The Publisher Lab: ChatGPT

In this week’s Publisher Lab episode, Tyler Bishop and Whitney Wright discuss the hottest topic in the digital publishing world right now–ChatGPT!

You can also listen to the podcast or watch it on YouTube.

ChatGPT has the record for the fastest-growing user base

While researching topics to discuss in this week’s podcast episode, almost every source had something about ChatGPT, whether it be how popular it is or how it is going to disrupt everything.

ChatGPT is now officially the fastest-growing consumer application in history–just two months after its release, it is estimated to have reached 100 million active monthly users last month, or 13 million unique visitors per day. Just to compare, it took TikTok 9 months to achieve this and Instagram 2.5 years!

This is especially interesting when you consider how viral people thought Instagram was, including Facebook. Instagram was still young with Facebook spending billions of dollars in purchasing it, making the social media giant look smart.

However popular the application may be, there are plenty of concerns about how it may change everything, especially in digital publishing

NLPs and SEO

A recap article recently published by Moz relays how some people are concerned about ‘the death of SEO’ due to ChatGPT and natural language processing (NLP) apps. Are people going to start using NLP assistants or ChatGPT clones instead of search engines?

The short answer is no, ChatGPT will not replace search engines. Search engine giants like Google are not going to simply be replaced by something that does the same thing better (like NLPs), but search engines will be replaced by something that solves the problem better. Additionally, NLPs are not very good yet at giving concrete examples and instead give broad information. This can be good for subtopics, headings, and meta descriptions, but as far as giving specific information, it still has a long way to go.

However, the future of search is certainly going to be different than it is now. The likelihood that in 10 years we’re still loading a static list of websites from a search engine like we do today is low.

All that to say, do we know exactly how this is all going to pan out? No. Trying to extrapolate out how everything is going to change with NLPs is just worrying about something that we have no control over and no way of knowing.

Search engine redesign could include AI-generated responses

Speaking of search engines, Google has reportedly started testing a new desktop search page design that integrates chat technology. This new search engine redesign would include results from Google AI and push organic search results down further. This could look similar to how snippets of an organic search result are pulled to the top of the search page to answer the query quickly, but instead of organic results, it would be AI-generated content.

This is a genuine concern, especially when you think about how Google recently purchased a ‘data pipe’ from Wikipedia; this can then have metadata added and parse the information to produce answers to queries.

When thinking about this, however, you have to remember that Google can’t just steal your content and put it into Knowledge Graph to spit out AI-generated content then. However, if they are lawfully gaining access to content and then spitting it out to answer queries, then there is nothing stopping them. The balance and relationship between publishers and Google could certainly change if this is the case.

As stated before though, AI-generated content is not at the point where it can give everything a person may be searching; it is good for giving quick answers, like ‘what is the weather’ but if you want really good information and content, you’re still going to have to go through organic search results.

If publishers have been focusing on E-A-T—Expertise, Authoritativeness, and Trustworthiness—then the cards are in your favor, because this is exactly what a bot cannot offer. Consider something like Porter’s 5 Forces; this content is specific and unique. There is not an AI right now that could replace him.

Is AI stealing your content a concern?

There currently isn’t a great way to opt out of having your content used to train large language models (LLMs) like ChatGPT. Currently, LLMs pull from sources like Wikipedia, government court records, and a popular dataset of web content called OpenWebText. OpenWebText consists of URLs found on Reddit posts that have at least three upvotes.

There is also something called Common Crawl data, which comes from a bot called CCBot that crawls the entire internet. CCBot obeys the rules of robots.txt, so it is possible to block CCBot through your robots.txt if you don’t want your website data making it into that dataset. There isn’t, however, a way to remove one’s data from an existing dataset.

We predict there will likely be a court case (or two) in the future on AI stealing content—whether it be written or art—and regurgitating it as AI-generated content which will determine how this is handled, probably at the Supreme Court level. There is currently a court case related to this about a stock image library.

These court cases will likely require AI technologies to allow investigations into how the AI is using the data it has pulled—is it taking pieces of what it has gathered to generate content or art, or is it learning from the data and then putting out ‘inspired’ work?

Another question that will need to be answered—do machines and humans have to follow the same laws, or will we need to create new laws that only pertain to machines? All of this won’t be figured out in the near future and so it’s a developing story to continue to follow.

It is likely that the way data will be absorbed and used for AI will change. You would think that NLPs like ChatGPT would ‘learn more’ by getting more data dumped into it, but that is not the case. AI is developing currently from learning from actual humans; warehouses of people are being hired to give their opinion on whether an AI-generated response answered the query accurately, what a person would expect, and how one human would relay that information to another human. This is likely the future of AI learning.

In fact, it’s already happening in some ways. Apple uses you and your iPhone to inform its satellites, and Google’s bots learn all the time through how you respond to reCAPTCHAS.

How should publishers be reacting to ChatGPT and the future of AI?

ChatGPT and AI-generated content, whether that be written or images, is likely as innovative as the iPhone was when it first came on the scene in 2007—it’s going to be disruptive, certainly, and change a lot of things, but it’s not going to flip the world upside down. Just like a cell phone made the way we access information better, so will AI.

If anything, it is a tool that publishers can use to their advantage, and there are endless opportunities. For example, you could need a chart for your article and you do not like making them or are not good at making them—you can have AI make you one by inputting what information needs to be included, and it will do it in the blink of an eye.

You can also ask ChatGPT if your information is accurate, have AI edit your content, write general content for you, and code a sidebar, ultimately eliminating multiple things from a publisher’s plate. Consider if you had writer’s block—you could have ChatGPT give you H2 suggestions for an article and even write you an outline.

ChatGPT still has a long way to go

AI still has a long way to go in regards to understanding why a person does something and being able to replicate that why in another iteration.

We’re in the phase of AI where people are imagining the craziest scenarios, for both good and bad. It is likely another innovation we will look back on and be grateful for.