Discover more from Sonja's Grey Matter
Learning ChatGPT Matters, and It’s Not About ChatGPT
Learning ChatGPT Matters, and it's Not About ChatGPT
A lot of people are just dabbling in ChatGPT use right now. It’s understandable - there’s a lot of work to do in the day, and introducing a new software doesn’t seem particularly important.
I believe it’s critical for people to begin to introduce ChatGPT into their workflow, and it has nothing to do directly with ChatGPT.
And y’know, I don’t throw around words like “critical” very easily. To me, such words should be reserved for software updates that threaten the safety of your computer. I don’t think there are very many things in life that are “critical” beyond food, sleep, and loved ones.
So it’s important to me that I do my best to help people understand why ChatGPT is “critical,” why I would use such language for something not everybody is even interested in right now.
As of April, 2023:
Roughly half of people in the US have only used ChatGPT once or twice.
Of those who've used it, more than half use it for less than 10 minutes per request.
There are a lot of reasons why people may not be using it. The ones I’ve seen often are:
It’s cool, but not really any faster than just doing the work myself
It makes too many mistakes
It’s theft / cheating / bad
The resulting writing is bland
Taken by themselves, those are all excellent reasons to walk away from ChatGPT and shake your head at people like me. 🤣 I totally hear you.
So, let’s zoom back and take a look at what else is happening right now.
What People ARE Using It For
People who are using LLMs (”Large Language Models,” which is a form of AI like what’s running under the hood of ChatGPT) are using it for all kinds of activities. LLMs are helping people:
Come up with new ideas when they’re stuck
Learn new skills, or understand complicated ideas
Respond to emails quickly and professionally
Improve customer service interactions
Develop new medicines
Work on climate change
… it’s even saving lives*
*I attended a training call through UW a couple months ago, where a crisis responder who works in other countries described how he’s using ChatGPT. They work with natural disasters and the like. He’s able to use ChatGPT to help quickly generate the strategies that they will use during a crisis response and develop it in a way that is culturally sensitive to the region. This shaves hours off his work. In his line of work, saving hours means many more lives saved.
If I put myself in the shoes of someone with the earlier objections, this would be kind of a scary list. This list would say to me, that people are writing bland content, incorrectly analyzing data or translating content, writing broken code, and possibly hurting people.
And it’s important to note, that interpretation is not wrong.
All of those things are happening.
And it’s not always the case.
LLMs don’t have to write bland content. There are plenty of ways to work with tools like ChatGPT to write really incredible content.
LLMs like ChatGPT are improving regularly - ChatGPT-4’s accuracy is WAY better than ChatGPT-3.5’s, and it’s only continuing to improve. Code that is written can be reviewed and edited with the help of ChatGPT to not be broken. And as far as hurting people … humans still gotta check their work, we learned that in elementary school and it’s still true. ChatGPT doesn’t take away responsibility for or ownership of content.
This is part of what I write about, ways to use AI tools successfully, so they help your job be more inspiring and fun.
But even still … if someone can still do their job just as well, and just as fast, without ChatGPT, why use it?
LLMs Are Going to Be in All the Things
Star Trek Time!
Now if you’re not a Trekky, you may have seen an episode or two, or at least bumped into a show or book at some point where the humans on the ship do two things:
Use controls on a control panel (often with few-to-no labels, wtf is that about)
Talk to the ship
The buttons and sliders on the control panel (assuming there were labels there for us to read) are generally understood to do whatever singular command they’re designed to do.
That’s how computers essentially work for us right now. You push a button, something changes from on to off, or an input (like a keyboard key) is made. You’ve said “do this one thing,” and the computer, hopefully, does that one thing.
Then there’s the other computer interaction. Someone stands in a ready room, on the bridge, or anywhere else, and just announces “Computer, what is the chemical composition of the entity in the nebula?” and a list of elements is rattled off. “What would cause it to be reacting to the ship?” And the computer will respond with multiple theories, with the most dramatic option typically uttered last.
The “computer” on the ship is used as a character, because it’s more interesting to watch than someone sitting at a desk working with a control panel. But it’s a perfect demonstration of where our tech is headed.
Microsoft Integrates AI … Everywhere
Microsoft is rolling out an AI interface called “Copilot.” In the coming weeks, using Windows will include both traditional navigation and also the ability to simply write language into an input window somewhere and get responses.
It’s more than a search. It’s … well, a copilot. You can ask it to change settings on your computer, using your own words, and it’ll (theoretically) be able to find and adjust the setting you need. It can summarize content on the sites you’re browsing, or pull up music to listen to in your music player. It can brainstorm with you, help you understand a concept, all just naturally through language.
In other words, it’s being directly integrated into the most-used desktop OS, right now.
(Apple is (presumably) working on it, but we’re not going to see anything from them just yet.)
Text and Image Editing Software Adding Generative AI
Google Docs, Microsoft Word, Notion, Canva, Photoshop, Mighty Networks - they’re all adding “generative AI” to their software.
This means, you can be in Gmail, and ask it (in plain language) to write an email draft for you. You can be in Notion, and ask it to help you with your next sentence. You can be in Photoshop and ask it to extend the background of a photo that isn’t quite large enough.
There are countless ways to use it in software, many (probably most) of which haven’t been thought up yet.
All of which is to say …
You’re Going to Have to Work to NOT Use AI
A lot of people are of the belief right now that they can choose not to use AI by just not using ChatGPT, Bing, or Midjourney.
But it’s being integrated into all major software. Not soon, but right now.
Pretty soon (within a couple years), distinguishing between “AI” and “software” will be moot. “Software” will include AI on some level more often than not.
The question then will be, do you still know how to use it effectively.
The natural language interface will make it easier to use, which is awesome. But it will also make all of your software capable of interaction and results that are fundamentally different than how you use software now.
Now Is the Time to Practice
Those complaints I listed in the beginning - it’s bland, slow, makes mistakes, it’s theft - these are all being addressed right now. They’re being addressed by the courts, by the AI tools themselves, and by the users who engage with AI regularly.
These are problems that are being solved. They will go away. Soon.
After I first played with ChatGPT, I thought, eh, okay. This is fine, but not that great.
The more I’ve used it, the more I’ve learned to find and avoid the mistakes, based on which tool I use and how I use it. I’ve learned how people can use it to “cheat,” and how people can use it effectively and ethically. I’ve learned how to make content that isn’t bland. I’ve learned where it speeds up my results, and where I need the tools to improve a bit and it’ll be more helpful later.
I understand how AI is helpful to me now - and there’s so much more still to explore.
AI is going to transform tech quickly. Whether you run a business or have a career, one of the most valuable skills you will have in the next 5 years is your ability to adapt to changing software.
Now is the time to practice.