29th 2024 7:03am, by Oladimeji Sowole Large Language Model Observability: The Breakdown Mar 28th 2024 1:46pm, by Alex Williams Dev News: Deno Supports Open Source Repository JSR and an Offline AI Mar 30th 2024 4:00am, by Loraine Lawson Using AI for Work: New Open Source Gateway Promises Better Control Mar 28th 2024 7:17am, by Loraine Lawson UI Libraries Are Dying: What's Next? Mar 27th 2024 9:41am, by Eden Ella Dev News: WordPress 6.5, Angular Signals and .NET Components Mar 23rd 2024 4:00am, by Loraine Lawson WP Engine Introduces AI-Powered Search for WordPress Sites Mar 22nd 2024 7:44am, by Loraine Lawson AI Code Generation: 6 FAQs for Developers Apr 2nd 2024 8:57am, by Manish Kapur How to Code First with Design-First Benefits Apr 2nd 2024 7:16am, by Rak Siva From Postgres to ScyllaDB NoSQL, with a 349x Speed Boost Apr 1st 2024 11:27am, by Cynthia Dunlop 5 Software Development Skills AI Will Render Obsolete Apr 1st 2024 7:16am, by Eran Yahav What Are Python Lambda Functions and How Do You Use Them? Apr 1st 2024 5:00am, by Jack Wallen API Design Is Pretty Bad -- Here's How to Fix It Apr 3rd 2024 6:01am, by Lebin Cheng Using AI for Work: New Open Source Gateway Promises Better Control Mar 28th 2024 7:17am, by Loraine Lawson Dev News: WordPress 6.5, Angular Signals and .NET Components Mar 23rd 2024 4:00am, by Loraine Lawson The Economics of API Attacks and How Developers Can Stop Them Mar 12th 2024 8:03am, by Loraine Lawson Learn to Love Day 2 Operations with GitOps-Driven API Management Mar 8th 2024 7:04am, by Emile Vauge What Are Python Lambda Functions and How Do You Use Them? Apr 1st 2024 5:00am, by Jack Wallen How to Build a RAG-Powered LLM Chat App with ChromaDB and Python Mar 29th 2024 7:03am, by Oladimeji Sowole NVIDIA Wants More Programming Languages to Support CUDA Mar 28th 2024 10:30am, by Agam Shah Python's Collection Module for Specialized Data Structures Mar 27th 2024 5:00pm, by Jack Wallen AlgoKit -- Python Toolkit for Blockchain Mar 27th 2024 7:58am, by Jessica Wachtel Dev News: Deno Supports Open Source Repository JSR and an Offline AI Mar 30th 2024 4:00am, by Loraine Lawson Dev News: WordPress 6.5, Angular Signals and .NET Components Mar 23rd 2024 4:00am, by Loraine Lawson Dev News: Android Apps on Rust, Astro DB, and Storybook 8 Mar 16th 2024 4:00am, by Loraine Lawson Free Tool Helps Web Devs with Google's 'Complex' SEO Update Mar 14th 2024 9:25am, by Loraine Lawson How to Build Embed Components with Astro, Qwik and StackBlitz Mar 9th 2024 5:00am, by Paul Scanlon Dev News: Deno Supports Open Source Repository JSR and an Offline AI Mar 30th 2024 4:00am, by Loraine Lawson Advanced OOP in TypeScript: Interfaces and Abstract Classes Mar 22nd 2024 10:30am, by Bob Reselman How to Get Advantages of TypeScript in JavaScript Oct 27th 2023 10:51am, by Phil Nash Dev News: Udemy's New Docker Program, Plus TypeScript Beta Oct 7th 2023 5:01am, by Loraine Lawson The Angular Renaissance: Why Frontend Devs Should Revisit It Sep 26th 2023 8:15am, by Loraine Lawson KubeCon Europe: WebAssembly, eBPF Are Huge for Cloud Native Mar 29th 2024 8:24am, by B. Cameron Gain Fermyon Says WebAssembly on Kubernetes Is Now Doable Mar 28th 2024 6:19am, by B. Cameron Gain Why Wasm Wins Where Java Applets Failed Mar 12th 2024 10:22am, by Liam Crilly Adobe Developers Use WebAssembly to Improve Users' Lives Mar 8th 2024 6:49am, by B. Cameron Gain WASI 0.2: Unlocking WebAssembly's Promise Outside the Browser Mar 7th 2024 5:00am, by Tyler McMullen and Luke Wagner IT and Climate Change: A Startup Builds for a Greener Future Apr 1st 2024 10:06am, by Heather Joslyn OVHcloud Unveils Roadmap to Take on Hyperscalers from Europe Mar 29th 2024 6:14am, by B. Cameron Gain Cloud Data Migration or Cloud Data Tiering? Mar 25th 2024 10:00am, by Kumar Goswami The Future of AI: Hybrid Edge Deployments Are Indispensable Mar 22nd 2024 10:00am, by Luis Ceze How to Introduce Real-Time Data Predictions with Redpanda Mar 22nd 2024 8:18am, by Christina Lin How to Use Data Manipulation Language (DML) in SQL Apr 3rd 2024 11:18am, by Gerald Venzl RAG without OpenAI: BentoML, OctoAI and Milvus Apr 3rd 2024 7:49am, by Yujian Tang Beyond Git: A New Collaboration Model for AI/ML Development Apr 2nd 2024 10:00am, by Gorkem Ercan From Postgres to ScyllaDB NoSQL, with a 349x Speed Boost Apr 1st 2024 11:27am, by Cynthia Dunlop Why Vector Size Matters Apr 1st 2024 8:23am, by Jonathan Ellis API Design Is Pretty Bad -- Here's How to Fix It Apr 3rd 2024 6:01am, by Lebin