Your Data, Your Moat

Welcome to another edition of the best damn newsletter in AI.

As a reminder, we’ve opened up Reader Questions. Our mission is to make using AI accessible to every business, and a huge part of that is making sure you’re not stuck and that we continue answering diverse questions. So, submit your question here. We’ll review every single one and cover answers here in the newsletter.

Here's what we're covering today:

• How to find valuable data for your AI models

• A fun ChatGPT tip for the wannabe green thumbs out there

• What we’re reading this week about AI

• & more

Let’s get to it! 👇

TODAY’S PERSPECTIVE

How to *actually get value out of your data.

In today's post-ChatGPT world, data isn't just valuable—it's the cornerstone of competitive advantage. 

AI wearables are collecting it in troves. Meta might make people pay for their data to not be collected. Every enterprise AI company makes big claims about your data, your moat.

Just like the big data trends of the past, companies are being pushed into a “collect everything” trap. Record every meeting. Document every decision. Log every customer interaction. “We need it for the AI”. 👿

Some of that is true.

But anyone who’s worked with data will tell you: the hard part isn’t getting data, it’s getting value out of it. 

Today, we’re going to help you navigate HOW to find valuable data and common TYPES of valuable data so that you can speed up your time to get value out of it.

How to find valuable data

1. Unique to you. Areas of the business where systems and processes are unique to you (you wouldn’t find that ‘How to escalate X issue” support guide on the internet)

2. Inhuman quantities. There’s a reason that customer interaction data is so valuable. Most companies have too many customers and too many interactions for people to remember. Collecting this is valuable.

3. Gold standards. Just like many people, AI learns best by being shown what to do, not just told. So collecting gold standards is valuable.

Types of data to collect & get value out of

Here’s the 6 types of data you’re probably generating in your day to day work and how to get started getting value out of them.

Instructions

What they are: Step-by-step guides or manuals that explain how tasks should be done

How to get value: You can automate some routine writing-heavy tasks by uploading these documents to a custom GPT. Or for more complicated processes, you can use these instructions to create interactive training GPTs for onboarding new employees.

Frameworks

What they are: Structured templates or outlines that guide thought processes and decision-making

How to get value: Frameworks are great as a baseline for AI tools like ChatGPT to generate complex documents like strategic plans and project proposals. By feeding structured frameworks into AI, you can teach it your preferred methodologies and get higher quality outputs.

Examples

What they are: samples that demonstrate expected inputs or outputs, which can serve as reference points for what “good” looks like.

How to get value: Train AI models with these examples to recognize standards of quality and output in specific contexts, such as customer service responses or technical support solutions. This training can help AI systems better understand the nuances of your business operations and improve their accuracy.

Q&As

What they are: These are question and answer documents, often found in FAQs or support documents.

How to get value: Upload these into AI-driven chatbots or support tools to provide quick and consistent responses to customer inquiries. We are fans of Chatbase.

Transcripts

What they are: Verbatim records of conversations, meetings, or interviews, which are typically less structured and more verbose than Q&As.

How to get value: We actually recommend NOT using these. Instead, summarize and analyze these transcripts using AI to extract key themes, action items, or decisions to make Context (see next section).

Context

What they are: Summarized overview of concepts, key points, and other information. These can be dense and less logical or human-readable, focused on condensing as much context that the LLM might need for a task as possible.

How to get value: Build these to summarize quarterly business results, entire customer accounts, industry analysis, and more. Think of creating these as “Build an executive summary on XYZ”. (and yes, you can use ChatGPT to do that)

USE CASE

ChatGPT: Your Green-Thumbed AI Pal

If you love your garden but hate gardening, this one is for you.

From identifying plants to caring for them, here's how you can use ChatGPT to make green thumbs greener:

• Plant Identification: Snap a pic of any plant, and ChatGPT will tell you what it is.

• Personalized Suggestions: Based on where you live, ChatGPT can recommend plants that will thrive in your climate.

• Lawn and Plant Care Advice: Take a picture of your dying plant and ChatGPT will diagnose the issue and offer advice to save it.

• Artificial Plant Detection: ChatGPT can even distinguish between real and artificial plants!

You’ll never garden the same way again.

LINKS

For your reading list 📚

AI is shaping the future of MANY industries...

• Boston Dynamics is retiring its pioneering humanoid robot Atlas to focus on more commercial technologies, hinting at a successor coming soon.

• Adobe is developing new Firefly generative AI video capabilities for Premiere Pro, aiming to provide creative professionals with robust generative video options directly within Adobe's ecosystem.

• The Weather Company's new ReelSphere tool helps meteorologists quickly create localized weather forecast videos for multiple platforms using AI voiceovers and automated graphics.

AI regulation and safety are still progressing...

• The UK is drafting legislation to regulate AI models following the establishment of the UK's AI Safety Institute to evaluate powerful AI systems.

AI may also disrupt the way we interact with social media...

• TikTok plans to add virtual influencer AI avatars to promote products on its TikTok Shop, potentially competing with sponsored ads from human influencers.

AI in education is not without its challenges...

• The Texas Education Agency is implementing an AI scoring system for open-ended questions on STAAR exams, but some educators are concerned after seeing more zero scores during initial use.

AI image generators are struggling with representation...

• Major AI image generators, including Meta's Instagram, Google Gemini, DALL-E, and Midjourney, struggle to generate images of Asian men with white women without relying on stereotypes or producing odd results, revealing potential racial biases in training data.

Have a burning AI question? Or stuck trying to help your team adopt AI?

We read every single question submitted and use these to inform the topics we go deep on. 👇️ Submit your question below!

That's all!

We'll see you again soon! Thoughts, feedback and questions are much appreciated - respond here or shoot us a note at [email protected].

Cheers,

🪄 The AI Exchange Team