Decoding Biden's AI Executive Order
PLUS: Master ChatGPT Prompting with "WISER" Framework
Hey hey fellow attorneys,
Let’s dive into this week's newsletter, where we dissect Biden's AI stance, unravel the latest in the AI art copyright saga, and spotlight Singapore's innovative AI safety sandbox. Plus, get the lowdown on mastering ChatGPT with the "WISER" framework tailored for attorneys. And for a sprinkle of humor, don't miss our lawyer meme of the week! Ready to get enlightened? Let's dive in!
On the docket today:
Decoding Biden’s AI Executive Order: A Progressive Leap or a Regulatory Hurdle?
AI Art Copyright Battle: Judge Sides with DeviantArt and Midjourney, but Stability AI Remains in the Crosshairs
Singapore’s Bold Move: Launching AI Safety Sandbox with Global Tech Leaders
Master ChatGPT Prompting with “WISER” Framework
Hilarious Lawyer Meme
New here? → Subscribe
Chip Somodevilla/Getty Images
President Biden's Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence has elicited diverse reactions. Aimed at regulating powerful AI systems, its reception ranges from being seen as a landmark initiative to potentially restrictive.
The order underscores the ambition to strike a balance between AI's risks and rewards. It mandates federal agencies to establish AI standards across various sectors, emphasizing transparency by requiring companies to report on high-risk AI models. However, its effectiveness is contingent upon its execution by these agencies and potential legal battles.
Another perspective offers a tempered view, suggesting the order mainly consolidates existing AI initiatives without introducing groundbreaking proposals. The incorporation of elements from global AI regulations and its pragmatic approach to foundation model notifications are commended.
In contrast, there's critique for being overly prescriptive, potentially stifling AI research and innovation. Concerns about the order's data sharing legality and the bureaucratic challenges it might introduce are highlighted.
In summary, while some view Biden's AI Executive Order as a progressive step towards AI safety, others caution against its potential unintended consequences, emphasizing the need for a balanced approach to AI governance.
In a significant development regarding the contentious issue of AI art generators and copyright infringement, U.S. District Court Judge William H. Orrick has made a ruling in a lawsuit against Stability AI, Midjourney, and DeviantArt, all of whom utilize the Stable Diffusion text-to-image AI generator. The lawsuit, initiated by artists Sarah Anderson, Kelly McKernan, and Karla Ortiz, alleges that these companies used copyrighted artwork without permission to train their AI models.
Judge Orrick largely dismissed the claims against Midjourney and DeviantArt, emphasizing that neither company directly scraped images or trained the AI. The primary training dataset was assembled by LAION, a non-profit, which reportedly collected public images from the internet under Stability's direction. Consequently, Stability remains the primary defendant. The judge highlighted that for a copyright claim to be valid, there must be a 'substantial similarity' between the original and the AI-generated content. Simply put, AI outputs, even if based on copyrighted works, might not be similar enough to the original to constitute infringement.
However, the battle is far from over. The artists have been granted the opportunity to amend their claims and refile a more specific lawsuit. While Stability AI asserts its commitment to fairness, safety, and legality, the broader implications of this case could set a precedent for how AI-generated content is viewed in relation to copyright laws.
Singapore's Infocomm Media Development Authority (IMDA) and the AI Verify Foundation have introduced the "Generative AI Evaluation Sandbox." This initiative aims to consolidate current evaluation benchmarks and testing methods for generative AI (GAI) systems. The Sandbox serves as both a comprehensive catalogue and a platform for experimentation and development, allowing for independent, open-source, and third-party model evaluations. Major tech entities like Google, Microsoft, Anthropic, IBM, Nvidia, Stability AI, and AWS are actively participating.
Singapore's approach stands out in the global AI regulatory landscape. Instead of rushing into AI regulations, Singapore focuses on practical solutions that assist AI developers and businesses at the operational level, emphasizing AI safety from a design and technical perspective. This proactive approach contrasts with other jurisdictions, such as the EU and China, which have regulatory requirements but lack the practical guidance that Singapore offers.
The Sandbox initiative will:
Offer a research-based categorization of current evaluation benchmarks and methods.
Build a body of knowledge on how Gen AI products should be tested, involving various stakeholders in the AI ecosystem.
Develop new benchmarks and tests, especially in domain-specific and culturally specific areas.
Prominent companies, including Google, Microsoft, Anthropic, IBM, NVIDIA, Stability.AI, and AWS, have joined the Sandbox, contributing to a more robust testing environment. The collaboration between Anthropic and IMDA, for instance, focuses on red-teaming methodologies tailored for Singapore's diverse linguistic and cultural landscape.
Singapore continues to make significant strides in fostering responsible AI, with the Sandbox marking another step forward by leveraging the global open-source community's collective power and contributions.
Master ChatGPT Prompting with “WISER” Framework
I really want attorneys to harness the full potential of ChatGPT, as it offers invaluable assistance in areas like drafting, simplifying legal terms, brainstorming case strategies, enhancing time management and productivity, and simulating client interactions. Mastering the art of prompting is essential for legal professionals to utilize this tool effectively.
So, I've adapted an existing formula just for attorneys: the "WISER" framework.
👉 WISER = Who + Instructions + Sub Tasks + Examples + Review 👈
👇 Let's break it down 👇
W - who is it. Begin your prompt with "you are a ____." You are a constitutional law expert. You are a corporate attorney. You are a criminal defense lawyer with a flair for dramatic courtroom moments. Assign it a specialization. Set the scene.
I - instructions. Provide clear directives—what do you want the model to achieve? If it’s a corporate attorney, instruct it to simplify a complex merger process. Ensure your instructions are explicit and straightforward.
S - sub tasks. Segment your request into smaller tasks for better results. For a constitutional law expert, ask it to define a legal term, its origins, its implications, and its significance in modern law. Or request an explanation of the First Amendment rights for different age groups (5 yo, then 13 yo, then 21 yo). Guide it through a step-by-step process.
E - examples. ChatGPT excels at imitation, so offer examples or a template. If you want a legal brief, provide a sample brief. If you're drafting a contract clause, give a starting point. Multiple, diverse examples yield the best results.
R - review. Did the output meet your expectations? Was something missing? What's the next step? Utilize the chat interface; ask for clarifications, delve deeper, or expand on ChatGPT’s initial response. Request a more concise explanation, or a detailed breakdown. Don't hesitate to restart or initiate a new chat thread.
Meme Of The Week:
That’s all for today!
Catch ya’ll on LinkedIn
What'd you think of today's email?