Cheng Enhancing Business Security and Compliance with Service Mesh Apr 1st 2024 10:00am, by Ninad Desai Linux xz Backdoor Damage Could Be Greater Than Feared Mar 31st 2024 10:00am, by Joab Jackson Malicious Code in Linux xz Libraries Endangers SSH Mar 30th 2024 12:25pm, by Steven J. Vaughan-Nichols Chainguard: Outdated Containers Accumulate Vulnerabilities Mar 29th 2024 3:00am, by Joab Jackson Platform Engineering Operations CI/CD Tech Careers Tech Culture DevOps Kubernetes Observability Service Mesh Using an Internal Developer Portal for FinOps Visibility Apr 3rd 2024 9:05am, by Aidan O'Connor Platform Engineering Is Not Just about the Tools Mar 28th 2024 10:00am, by Atulpriya Sharma Platform as a Product in 4 Steps Mar 28th 2024 6:57am, by Aeris Ransom How Spring and Java Shaped Internal Developer Platforms Mar 26th 2024 1:10pm, by Charles Humble How TV 2 Prioritizes and Measures Developer Experience Mar 25th 2024 3:00am, by Jennifer Riggins How to Code First with Design-First Benefits Apr 2nd 2024 7:16am, by Rak Siva Linux: Migrate the Home Directory to Another Drive Mar 30th 2024 6:00am, by Jack Wallen Infrastructure as Code Is Dead: Long Live Infrastructure from Code Mar 29th 2024 10:00am, by Asif Awan Why Observability Can't Measure What Really Matters Mar 29th 2024 9:16am, by Vicki Walker How We Added QUIC Support to OpenSSL without Patches or Rebuilds Mar 28th 2024 12:07pm, by Michael Vernik and Roman Arutyunyan Beyond Git: A New Collaboration Model for AI/ML Development Apr 2nd 2024 10:00am, by Gorkem Ercan Stack Exchange Expands Beyond Q&A Format with 'Discussions' Mar 27th 2024 6:00am, by David Cassel Evolve Manual and Templated Dockerfiles with Automation Mar 26th 2024 10:33am, by Rak Siva Platform and Mobile App Engineering: How They Work Together Mar 22nd 2024 9:00am, by Usman Siddiqui Building Polyglot Developer Experiences in 2024 Mar 19th 2024 10:14am, by Mauricio Salatino Can Military Veterans Alleviate Your IT Hiring Woes? Mar 28th 2024 9:20am, by Joe Fay Using AI to Improve Bad Business Writing Mar 26th 2024 5:00am, by Jon Udell Developers Share What Helped Them Land New Roles Mar 25th 2024 6:45am, by Jeff James US Tech Cannot Comprehend the Digital Nomad Way of Life Mar 23rd 2024 3:00am, by Paul Scanlon Tech Works: How to Identify and Address Burnout on Your Team Mar 22nd 2024 5:00am, by Jennifer Riggins Breaking the Big Tech Grip: An AI Playing Field for Everyone's Benefit Apr 3rd 2024 10:00am, by Tim Davis How Should We Define 'Open' AI? Mar 31st 2024 6:00am, by David Cassel Stack Exchange Expands Beyond Q&A Format with 'Discussions' Mar 27th 2024 6:00am, by David Cassel How Spring and Java Shaped Internal Developer Platforms Mar 26th 2024 1:10pm, by Charles Humble How TV 2 Prioritizes and Measures Developer Experience Mar 25th 2024 3:00am, by Jennifer Riggins Will Spotify Open Source its Microservices Framework? Apr 2nd 2024 7:45am, by Loraine Lawson Infrastructure as Code Is Dead: Long Live Infrastructure from Code Mar 29th 2024 10:00am, by Asif Awan Platform Engineering Is Not Just about the Tools Mar 28th 2024 10:00am, by Atulpriya Sharma Platform as a Product in 4 Steps Mar 28th 2024 6:57am, by Aeris Ransom How TV 2 Prioritizes and Measures Developer Experience Mar 25th 2024 3:00am, by Jennifer Riggins Kubecost 2.2 Covers Carbon Cost Monitoring (and More) Apr 2nd 2024 3:00am, by Meredith Shubel KubeCon Europe: WebAssembly, eBPF Are Huge for Cloud Native Mar 29th 2024 8:24am, by B. Cameron Gain OVHcloud Unveils Roadmap to Take on Hyperscalers from Europe Mar 29th 2024 6:14am, by B. Cameron Gain Kubernetes and AI: Are They a Fit? Mar 28th 2024 12:27pm, by Alex Williams KubeCon24: Mirantis Doubles Down on Open Source Mar 28th 2024 8:03am, by Steven J. Vaughan-Nichols Why Observability Can't Measure What Really Matters Mar 29th 2024 9:16am, by Vicki Walker Large Language Model Observability: The Breakdown Mar 28th 2024 1:46pm, by Alex Williams Will the Real Test Observability Please Stand up? Mar 27th 2024 7:27am, by Ken Hamric 10 Ways Kubernetes Observability Boosts Productivity, Cuts Costs Mar 27th 2024 6:08am, by Eric Schabell 'Observability' Is Not Observability When It Comes to Business KPIs Mar 26th 2024 10:00am, by Eric Futoran Enhancing Business Security and Compliance with Service Mesh Apr 1st 2024 10:00am, by Ninad Desai Some Linkerd Users Must Pay: Fear and Anger Explained Feb 28th 2024 9:21am, by B. Cameron Gain Buoyant Revises Release Model for the Linkerd Service Mesh Feb 