How legal is all of this anyways?

Welcome to another edition of what we’re determined to make the best damn newsletter in AI. Here we’ll break down AI topics that matter, open your mind to use cases, and keep you ahead of the curve.

Our #1 goal is to be useful. So please shoot us an email 📩 if you have questions or feedback, and especially if you implement something we share!

Here's what we're covering today:

• A brief overview of the controversy between copyright and AI + a link to our research

• Prompt engineering deep dive w/ curated prompting tips

• What we're reading - and think you should be too

Let’s get to it! 👇

TODAY'S PERSPECTIVE

AI and Copyright Law: A Controversial Mix

As ChatGPT and other generative AI products have blown up, you've probably been asking yourself, how legal is all of this anyways?

You’re not alone - it’s a hot topic! 🔥

Here’s what we think you should know —

1) This isn’t new — AI researchers have been allowed to use copyright protected work scraped from the public internet to train AI models under the doctrine of Fair Use for years!

2) But recent court cases might change that perspective — In February, US Copyright Office ruled that AI art can't be copyrighted.

Then, in September, a comic book artist received first known US copyright registration with AI art contained in their work.

But in December, the US Copyright Office initiated a proceeding to revert that decision.

Other open cases including the legality of Andy Warhol's illustration of Prince and GitHub Copilot developers being sued for large-scale software piracy might beg to differ.

3) The explosion of AI in pop culture could drive change — Some argue that copyright protection in AI was too niche of a topic in the past to have robust legal rulings. But AI is quickly going mainstream thanks to ChatGPT, DALL-E, MidJourney, etc.

4) Using copyrighted work to train an AI to predict vs generate could be seen as two very different purposes. And established legal examples are based on prediction.

Ok so…? It’s impossible to predict the future here. But generally, we think the risk is low in most use cases of generative AI. One area we personally are avoiding is using artists specific names in art generation. We’d rather create something that feels more stylistically ours!

“Following the money” can also suggest a way these rulings might play out. So we asked ChatGPT who could benefit from a contrary ruling.

If you're interested in a further deep dive to understand the varying sides of the debate, we've synthesized our research on generative AI and US copyright law here!

Please note that this is not legal advice. Consult your lawyer with detailed questions.

USE CASE DEEP DIVE

5 minutes to write better prompts ✍️

The key to generating better AI outputs is leveling up your inputs. 

“Prompt engineering” is the process of discovering inputs that yield better outputs! Remember - “garbage in, garbage out.”

Go from a beginner -> advanced prompt engineer with these tips:

1. Spend time with the tools - become extremely familiar with OpenAI's GPT playground (you'll thank us later)

2. Use your expertise in other areas to guide your prompting - prime GPT with your expertise (like SEO best practices), then ask it questions. It's a fast study.

3. Become a prompt researcher instead of engineer - try as many different variations and formulations of your prompt as possible (pro tip: GPT can do this for you - see below👇)

Here's some specific prompting tips + example prompts to get you started:

1) Separate your ChatGPT threads for different topics

2) Type "continue" to get ChatGPT to finish its response

3) Set the context surrounding the task in your prompt

Example prompt - "This is a conversation between a customer and a friendly, helpful customer support representative. Customer: <input customer question>. Customer representative:"

4) Use variables within your input indicators

Example prompt - "Write a sales call follow up email in a familiar tone using the following inputs: Customer name: John Smith. Company: Amazon. Call action items: ask about his vacation, extend special pricing terms, schedule follow up call. Output:"

5) Give AI a structure to follow

Example prompt - "Using a structure which includes a hook, body, and call to action, generate a TikTok script for why people should walk more often."

If you found these tips helpful, check out our Ultimate Prompt Engineering Guide - a free resource we created with 40+ tips proven to help you prompt better. 

Happy prompting! 🎉

LINKS

For your reading list 📚

What's poppin' in the AI world right now

• Daily revenue for AI-avatar generator, Lensa, down 92% from its Dec '22 peak

• OpenAI to be valued at $29B if tender offer is completed

• Anthropic AI to develop GPT-3.5 competitor, Claude - it's not out yet, but we're definitely watching

And if you're really nerdy ...

• If you're interested in training your own LLM, here's a paper on training a language model on a single GPU in one day, and the code

• An 80+ page free guide on the ins and outs of prompting with DALL·E 2

That's all!

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

Cheers,

🪄 The AI Exchange Team