Issue #8: GPT-4 news, AI in Google Docs, The Future of Work, and Instagram Prompts
And here it is, GPT-4.
If you have 25 minutes, I urge you to watch the OpenAI GPT-4 Developer Livestream here. It’s well worth your time.
(Right now, GPT-4 is rolling out to ChatGPT Plus users. It’s the best $20/month I’ve ever spent and I’d pay 100x more, easily).
I’ve been immersed in GPT-4 for hours, running through a few hundred of my prompts (last count I had 1,854-ish different prompts, not counting variants).
And making up new ones on the fly.
They all work, of course.
But my prompts are getting much better quality outputs than with GPT-3 and ChatGPT (which mostly ran on GPT-3.5).
So, I’m preparing a special deep-dive issue that goes out next week, with more details, breakdowns, and examples for you all.
You won’t miss a thing (if you’re subscribed).
(And don’t stress about keeping up with the torrential downpour of AI news coming at you every day, I’ve got you).
For now, here’s the GPT-4 TL;DR:
🤖 More creativity with superior reasoning, problem solving, broader general knowledge, association, and quality of output. It’s astonishing what it can do and the quality upgrade from GPT-3 and ChatGPT.
🤖 It actually passes many standardized exams (BAR, LSAT, etc.) This is huge because of what it implies (reasoning abilities, probably close to displaying a level of human intelligence in some instances and under the right circumstances).
🤖 Multimodal: API accepts images as inputs to generate captions and analyses. It can “look” at images and describe them in detail.
🤖 Huge context upgrade: You can include the text of full documents within a single prompt. Via API, it can go as high as 50-ish pages worth of text. And your outputs can be longer, too.
🤖 It can do drug discovery (find similar compounds, modify them, etc.) This means it can do the same for supplements and even for other product types altogether (product development will go up 10x).
🤖 Write working code for a Discord bot (and figure out what’s wrong with error messages with little to no direction, including debugging and reading documentation).
🤖 Create website code from a pencil mockup made in a physical notebook(!) Seeing this in the demo was… goosebumps.
🤖 Capability to organize huge knowledge bases, helping you “interact” with it all, instantly. (Is search and SEO changed forever? Yes. Is Wikipedia a thing of the past? Yes).
There’s so much more I could add.
And I will (and go deeper), when the special GPT-4 issue comes out next week.
What happens now?
Adding GPT-4 to the mix is turning a raging 5-alarm fire into a nuclear bomb of possibilities.
And on top of all that: Google just launched Generative AI across ALL of Google Workspace (Gmail, Docs, Sheets, Slides, Images).
Here are just a few of the many things you will be able to do with their AI (which is not any of the GPT models from OpenAI, it’s their own):
Draft, reply, summarize, and prioritize your Gmail.
Brainstorm, proofread, write, and rewrite in Docs.
Bring your creative vision to life with auto-generated images, audio, and video in Slides.
Go from raw data to insights and analysis via auto completion, formula generation, and contextual categorization in Sheets.
Generate new backgrounds and capture notes in Meet.
Enable workflows for getting things done in Chat.
That's only scratching the surface.
And there’s one more thing!
Microsoft is hosting a “Reinventing Productivity: The Future of Work” event tomorrow (March 16th).
In their words: “Learn how AI will usher in a new way of working for every person and organization.”
My bet is it’ll be AI integration across their whole Office suite (and everywhere).
Are you passing out yet from the overwhelm?
Don’t. Just read my emails. I’ll make sure you get the goods without spending days or weeks drinking from the firehose.
In today’s issue:
Mini-Tutorial: ChatGPT Prompting for Instagram.
20 ChatGPT prompts to transform your SEO.
Boost your CTR with AI-animated visual assets.
LinkedIn introduces collaborative articles with “AI-powered conversation starters”.
Are search engines and SEO becoming obsolete? Is SEO dead?
Replace hours scouring docs, and access valuable information with simple questions.
Let’s dive in.
Lots of people are using various AI/ML tools for text.
But images via Midjourney or Stable Diffusion seems to be harder.
Mini-Tutorial: ChatGPT Prompting for Instagram
This isn’t an email about using ChatGPT for Instagram captions.
Nine times out of ten, just asking ChatGPT to “write 10 Instagram captions describing X. Include hashtags” will get you where you need to go.
Short-form writing like Instagram captions are right in ChatGPT’s sweet spot.
So, what’s the real generative AI differentiator for Instagram?