21st 2024 9:30am, by Joab Jackson Istio Advisor Plus GPT: Expert System Meets AI for Service Mesh Dec 14th 2023 12:15pm, by Steven J. Vaughan-Nichols Using JWTs to Authenticate Services Unravels API Gateways Nov 8th 2023 6:53am, by Christian Posta and Peter Jausovec 2024-03-31 06:00:12 How Should We Define 'Open' AI? AI / Data / Tech Culture How Should We Define 'Open' AI? The term "open" has no agreed-upon definition in the context of AI, and is applied to widely divergent offerings with little reference to a stable meaning. Mar 31st, 2024 6:00am by David Cassel Featued image for: How Should We Define 'Open' AI? VOXPOP Try our new 5 second poll. It's fast. And it's fun! What Is our P(doom) Level? P(doom) is Silicon Valley's latest hot metric, an approximation of how worried you are that AI, once fully cognizant, will destroy humanity. Inversely proportional to e/accs (the "effective accelerationists"). 0-25: AI will usher in a new age of utopianism. 26-50: What, me worry? 51-75: We'll probably be OK with some government oversight. 76-100: I'm stocking up on batteries and canned food now. I HAVE AN OPINION We'd love to hear what you think. "Our vision is a world where all benefit from the unfettered exchange of information," explains the official website for the National Information Standards Organization. Accredited by the American National Standards Institute (ANSI), it's the nonprofit standards group for library/publishing/bibliographic applications for both "traditional" and new technologies. But there's one very important standard that's still developing in the world: What exactly constitutes an open artificial intelligence, or "open AI"? It's a question Archive.org's deputy director Thomas Padilla tackled this year in his keynote for that year's NISO Plus conference. "It's a talk about the AI," Padilla promised his audience. "But like most talks about technology, it ends up being to talk about us and about what we want to achieve." Is Llama 2 Open Source? While the Open Source Initiative is still working on their official definition of "open source AI" -- through a consultative community process -- some important distinctions are already clear. Identifying five characteristics that open AI should have, Padilla said he believes "open" AI should be reusable, with users weighing trade-offs between performance and "broader reusability and the potential for others in our community to build upon the bottom model." But how does that play out in the real world? Meta's Llama 2 declares it's open source, and free for research and commercial use -- but Padilla reviewed the license terms at its GitHub repository, "putting my librarian hat on." Under "Additional Commercial Terms," it warns that if a licensee has more than 700 million active monthly users, "You must request a license from Meta, which Meta may grant to you in its sole discretion." And otherwise, "You are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights." Padilla called that "some shade against the other big tech companies, for the most part" -- although Red Monk co-founder Steve O'Grady is fairly clear that that's not pure open source: useful clarification, which makes the targets clear. you can use it unless you work at, say, google. imagine if linux was open source, unless you worked at facebook. and that's just one of multiple field of use restrictions, any one of which means it's not open source. https://t.co/x2CYrKLifF -- steve o'grady (@sogrady) July 19, 2023 But Padilla thinks there's another clause even more deserving of a "Llama side-eye." Screenshot from Thomas Padilla keynote at NISO Plus 2024 (Llama Side Eye to Meta's licensing). the license says users "will not use the Llama Materials or any output or results... to improve any other large language model (excluding Llama 2 or derivative works thereof)." Padilla feels that very much is "... not in the spirit of open source. It feels to me like it sort of -- it explodes the whole thing. How is that open source? It's not!" But Padilla adds that "other models have come onto the market that do show a degree of promise, I think, for having a stronger alignment with the spirit of open source ..." -- citing specifically the open language model OLMo. OLMo describes itself as a "truly open LLM and framework," and emphasizes that "all code, weights, and intermediate checkpoints are released under the Apache 2.0 License." Mysterious Answers Meta's Llama isn't the only game in town. ChatGPT's developers