232 | 🚨🔒 How to Protect Your LinkedIn Data from Training Their AI

In today’s Brainyacts:

1. You are the last mile of training LLMs

2. LinkedIn is using your data - how to stop them

3. Microsoft lights up nuclear plant to power AI and other AI model news

4. Darth Vader’s voice lives and more news you can use

   👋 to all subscribers!

To read previous editions, click here.

Lead Memo

🗝️ 💪 Revisiting the 5P’s of prompting (aka, you are the last mile of AI training)

A law student perspective by: Ishani Sachdeva, Jenna Sur, and Nicole Eckert

For the novice user of generative AI and LLMs, it is easy to assume that AI is THINKING. We’ve all had that experience. However, that is not the case. The LLM models are designed to analyze patterns in words and convert them into a sequence of numbers, known as vectors. While the models may seem to understand and process questions like a human, they really are just doing math. That said, they still rely heavily on human input. Human involvement is crucial for training and fine-tuning these models. What many users of LLMs don’t realize is that their prompts are the crucial “last mile” of training for these models.

So, let's talk about prompts. A vague prompt like "Generate an image of students" will produce—you guessed it—vague results. Do you want law students? High school students? A courtroom setting or a classroom? Should the tone be formal and buttoned-up or casual and creative? The more specific you are, the better your AI-generated outputs will be, and it saves you from getting random, off-the-wall results. So, next time, skip the ambiguity and be crystal clear—your AI (and your sanity) will thank you!

Our Experiment: The Prompt Paradigm

To explore the impact of prompts, our Generative AI class conducted an enlightening experiment. We asked Chat-GPT and Claude to compose song lyrics about a law student learning AI. We then polished our prompts with descriptors like "a law student who is bored" and "you are an amateur and a bad songwriter." The stark contrast in responses was both surprising and informative.

Intrigued by our lyrical findings, we extended our investigation to the visual realm. Enter DALL-E 3, our graphical test subject. We began with a simple prompt: "Create an image of students in a classroom." Predictably, the result was as generic as it was uninspiring.

From this baseline, we progressively enhanced our prompt. We added context, described the students and professor in detail. Our objective was to generate an image that captured the essence of our Generative AI classroom, featuring our esteemed Professor Kubicki.

Inspired by the results, we decided to push the boundaries further. We introduced an artistic element to our prompt, instructing DALL-E 3 to adopt the persona of "an artist known for your abstract and experimental painting style." The outcome was truly remarkable, transforming our classroom scene into a creative piece that blurred the lines between reality and abstraction.

The journey from a generic classroom image to this artistic representation was nothing short of astounding. What's more, these dramatic shifts were achieved through subtle adjustments in our prompting technique.

This experiment serves as a powerful reminder of the critical role that effective prompting plays in harnessing the full potential of generative AI.

For legal professionals, the implications are significant: mastering the art of prompting can lead to more precise and relevant AI-generated content, potentially revolutionizing legal research, document drafting, and case analysis. As we navigate the integration of AI in legal practice, the ability to craft effective prompts may well become an essential skill in the modern lawyer's toolkit.

The 5 P’s Framework

The key to unlocking the potential of AI lies in how you prompt it, and that's where the 5 P’s of Effective Prompting come into play.

1. Prime: Start with Context

Providing background or context when interacting with AI is crucial. Clearly define the issue or scenario to ensure the AI is aligned with your goals from the start.

2. Persona: Set the Tone and Expertise

AI can adapt to different styles and tones based on how you prompt it. Are you looking for a formal, expert-level legal analysis or a casual brainstorming session? By incorporating the right persona—whether it's professional, academic, or conversational—you'll get results that match your specific needs.

3. Privacy: Protect Sensitive Information

When interacting with AI, always be mindful of the data you're sharing. Avoid disclosing sensitive or personal information unless you are certain the AI tool is secure. Treat the AI like a public forum—think before you share.

