What is ChatGPT and how is it different from AI writing assistants?

It’s easy to understand the hype behind ChatGPT, the new artificial intelligence writing assistant on the block by OpenAI.

ChatGPT is essentially a chatbot where users can ask it any question, and receive an answer in a few seconds. While this is hardly revolutionary, where it differs from the chatbots that most are used to, is that the responses nearly seem like a human was behind the screen.

But a human with Google for a brain.

This brings about a wide spread use of applications:

• Answering questions you would normally type into a search engine
• Write code
• Answer math equations
• Generate copy on nearly any topic
• Write a rap about the superiority of Electric Vehicles in the style of Ice Cube

Some are saying that it won’t be long until ChatGPT replaces Google entirely.

Google themselves are worried about what this technology might mean for them – the number 1 search engine in the world. Paul Buchheit (the guy behind Gmail) thinks the writing might already be on the wall.

What is the difference between ChatGPT, OpenAI, and other AI writing assistants?

OpenAI is an artificial intelligence research and development company, and the creators behind GPT-3. GPT-3 is the NLP (Natural Language Programming) model that has been used to train ChatGPT, as well as many other AI writing assistants – including Mark Copy AI – making Open AI a partner.

Even though this is an incredible base, there is still work to be done to transform this technology into something useable.

AI writing assistants have more specific use cases

Taking these learning models, they are often further trained and refined for specific customer use cases – creating writing styles. For example: a real estate listing is likely to contain very different language than an Instagram Post. Creating these separate workflows means that the outputs from the AI are more detailed, specific, and most important – useable.

AI writing assistants use more than one model

This means that outputs are able to pull from many sources, not just one. Leading to a more rounded and less dependant systems, capable of producing more nuanced language.

To gain even more control, we train our own models.

Most companies leveraging this technology, will use this as a base.

GPT-4 might be the best baguette in the world, but to make it the best sandwich in the world you need to add your own special ingredients 🥖

ChatGPT gives you one answer

Think of the current search engine model – ask a question, get seemingly endless results. The ones most applicable or the ones that have been found to be the most useful, rightfully filter to the top. All is right with the world.

Now consider the chatbot model – ask a question, get one answer. This works great for things that are reasonably linear (think math applications, how do I do x, etc).

The danger here is that even though AI has comes in leaps and bounds, when pulling in information it can still make huge errors. This is because it is still unable to actually think.

Because of this it is not entirely able to discern what is the right answer, what is bias, arbitrary, propaganda, or even hate speech.

When opinions are delivered as facts, we enter into dangerous territory.

In contrast, AI writing assistants will deliver multiple opinions for you, the subject matter expert, to use as a base. Adding your own opinion, tone, etc to make the content truly yours.

Use cases for ChatGPT

We’ve covered some of the use cases for ChatGPT, but let’s dive a little deeper into the best of them:

Chatbots

Starting with the obvious, as the name suggests, ChatGPT is intended to be used as a chatbot. This is where it is truly revolutionary. Its ability to take a prompt and generate an output near to human language is unlike any chatbot out there – giving you answers instantly.

General knowledge

Where it truly comes into its own is answering general knowledge questions – the kind of question usually destined for a search engine. And just like a search engine, you are more likely to get good results when your question in linear, leaves little room for interpretation, and there is a right answer to be had.

TechCrunch had a great example: explain all of the different Pokemon type strengths and weaknesses. And boom – all the answers 12-year-old me would normally be racking his brain for, and 30-year-old me wishes Google could spit out immediately.

Let’s compare this to the long list that Google returns, that has the answer somewhere in one of those articles:

Use cases for AI writing assistants

In comparison, AI writing assistants use this technology to be very specific in what they do, and come into their own when looking at more specific use cases. Let’s throw out a few common examples.

Marketing departments

AI writing assistants have been working hand in hand with marketing, content, and growth teams ever since spell checkers sent dictionaries spiralling into near oblivion.

And whilst that relationship has only gotten stronger over the years despite some scepticism about AI, it’s important that understand what the future looks like for AI and marketeers.

Especially when thinking about what is most important to CMO’s everywhere: delivering personalisation at scale, content driven marketing strategies, speeding up time to market.

With this in mind, creating content in the right tone of voice, for the right audience, at scale; becomes vital for driving a demand generation strategy.

Think of how the following could be optimised:

• Blog articles
• Email campaigns
• Social media content
• Product descriptions
• Ad copy

All areas where AI writing assistants, trained for these use cases can help.

The use cases listed for marketing teams are only multiplied for marketing agencies who are tasked.

SEO

The importance of creating content optimised for SEO can not be overstated. Marketeers want to give their audience what they want with the least amount of friction. It will come as a surprise to no one that the first three 3 results get the vast majority of clicks.

Some AI writing assistants have SEO tools built into their interface to help out with this piece of the puzzle.

There’s no point creating great content if your audience is unable to find it.

AI detected content

Google is out to punish bad content.

It doesn’t matter if it is created by AI Natural Language Programming models based on hundreds of billions of parameters and decades of training, or Barry from accounts – the release of Google’s Helpful Content update sends a clear message: bad content will feel Google’s wrath.

Good. It should. Bye.

With this new update, Google will be able to “automatically identify content that seems to have little value, low-added value or is otherwise not particularly helpful to those doing searches.”, punishing those in breach of these content guidelines. 

Unfortunately, ChatGPT is easy to flag. Using even more AI.

The irony gods had a field day cooking up this one I’m sure.

Originality and plagiarism are cut from the same cloth, which is why at Mark we are taking this update incredible seriously – integrating originality and plagiarism features directly into our platforms to avoid any possible punishment from Google in the SERP ranking.

Guaranteeing 100% original, safe content.

But it’s important to mention the purpose of AI writing assistants, is to assist. It’s up to the user to personalise the inputs, change the syntax, and add your expertise and opinion to make the content truly your own.

AI is there to help with the heavy lifting – not replace.

Data privacy & security

Coexisting with AI

Forrester said it best in their latest report:

“Ten percent of Fortune 500 enterprises will generate content with AI tools. Human-produced content creation will never be fast enough to address the need for personalized content at scale, and in the next year, we expect to see at least 10% of companies invest in AI-supported digital content creation. Models for text, image, audio, and video creation will become an intrinsic part of daily workflows in both existing tools and new applications.”

And this is for enterprise companies – more agile agencies and start-ups have already started leveraging AI as a competitive edge over their competition.

With the advance of AI, human and machine are likely to have a much closer relationship in the next few decades.

But AI cannot replace a subject matter expert – nor should it.

AI’s role is, and always will be, an assistant to the subject matter experts out there.

Creating ideas, outlines, content and sentence structure, spelling, grammar; AI does a great job of generating and correcting content in seconds.

But think of humour, nuance, emotion, objectivity, subjectivity, love, anger; there’s a long list of human characteristics that AI can attempt, but is as yet unable to perfect.

This is why AI will always need a human touch.

AI will improve – it’s getting better and more intelligent everyday. However, we are long long way from AI completely taking over human creativity.