even named their technology OpenAI -- and here, Padilla tells his audience, "I also believe OpenAI is open." But there's still room for criticism. Transparency is part of Padilla's definition of openness -- it plays a role in the "integrity of knowledge"-- defined as making sure authors and "knowledge producers" are credited. Instead, today content producers are being "ghosted," as Padilla sees it -- with crediting either not existing "... or it's vaguely referenced to. I think that's bad." With much of today's generative AI, it's more like mysterious answers summoned from a magical seance. "You get responses like, 'I don't have access to how I provided you this answer.'" Responses like that would never fly for a peer-reviewed paper, Padilla points out: I asked Chatgpt how much of its data comes from Wikipedia: I don't have access to my training data, but I was trained on a diverse range of data from the internet, including sources such as books, websites, and other texts to develop a wide understanding of language. -- James (Jim) McTague (@Mctaguej) December 8, 2023 There's another issue. In February, Fast Company spoke to the head of the OLMo project, AI2 senior director of research Hanna Hajishirzi, about Meta's models. Their conclusion? The models were very valuable, but the data behind them still wasn't available. "We don't understand the connections starting from the data all the way to capabilities," the magazine noted. And even beyond that, "The details of the training code is not available. A lot of things are still hidden." So another aspect of transparency is good documentation -- and here, Padilla sees promise in Hugging Face's model cards. When users upload a model to Hugging Face's tools, they're prompted to specify the model's parameters and which datasets were used (according to Hugging Face's definition of model cards) -- as well as the model's intended use, and its "potential limitations, including biases and ethical considerations." Padilla says he'd "love to see more uptake of this" -- and "possibly even a requirement" to upload them to Hugging Face. "Perhaps that's around the corner." "Without transparency, you have no assurance of knowledge integrity," Padilla tells the audience. "And if you have no knowledge integrity -- then why bother be involved in something like this at all? I really think it weakens the value proposition of the entire thing." How Should AI Be Held to Account? In addition, open AI should also very much be accountable, Padilla believes -- meaning that it's developed (and used) according to specific community needs. For "accountability," Padilla applauded policy and regulation and executive orders "attempting to enshrine certain protections and safeties to guide the development and implementation of AI," including the White House's October "Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence." But Padilla also cited the real-world accountability principles espoused by Distributed AI Research Institute (founded by AI researcher Timnit Gebru), on the organization's Research Philosophy page. One of their aims is to reduce "the distance between researchers and community collaborators." Another aim of theirs is to focus on trust and time, to forge "meaningful relationships with communities." And they in particular are talking about minoritized or marginalized communities that are not often at the center of AI research, but are often on the receiving end of negative impacts of AI research that does not take into account the lived experience. "I think DAIR provides us a nice roadmap," Padilla said. Sustainability Padilla said open AI should be adopted in a "sustainable" way, with an awareness of "interdependence, threats and opportunities." To that end, Padilla proposes three ways of looking at it. To demonstrate what he calls the "exploded" view of interdependence, Padilla showed a 2018 data visualization exhibited at New York's Museum of Modern Art which (according to its gallery label) "analyzes the vast networks that underpin the 'birth, life, and death' of a single Amazon Echo smart speaker." (It moves from the periodic table of elements, Padilla told the audience, "up to the strip mining that needs to happen to gain the raw materials to produce a Dot, up to smelters and refiners and then favorable labor conditions, up to the component manufacturing and so forth. "I think it's one way to kind of get a sense of the interdependency that's at play when we adopt, not