- The Brainyacts
- Posts
- 030 | 800lb Gorilla
030 | 800lb Gorilla
Brainyacts #30
The Generative AI newsletter for legal pros everywhere.
🚀 Hello! We have reached 830+ subscribers in 29 days! I cannot thank you enough for subscribing!! 🎉🔥
NEW SUBSCRIBERS, to read previous posts, go here.
30th Day since I began this challenge.
70 more days to come. I cannot wait!
Thank you for reading, sharing, and engaging! You keep the pressure on me to deliver every day. I won’t let you down!
Ok, today we will:
address the 800lb Gorilla
summarize books
share some fun and curious AI sites
talk news you can use and news you can lose
🦍 The 800lb Gorilla
You have likely heard that ChatGPT hallucinates or fabricates lies with authority. This is indeed true and I am about to show you an example. But first I want to share 2 best practices with you. There are more best practices but these 2 are highly relevant to this situation.
#1 Always review the ChatGPT response. Never just copy-n-paste it and send it along its merry way.
#2 Be careful of getting your prompts from Twitter, LinkedIn, or buying them. Please never buy prompts. They are likely garbage.
Ok, let me show you a really good lie.
I see many prompts as I work to generate use cases and strong stress-tested prompts to deliver high-quality results for you. The prompt below was shared by a GenAI influencer raking in big $$ and growing his newsletter by 1000s/day. It sort of drives me nuts - but I digress.
He shared a prompt on how you can ask ChatGPT to summarize any article you find on the internet. People were like WOW! Can ChatGPT really read the current internet? The answer, like many to do with Generative AI, is “sort of, but it’s not that easy.”
Let’s not get technical here. Let’s get to the lies, oh, the shame of the lies.
The prompt he gave was simple:
‘summarize: [paste the URL of the article you want to be summarized]’
Bam! It works. It really does “look” at the article and summarizes it.
Here is the link to my chat session.
Lie #1 - Who the heck is Amy Quackenboss?
Amy Quackenboss is not mentioned in the article at all. Neither is the quote ChatGPT attributed to her. A bit of sleuthing reveals that there is an Amy Quackenboss and she is relevant to this article. She is the Executive Director of the American Bankruptcy Institute, and that is referenced. Ok, she is a real person but not her or her quote are in this article. Yikes! 😢
Lie #2 - Numbers always lie!
We all know statistics can lie but this is egregious.👺 😈 ChatGPT returned this otherwise authoritative-looking nugget “ . . . here were 52,518 total business bankruptcy filings in the United States in 2021, which is up by 19% from 2020.”
Looks good. Sounds good. Except it is totally false. I truly cannot tell where this one came from. Perhaps it tried to do some math (which ChatGPT is notoriously weak at without proper priming).
The takeaways:
ChatGPT is complex technology that even its founders struggle to understand and explain at times.
Yes, it makes stuff up and does so with bravado and confidence making it truly tough to discern fact from fiction.
It might be smarter than we give it credit for. Perhaps in the article I wanted it to summarize, ChatGPT looked into linked articles, and metadata within the HTML code of the page, or it simply “read” the text not knowing it was an article and just thought I wanted a new version or something. Who knows?
Always read the responses.
If you are using known data or information (like legal case names), be especially diligent.
ChatGPT excels at reasoning, not fact-finding. Realize the difference here and you will be fine.
You can trust it, you just have to know for what. And that is why you read Brainyacts religiously! 🧠
Use Case: Summarizing Books
But wait! I literally just wrote above that ChatGPT lies when it summarizes. Yes, I did that. But no that does not mean it is still not useful in creating summaries.
Let me show you.
Because ChatGPT was trained on a massive data set, it has consumed many things that existed prior to mid-2021. Above we asked it to “read” the internet. What did it actually do? I have no idea. I mean it tried right?
But when it comes to well-known books that existed prior to mid-2021, it does one heck of a job.
So, are there books that you have read but forgotten? Or ones you haven’t gotten around to? How about those books behind you as the backdrop for your Zoom calls? Read all of them?
Here is a handy prompt sequence that helps you digest these books and extract the key concepts and takeaways.
▶︎▶︎Prompt
First, we want to find some books, if you cannot think of any. If you already know the book you want to summarize, skip to the second prompt.
▶︎▶︎PROMPT
What are the top 5 books for [client development and sales in professional services]?
It returned a great list. And yes, all of them are real books. 😉
So, let’s choose one to drill down into.
▶︎▶︎PROMPT
Please create a chapter-wise summary of The Art of Possibility.
Here is the link to the response.
It stopped after Chapter 10, so I prompted it to continue. It did but skipped the final two chapters. See, be careful. Note I did rerun the prompt again and it did complete it in one pass and correctly detailed the chapters.
Now we want to get a bit deeper into the chapters.
▶︎▶︎PROMPT
Please provide an in-depth and detailed summary of Chapter 5.
Here is a link to the response.
It appears to be a really good chapter summary.
💣💥🤯 BOOM!!
You never have to read a book cover-to-cover again. No not really. Please keep (or start) reading those books behind you. 😆
Fun & Curious AI Sites
Any Cards Against Humanity fans here? Here is a site for Cards Against ChatGPT. It covers negative stories about ChatGPT.
Looking to get paid based on all the knowledge you get from Brainyacts? Here is a job site for Prompt Engineers.
Want your own Chatbot for your own data? Here is an active directory of Custom AI Chatbots.
News you can Use:
Google might be changing the way it does search.
The days of Googling something and getting back a long list of links might end soon.
— Mashable (@mashable)
3:35 PM • Apr 6, 2023
News you can Lose:
Even though DoNotPay has been taking some heat lately, I have been a fan of its mission since its inception. Their execution of that mission? Well, questionable at times.
They just released a new bot to troll Spammers. Ok, I get it. But I don’t think this adds to civilization in any constructive way.
Today, DoNotPay is launching a new GPT-4 email extension to troll scam and marketing email/text messages by engaging them in an endless A.l. conversation. For example, here it is trolling an NFT scammer.
Here is how it works:
— Joshua Browder (@jbrowder1)
4:17 PM • Apr 5, 2023
Was this newsletter useful? Help me to improve!With your feedback, I can improve the letter. Click on a link to vote: |
That's a wrap for today. Stay thirsty & see ya next time! If you want more, be sure to follow me on Twitter and LinkedIn.
DISCLAIMER: None of this is legal advice. This newsletter is strictly educational and is not legal advice or a solicitation to buy or sell any assets or to make any legal decisions. Please be careful and do your own research.