The Basics of Image Prompting
Generative AI image tools like DALL·E can create images with a basic prompt.
Just put your text into a box and you’ll get an instant image.
But here’s the problem:
Without careful control over your prompts, you have little influence over the image you get.
AI image prompt crafting is the art of refining your prompts to suit your style and match your intent.
Image Prompt Essentials
Here are basics of image prompting:
Be as detailed as possible, using adjectives and style modifiers.
Add emphasis with repetition and weighted terms.
Refine with negative prompts.
ChatGPT helps you infuse your prompts with detail while keeping them short enough to be processed. (DALL·E, for example, has a limit of 400 characters.)
Steps 2 & 3 need to be done manually—but if you’re happy with the output after step #1, you don’t need to worry about them. You’re good to go.
How to Construct a Prompt
When you’re creating a prompt for image generation, imagine you’re “zooming out.”
Start with what’s most essential to the output you want. Then, layer on details.
An oil painting of a cat dressed in a chef’s outfit holding a knife and cutting an onion, standing in a gleaming shining industrial kitchen, in the style of Norman Rockwell, painted in vivid detail.
Here we have:
[medium and main subject] an oil painting of a cat
[details] dressed in a chef’s outfit holding a knife and cutting an onion
[setting] standing in a gleaming shining industrial kitchen
[style] in the style of norman rockwell, painted in vivid detail
I’m happy with the output of this prompt, but it’s only 191 characters—which means I may be missing opportunities to add details that improve my output.
That’s where ChatGPT shines.
Using ChatGPT for Image Prompts
Here are three ways to use ChatGPT for image prompts.
1. Brainstorm: When you start prompting, you may not know exactly the vibe you’re going for. ChatGPT can help you find different paths to go down.
Image Prompt #1:
Write a list of five detailed prompts about the idea written after CONCEPT. Use as much detail as possible. Give a description of the scene, modifiers divided by commas to alter the mood, style, lighting.
Once you find one you’re happy with, it’s time to expand your description.
2. Add details: Just by adding detail, ChatGPT can make your image prompts much more formidable.
Here’s a simple prompt:
Image Prompt #2
Expand the below prompt, adding detail, adjectives, and style modifiers. Shorten output to 400 characters.
[Describe your scene here]
3. Descriptions + Formatting: For a more consistent output, you can give ChatGPT instructions to become a “DALL·E Prompt Generator” and follow a specific framework:
Image Prompt #3
You are a DALL-E prompt generator, a bot that generates prompts for AI image generation. Expand the below prompt, adding detail, adjectives, and style modifiers. Output the prompt on a single line in the following order: type of image, main subject, subject details, setting details, style. Shorten output to 400 characters. Understand?
Once ChatGPT replies that it understands, input a brief description of the scene you would like to create and let ChatGPT expand on it.
Refining Your Images
Generative images are about experimentation—even more so than generative text is.
If you’re not happy with your output, keep trying different descriptions and style modifiers.
You can also emphasize certain parts of your image with weighted terms (like cat:10) or repetition (like saying “cat cat cat” instead of “cat”).
You can also deemphasize other parts of your image with negatively weighted terms (like onion:-5).
Add The Power of Generative AI to Your Instagram
Instagram is the king of visual platforms.
Even if you only use generative AI to add unique filters to the images you post, you’ll stand out.
But that’s just the start. Once you master image prompting, you can create entirely new, dazzling content that’ll get people talking about you and your brand.
And ChatGPT can get you there faster.
A few more how-to’s for you:
ChatGPT prompts for SEO.
Being able to identify crucial keywords for search engine ranking, optimize the content for seamless integration and balance metadata is time-consuming and often confusing.
Luckily, here’s the perfect guide on leveraging ChatGPT to harness SEO.
The guide contains 20 prompts for AI-powered content.
Whatever your current SEO skills are like, ChatGPT can help you:
Identify the most popular questions to target
Classify keywords based on search intent
Generate relevant FAQs to flesh out any article
Provide facts and data to back up your content.
Defeat blank page syndrome, and get ready to smash your CTR.
Keep up the engagement hype and spruce up your social media presence, landing page or email signature with your very own AI-generated animation with this Animated AI Assets tutorial, using the power of Midjourney AI and Rive.
The tutorial outlines how to create a stylized, low-poly asset that can be exported as a GIF for a lightweight enhancement for any digital platform.
The best part? It takes less than half an hour.
Even if you don’t consider yourself artistically inclined, the short course is straightforward, making the most of what AI can do.