just AI, but any technology." There's also a "systems" view -- Padilla cites the world-systems theory, which explores how different parts of the world get relegated to their roles in the production of what Padilla calls "highly polished commodity products." And Padilla also provided an example for the kinds of questions explored in the "replacement" view. In October, a U.S. House subcommittee on technology oversight heard testimony from Emily M. Bender, a linguistics professor at the University of Washington. "I find that the phrase 'artificial intelligence' is best understood as a marketing term," Bender had said, "and one which only muddies the waters. It is clearer to talk about automation." This leads to different questions, including: * What is being automated? * Who is automating it and why? * Who benefits from that automation? * Who is being harmed and what recourse do they need? * Who has accountability for the functioning of the automated system? * What existing regulations already apply to the activities where the automation is being used? The arrival of AI shouldn't just be about worrying about which jobs will be replaced. Ideally, we also want open AI to have an "affirmative" impact, Padilla believes -- supporting the evolution of roles in organizations. Marketing Terms vs. Technical Descriptors Padilla had opened his talk with a damning observation from a paper written last August. While earning his PhD at Carnegie Mellon, David Gray Widder had teamed up with Sarah West, managing director of the AI Now research institute, and Meredith Whittaker, president of the Signal Foundation (a nonprofit focused on open source privacy technology and developers of the Signal messaging app). "We find that the terms 'open' and 'open source' are used in confusing and diverse ways," the researchers wrote, "often constituting more aspiration or marketing than technical descriptor, and frequently blending concepts from both open source software and open science." "This complicates an already complex landscape, in which there is currently no agreed on definition of 'open' in the context of AI, and as such the term is being applied to widely divergent offerings with little reference to a stable descriptor." So we're currently living in a world where licensing terms are hard to assess, Padilla said, not to mention the jolt of discovering a tech stack that is highly proprietary -- and expensive. "We think we know what open means and then we start experiencing kind of like this M.C. Escher-like experience...." Screenshot from Thomas Padilla keynote at NISO Plus 2024 (when open AI feels like an M C Escher painting) "I think it behooves us to have, you know, a more concrete sense of what we want from AI and the knowledge work that we do collectively together." TRENDING STORIES YOUTUBE.COM/THENEWSTACK Tech moves fast, don't miss an episode. Subscribe to our YouTube channel to stream all our podcasts, interviews, demos, and more. SUBSCRIBE Created with Sketch. [82081813-7] David Cassel is a proud resident of the San Francisco Bay Area, where he's been covering technology news for more than two decades. Over the years his articles have appeared everywhere from CNN, MSNBC, and the Wall Street Journal Interactive... Read more from David Cassel SHARE THIS STORY TRENDING STORIES SHARE THIS STORY TRENDING STORIES TNS DAILY NEWSLETTER Receive a free roundup of the most recent TNS articles in your inbox each day. [ ] SUBSCRIBE The New Stack does not sell your information or share it with unaffiliated third parties. By continuing, you agree to our Terms of Use and Privacy Policy. ARCHITECTURE Cloud Native Ecosystem Containers Edge Computing Microservices Networking Serverless Storage ENGINEERING AI Large Language Models Frontend Development Software Development API Management Python JavaScript TypeScript WebAssembly Cloud Services Data Security OPERATIONS Platform Engineering Operations CI/CD Tech Careers Tech Culture DevOps Kubernetes Observability Service Mesh CHANNELS Podcasts Ebooks Events Newsletter TNS RSS Feeds THE NEW STACK About / Contact Sponsors Sponsorship Contributions roadmap.sh Community created roadmaps, articles, resources and journeys for developers to help you choose your path and grow in your career. Frontend Developer Roadmap Backend Developer Roadmap Devops Roadmap (c) The New Stack 2024 Disclosures Terms of Use Advertising Terms & Conditions Privacy Policy Cookie Policy FOLLOW TNS FOLLOW TNS TNS DAILY SUBSCRIBE