4. Product: Define Your Goal

Before prompting, be clear about what you want to achieve. Are you seeking legal research, case analysis, or creative solutions? By outlining the product or outcome you desire, you guide the AI towards delivering a focused and relevant response.

 5. Polish: Refine for Precision

Once the AI provides an initial response, it's time to refine. Ask follow-up questions, dig deeper into the reasoning, and don’t settle for a surface-level answer. One effective strategy is to request both positive and negative perspectives on the AI's output.

Why the 5 P’s Matter

Busy professionals often make the mistake of typing a single question into an AI tool and expecting the perfect answer on the first try. Unlike Google, AI requires engagement and fine-tuning to deliver optimal results. By applying the 5 P’s, you actively shape the conversation and maximize the likelihood of success.

And remember, AI is just like a law student after finals: it'll do its best, but sometimes it needs a little nudge to stay on track. So next time you engage with an AI tool, use the 5 P’s—and maybe offer it some virtual coffee while you're at it!

P.S. Note from Josh. For me first class I wore a Beatles t-shirt in homage to the “All you need is attention” paper that introduced the transformer technology that fuels the current LLM/generative AI capabilities. They got that title from “All you need is love.”

Spotlight

🤯 👂 LinkedIn used your data to train Microsoft AI

🚨 🚨LinkedIn scraped user data for training before updating its terms of service.

🔒 Protect Your LinkedIn Data from AI Training | Step-by-Step Guide

Learn how to opt out of LinkedIn's default setting that uses your data to train AI models. This quick tutorial shows you:

✅ Where to find the privacy settings

✅ How to navigate to the AI data usage option

✅ Steps to turn off data sharing for AI training

Don't let your professional information be used without your consent! Watch now to take control of your LinkedIn data privacy.

AI Model Notables

► The Three Mile Island nuclear plant is being restarted. All 835 megawatts will go to power Microsoft's data centers. This is both the first time a U.S. nuclear plant will come back into service after being decommissioned, and that an entire plant’s output will go to one customer.

► OpenAI threatens to ban users who probe its new models (o1).

► Mumsnet, the long-running, enormously popular, controversy-spurring UK-based parenting forum for mothers is threatening legal action against OpenAI.

► Amazon launched Amelia, an AI assistant for third-party sellers to help resolve account issues and access sales data, now available in beta for select US sellers.

► Nice helpful overivew of leading LLMs to use and try.

► Meta, Spotify, and Ericsson, along with 57 other companies including have written an open letter to the EU calling for AI regulatory certainty. The letter was published today in multiple outlets across the EU.

► Mark your Calendar: Next week is Meta’s Connect event - expect some AI news to flow out.

News You Can Use:

➭ Colorado’s new AI bill is an “amorphous, poorly defined big accountability stick” according to LinkedIn founder Reid Hoffman.

➭ Data center emissions probably 662% higher than big tech claims. Can it keep up the ruse?

➭ Scientists preparing “Humanity’s Last Exam” to test powerful AI.

➭ In Venezuela, AI news anchors aren’t replacing journalists. They’re protecting them.

➭ Does BARBRI wants to get ahead of its competition and offer AI training before other Bar prep sites?

➭ James Earl Jones' controversial AI decision will let Darth Vader live on, but it raises concerns among actors.

➭ The Animation Guild is currently negotiating a new contract with the major Hollywood studios seeking protections against AI.

➭ Lawyers say delivering work faster is the biggest benefit of AI, yet it is prompting many to rethink the suitability of their pricing structure

Who is the author, Josh Kubicki?

Some of you know me. Others do not. Here is a short intro. I am a lawyer, entrepreneur, and teacher. I have transformed legal practices and built multi-million dollar businesses. Not a theorist, I am an applied researcher and former Chief Strategy Officer, recognized by Fast Company and Bloomberg Law for my unique work. Through this newsletter, I offer you pragmatic insights into leveraging AI to inform and improve your daily life in legal services.