Who knows—this could be the beginning of your animation career if AI takes your job.
The total vertical integration of AI tools across software continues.
LinkedIn has introduced a feature called collaborative articles, which uses “AI-powered conversation starters” to start discussions between “experts” that use the platform
Good business communication is an art form, but the median business-related communication, especially at the start of a discussion, is pretty crappy.
LinkedIn is not the only platform doing this.
Snap has created its own chatbot, and Meta is working on AI ‘personas.’
It seems future social networks will be increasingly run by AI’s of different kinds.
When a tool as capable as ChatGPT becomes available worldwide, the diversity of the global population means it will be used for good and for bad.
According to Forbes, this year alone, over 50% of organizations are planning on incorporating AI to automate tasks and speed up production.
If things keep moving the way they are, we can expect AI to be pretty much ubiquitous within the next five years.
Does this mean search engines will become obsolete?
Let’s take a look.
Google Ads account for around 80% of Alphabet’s revenue, the digital directory’s parent company.
And what does the creator of Gmail have to “tweet” about it:
If search engine’s decline as rapidly as AI grows, then yes:
Most plausibly, they’ll be replaced with AEO, “Answer Engine Optimization.”
But before we get ahead of ourselves, let’s go back to a time before the internet.
Remember buying specific maps to navigate to a relative’s house out in the sticks, or flicking through the Yellow Pages to reserve a table at a local restaurant? Maybe you recall visiting the local library to dig up a specific homework answer.
That leads us to the one crucial question:
Is the rise of AI going to do what Google did to physical directories?
The answer is not quite. Maps still exist, libraries are still popular, and the Yellow Pages is now a very successful digital marketing directory.
They didn’t vanish. They evolved.
And Google is no neanderthal.
Progress across generative tech is about to be exacerbated by the war raging between two of the world’s largest companies.
Whatever the outcome, it’ll be fascinating to see exactly what Bard can do, and if it can extinguish those “Google-killer” whispers.
We’re in for a pretty crazy 2023.
$50M: Revolutionizing the customer experience is Ushur. The recent investment round will fund AI-powered workflows for claims processing, customer support, and appointment scheduling.
$25M: Inscribe is bringing automation to cybersecurity: this startup is using AI to safeguard against fraud. Accurate and efficient, the app is designed to protect organizations from dodgy documents and phishing schemes.
$23M: Natural language-driven computer programming assistant Magic is making coding accessible. The tool allows developers to generate code from plain English.
Got a business trip booked?
Trying to determine the best work spots, restaurants and possible day trips for your stay? Can’t figure out which ones are best for the team, and which just aren’t worth it?
🤖 Traveling just got a whole lot easier with Travel Advisor.
The app is impressively easy to use and instantly provides specific, but succinct, specific travel itineraries for anywhere in the world, for any duration.
No more endless hours spent browsing websites and planning itineraries—all the information you need is all in one place.
Bon Voyage! (Imagine doing this with your geofencing marketing, hmmm…)
🤖 On another route, discover the power of instant knowledge with Humata.ai.
Replace hours scouring docs, and access valuable information with simple questions.
Humata responds with answers about any file in real-time.
The tool can be used for just about any file you can think of:
Terms and conditions
The AI is one of many that instantly frees up your time to focus on other things that need your attention.
This is a game-changer for anyone in need of fast, reliable information.
Are you creating videos for YouTube?
YouTube will soon incorporate new generative AI features into its video-sharing platform.
Good times ahead.
THE LAST BYTE
I normally don’t read Seth Godin.
But someone shared this post with me, The coming ubiquity.
And I agree: AI will soon be everywhere, in everything, all at once.
AI and Machine Learning tools can do the job of any creative task, like write a book, craft copy, come up with business ideas, design products, create graphics, write songs and a lot more.
So when all this is happening (both now and in the future), what’s one thing that will both help you—and be one of the few things left for a human to do?
Exercising something called discernment.
It’s the ability to judge well. It’s the process of making careful distinctions in our thinking about truth, right and wrong.
It’s your ability to see clearly what’s in front of you. And decide. Judge. Evaluate.
Without a strong artistic taste, a head for understanding engagement, and a passion for your craft, the AI-generated content you produce will sorely lack creativity and consequently, success.
You can see the need for discernment playing out already in art:
The work (one of several fan recreations) was made using artificial intelligence.
Should it be thought of as art at all?
Expect more of this. With everything. Everywhere. All at once.
See you next week,