A state-of-the-art language model that uses deep learning to generate human-like responses to text-based inputs.
ChatGPT is what's called a large language model, which is basically what's called a neural network. So it's designed to basically learn, so you throw a lot of stuff at it. In this case, it's basically ingesting all the text on the internet. And what it's doing is learning to replicate human-like speech. And then, as a side effect, also giving pretty accurate answers to certain questions.
A rapidly changing technology with evolving limitations.
I'm not sure what the upper limit is. I mean, we've been experimenting with it. And you know, it can generate pages and pages, there have been books that have been written with it, people are writing scripts. So you know, the sky is kind of the limit with this. And the other thing I'll say is that it is just evolving like crazy. You may have seen an announcement; there is now GPT-4. So unlike other technology, where there's kind of a gradual progression, you know, we have this thing called Moore's law, which is, you know, this the power of our transistors, the density doubles every couple of years. This is not a case of that. This is where self-learning and improvement are happening very, very quickly. So even what we talked about today will probably be irrelevant six to 12 months from now. So it's moving very, very quickly.
02:57 New tools to allow humans to focus on higher-value work
ChatGPT and other AI models are well suited for repetitive mundane work.
There's a lot of activity that happens in your CMS. This first use case that we're talking about is around text. So we have content editors that are going in there writing blog posts or updating websites. So making a lot of decisions around the content that they're creating. And what they also have to do is a lot of lower value activity, like tagging things, summarizing things, and also doing some things that they actually aren't capable of doing, like translating that content. These are all things that ChatGPT and other models like this happen to be very good at. And so what we're trying to do is bring tools to our content editors that allow them to really focus on the higher-value work and the strategy, and less of the, Hey, I gotta rewrite this headline a couple of times.
You will find more use cases for AI when it's at your fingertips.
It's really powerful when you're able to reference other pieces of text, you know, in the same blog post, right? You can say, hey, I want this title to specifically capture the essence of, you know, this part of the blog, right? So instead of having to move things back and forth in your own account, having a right there in context not only makes that a lot easier to do but also it will help you think of use cases you may be worth thinking of using AI for because it's right there at your fingertips. So it's kind of like a reminder that, hey, maybe there's a way that I can do this better using the AI.
As a limited rollout, it is installable as an app into Contentstack's CMS.
"So we're definitely in the exploratory phase that is totally usable for actual work. But right now, we're kind of doing a limited rollout just to make sure that we don't overwhelm the system and that we can make it work consistently for everybody. And so this is installable as an app into our CMS. So we've got a really great marketplace and App Framework. We are eating our own dog food; if you will, we built this as an app. So if you opt-in, you can install it into your CMS. And it will show up along any kind of text fields that you're working in, which is the main use case for the AI assistant in this first iteration."
"We do ask if you do opt-in, that we can do some feedback sessions with you. We'd love to hear what worked and what didn't work. Is there anything we can add to it? Because we do have some, you know, out of the box, sort of macros that we've put in there based on our experience, but we'd love to hear, you know, the innovative, innovative ways that you'd like to use it and make your job easier also."
"What I can do is say, well, maybe I'd like to start by having a title recommended to me. So you can see on each text field there's this little wand, which is the AI assistant. So if I open that up, I can walk you through what we have here. So you have the ability to rewrite text because you might want to try a couple of different ones. Shortening text is another really common one, and then also translating to a different language."
"And if I wanted to write a summary, this is going to be like a paragraph of text. I could also be doing an outline. It can choose the SEO tags, which is a really helpful thing because you want the most relevant bits from this article pulled out. And so you can see how we're going from maybe a small kernel, and then going up all the way into all this additional text that you would just kind of spend your time on, which maybe isn't the best thing that you could be doing in an afternoon."
"We're looking at everything in our backlog through a completely different lens now. So it's very much a different world, where the availability of this is so ubiquitous. I think people are also going to get used to having it at their disposal. So we're looking at products that we're currently building that, you know, maybe are coming out later this year, early next year. And basically rethinking, do we have the right solution? Is there anything in this product that we can help speed up? Are there any low-value activities that you have to do as part of this?"
14:03 Now is not the time to bury your head in the sand
Now is the time to start adopting AI as a tool.
"I would recommend that everyone start adopting this as a tool because the one way I think people will get left behind is if they say, like, Oh, it's a fad. Or if I play with it, then I'm gonna lose my job somehow, like, just like a calculator is in everybody's pocket. And we all use them now, like, this is just going to be the same thing. So I don't think it will replace people, but it will fundamentally change the type of work that you do because it can handle those easier things to do. So now's not the time to bury your head in the sand, right? No, now's the time to understand it."
"So we're gonna see a ton of activity here. And we're thinking as well, what is the way that we as the CMS provider can enable our content editors, and there might be certain things that in the past, you would have to go to a stock photo site, you know, it's like, I need a group of, you know, happy looking multi-ethnic people crowded around a computer pointing at the screen, right, like classic stock, image stuff, that kind of stuff. If you're writing a blog, you could probably replace it with something like this. So there isn't someone in your company that is necessarily going and taking that photo. You're already getting this from a different source. But now you can replace that with something like AI and actually get something that's a little better of a fit for the content that you're creating. So we're definitely exploring that we have an asset manager. We handle a lot of customer images. So that's definitely an area that we're exploring. And I think for certain use cases, it can make a lot of sense, if you just need sort of stock imagery."
Talk to an expert
Interested in talking to a Contentstack expert about this episode?
Matt Black, Director of Information Security at Contentstack, explores essential security practices for safeguarding your composable DXP, focusing on user management and API token strategies within an enterprise headless CMS environment.
Power productivity: Thinking composable? Think workflows
Dhaval Majithia, Product Manager
Dhaval Majithia, Product Manager at Contentstack, explores the essence of workflows for content creators to understand their importance in digital management and their role in enhancing team productivity.