130 | 😎 🔥 Bro, that's fire!

In today’s Brainyacts:

1. Briainyacts Course Update!

2. Knowledge Cutoff vs Internet Browsing

3. Bard shares your conversation in Google search results and more AI model news

4. An iPhone for AI? and other related news

5. Bro! That’s fire!

👋 to new subscribers!

To read previous editions, click here.

Lead Memo

🧠 💥 Brainyacts Course Update

I am in the midst of updating my course - The Generative AI Blueprint. The first 2 Modules contain over 3 hours of content that will take you from beginner to advanced user of GenAI tools. It gives you the secret sauce on how to prompt, ways to work with tech for practice of law and business of law needs, and unlocks serious skills for you.

Now I am about to release Module 3: Advanced Skills - 40 mins of jam-packed highly pragmatic and useful skills.

⚡️ 🟢 If you spend 40 minutes in this module, I guarantee you will be able to save at least 1 hour a week (at least) in work!

I show you how to:

1. Have OpenAI create a slide deck for you using nothing more than prompts. You get it to write a detailed presentation complete with the content. Then you get it to write in code! You copy and paste this code into PowerPoint and BAM! you will have all your content on slides formatted as you want. Then you can quickly use the Design features in PPT to get to a great-looking deck. It might take you 5 mins!

2. How to get OpenAI to do real research that gives you real resources and citations - no hallucinations guaranteed.

3. How to use OpenAI as your mobile assistant - mobile as in walking and talking to it as it works to turn your wandering rambles into tightly focused and organized tasks and work product.

4. How to use Custom Instruction to make your prompting easier, return better responses, and extract more value from OpenAI. I will be giving away 50+ legal-specific templates you can use!

Plus . . .

I am in the process of building out Module 4 soon. This will be a deep dive into other tools like Microsoft Copilot and the Google AI stack.

Current students get all of this for free. They already paid to take the course and I promised to keep sharing more lessons and insights!

If you want it, sign up now. It is $189 right now but once I release Module 3 (within 2 weeks- likely very soon), the price will go up to $249.

🎓 Yes, you get a certificate if you want one.

💰 Yes, I am confident you can expense this if you choose to. It is highly valuable educational content. I present snippets of this for CLEs, state bar associations, in-house teams, and firms. It generates real value.

Here is the course link: https://www.brainyacts.academy/courses/brainyacts-generative-ai-blueprint/preview

Some reviews to share:

Spotlight

The Two Faces of AI Understanding: Training Data and Real-Time Browsing

In the fast-paced world of technology, especially when it comes to AI and machine learning, people often grapple with how "smart" these systems actually are. One confusion often emerges around two seemingly intertwined concepts: the "knowledge cutoff date" and "real-time internet browsing." To simplify this complex web, let's break it down into chewable pieces.

OpenAI is bringing back internet browsing via Bing. They released it a few months ago but took it down due to some issues. It is being rolled out once more.

OpenAI also just updated its cutoff date for GPT-4, not ChatGPT.

Previously it was September 2021.

ChatGPT’s Cutoff date remains September 2021.

The Backbone: Training Data Up to Cutoff Date

Depth of Understanding

Imagine you're a chef trained for years under various masters, flipping through decades-old cookbooks, and spending countless hours in the kitchen. You'd probably cook up a storm. In a similar fashion, GPT-4 has been trained on a massive buffet of data up until January 2022. This training allows it to give answers that are nuanced, deeply contextual, and richly layered.

Static but Reliable

While a trained chef can whip up classic recipes with her eyes closed, she might not know the latest fusion dish that's the rage on social media. Likewise, GPT-4 doesn't have access to developments post-January 2022. But what it does offer is reliable information based on highly reputable sources, much like your chef's dependable beef bourguignon.

The Flashy Cousin: Real-Time Internet Access

Real-time but Surface-level

Now, consider our chef suddenly gets live Twitter feeds on the newest food trends. She can talk about them, but can she understand the intricacies of these dishes without having cooked them? When GPT-3 has real-time browsing, it can fetch up-to-the-minute data, but this doesn't mean it has an in-depth understanding of it.

Speed vs. Quality

Just as our chef could quickly rustle up some trending avocado toast, GPT-4 could pull in the latest news or papers. But speed often compromises quality. While it can get real-time data, there’s no assurance that it’s from reputable sources or that it understands its nuanced implications.

Limited Context

Imagine our chef hears about a new Korean dish but has never cooked Korean food. Her understanding will be quite limited, won't it? Similarly, real-time data that GPT-4 pulls would lack the contextual richness that comes from being integrated into a broad base of training data. It will just return basic info with a link (maybe).

The Overlap and The Gap

Overlap

Whether it's a chef's seasoned hands or her real-time awareness of trends, both aim to please the palate. Similarly, both OpenAI’s trained model and browsing capabilities aim to provide accurate and relevant information.

Gap

But here's the kicker. Just as a chef can't instantly master a dish from a live tweet, GPT-4 can't deeply comprehend real-time data without the broader context.

The Real Story: The Devil is in the Details

Yes, there has been a recent update to GPT-4’s training data, moving the cutoff from September 2021 to January 2022. But this isn't the same as it suddenly being able to fetch live data from the internet.

Using the browser capability, the data will be similar to you manually copying and pasting a webpage into your conversation. Useful? Absolutely. The same as training? Not even close.

AI Model Notables

• Google Bard has been sharing your conversations in Google’s search results!

• Google adds a switch for publishers to opt out of becoming AI training data (too bad Bard didn’t have a similar one for us users 😂)

• OpenAI reinstates ChatGPT's internet browsing privileges (but it is rolling out slowly)

• OpenAI’s GPT-5 and 6 are coming

• Meta’s new AI features

News You Can Use:

➭ An AI-wearable? OpenAI, Jony Ive (former Apple lead designer under Jobs) in talks to raise $1 billion from SoftBank for AI device venture - they’ve previously worked together on the Worldcoin AI retina scanner.

➭ Hollywood Writers Win Promise: No Robots Will Get Screen Credits. Contract includes some curbs on AI, but not an outright ban. Studios may use existing scripts to train AI software.

➭ Job Alert: Read the job description of the Exec. Director of Conversational AI Job at JP Morgan - Pay seem high, low, or about right?

➭ 2023 State AI Legislation Summary - PDF (update this month)

➭ Canada aims to be the first country in the world with Official AI Regulations.

➭ Law librarians share their list of AI resources.

➭ Biden teases forthcoming executive order on AI.

➭ FTC is calling for AI-related research at PrivacyCon2024.

➭ Oklahoma has established an AI task force.

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.