We have all seen the range of content generated by AI, both useful and novel, circulating the web. There have been complex concepts broken down in the style of a Shakespearean tragedy and instructional steps communicated in the style of an Eminem rap, all produced in a matter of seconds. So now that we have taken AI for a joyride, how can we leverage this tool in an administrative and commercial capacity, and what considerations are to be made when sending a machine to do a human's job?
What is AI content and how did we get here?
While AI screengrabs have only recently started to dominate the newsfeed, AI roots back to the 1950s with early papers posing the question of whether computers can master intelligent behaviour and tasks. More recently, AI has been developed to be used across a wide range of fields for purposes such as image recognition, and natural language processing. Since 2020, a significant percentage of articles, tutorials, research papers and online content have been produced through AI content.
As of 2022, ChatGPT is the most mainstream content AI platform, developed by Open AI and an iteration of GPT-3 (Generative Pre-training Transformer). GPT is an MLM (machine learning model) that has been designed to produce human-like content based on the commands of users. Having ‘learned’ from books, articles, chat logs and even transcripts of real-life conversations, GPT is unparalleled in its handling of the human language and can adjust responses according to context and commands. Simply put? AI content feels a whole lot more human than anything we have seen before.
Unsurprisingly, ChatGPT has been embraced by millions of users. In fact, according to Chartr, ChatGPT gained 1 million users in just 5 days after it launched on December 1 of 2022. To put this in perspective, Instagram reached 1 million users in 75 days and Spotify in 150 days.
What kinds of things can AI content write?
It’s not surprising that professionals in marketing, technology, administration and so many other sectors are scrambling to put GPT to work. Although, what kinds of things can AI content produce? Some menial tasks include:
- Question and answering (similar to returning search queries)
- Providing summaries
- Gathering statistics.
Some of the more skilled tasks that AI can perform include:
- Basic scheme code
- Poems and creative writing excerpts
- News articles
- Article and essay outlines.
If you want to know what percentage of Australian households earn more than $70,000 per annum, this would be a relatively easy statistic for a human to find on ABS. If you wanted to correlate that number to the crime statistics, employment rates and other demographic statistics, the brief becomes more involved and labour-intensive for a human. Comparatively, AI can quickly read all these data points and succinctly produce data points. Writers, content creators and even students can now access data and have it tell a story without having to painstakingly find and analyse the information itself. Tasking AI with complex and dry material is a great use of the system, especially when it will be used as part of a larger piece of content.
Are there any drawbacks to AI content?
Not great at long-form… currently
The current AI generators create concise content based on commands, but there are few examples of quality long-form. This may be in part due to the commands issued and the expectation of output so users hoping to have AI produce 1000-word articles or comprehensive guides may need to rethink AI for this purpose. Instead, content creators have experimented with command capability and have found that AI can provide robust article, guide and essay outlines. From there, you can drill down paragraph by paragraph tasking AI with information related to a specific area.
Will that stop ChatGPT users from expecting a long, well-cited and comprehensive piece of content? Probably not, but learning how to brief in outlines is your ticket to long-form content.
When considering the content output, the enduring question seems to be about who owns the content and what public domain content means for writers, designers and creators. AI scans information from all over, but that means that the relevant sources are not being cited which would result in others sharing others' work without passing on the credit or even the do-follow link juice. Designers have taken this head-on in a spectacular stand against AI, creating AI art that exactly resembles Disney characters like Mickey Mouse to goad lawsuits and a discussion around copywriting against AI.
Furthermore, if the same command is entered, will it produce the same results? This could result in a number of niche service providers having the same product descriptions, guides and related content. Duplicate content can be detrimental to SEO, as search engines are unsure of which source is the original creator and who should rank for the search query. Google filters for identical content which often results in a loss of rankings.
Humans double-check references. AI is like a student, anything it reads on the internet, it will believe. AI content just regurgitates, and it not discerning as it has no insights to compare it with. It's difficult to see how simple AI is because it presents itself with polish like a scholarly article... but it’s not telling you anything new.
Risk of hate speech
When utilising AI content, businesses should be diligent in placing guardrails around its usage with detailed oversight as to what is being produced. Given the vast amount of sources that GPT pulls from, AI could be subject to using discriminatory language without the contextual nous to identify what is fact and what are long-held prejudices or assumptions that can be widespread. For example, gender roles, religion, cultural norms and areas that require nuance when writing and creating. According to Bloomberg, ChatGPT has built-in responses to discriminatory commands, but this may only catch the bolder forms of hate speech. There may be businesses that are very risk-averse or have particular brand guidelines and policies that make a human creator the more appropriate choice.
GPT-3’s machine learning is made up of 60% ‘common crawl’ datasets which are 60 million domains, with the remaining learning coming from books, sources like Wikipedia and other relevant texts. Of course, of those 60 million domains, not all have the same ethical standards of reporting as the ABC, The Wall Street Journal or even government websites. This means that our friendly AI bot has also seen the inside of forums and other less-than-reputable parts of the internet.
Are there people who are gaming the system?
There have been early adopters who have been working with early iterations of AI to make it even smarter still. Perhaps smarter is not the right word, but rather more human. Businesses have adapted AI to add variation, semantic mistakes and even a few typos and errors for good measure. All of this is designed to mimic human writing so that Google and other programs cannot easily detect the fingerprints of AI. There is speculation that AI content will bear a watermark that is undetected by the naked eye, although there is no saying that future Google Core Updates will not be able to flag this watermarked content.
What does Google think of AI content?
Critics of AI content are quick to remind us that Google has started to punish websites using AI content, but that isn’t necessarily the case. Those who are copying and pasting basic AI content that has been generated to fill a web page without providing content value that serves a use are being penalised. Although websites that utilise AI content that has been created through well-thought-out commands will not be penalised.
Ultimately, Google is championing the user and their search intent. Relevant and helpful content will always be accommodated. Those creating medium to long-form content will find that content can often be repeated at the start of a paragraph, and this may be deemed as duplicate content and may be an obvious sign of AI content. As of this time, AI content has not been watermarked as it has been rumoured to be, although all eyes are on Google to see how they will identify AI content and what that will mean for ranking and indexed pages.
Given that Google can assess which websites are designed to make income from Google Adsense and excludes them from qualifying for display advertising, it could be suggested that similar algorithms can be applied to spot ‘auto-generated content’.
So, is AI a threatening force for marketers or a tool to be embraced and leveraged?
Well, we went straight to the source.
Love it or hate, the age of AI is upon us, leaving businesses unsure of what this means for BAU, but there is certainly an opportunity for leaders to embrace this technology and put it to work under their own banner. Bolster your content with statistics, facts, insights and outlines that will produce compelling content - although ensure that it is adapted for your target audience.
Want to know more?
We have a whole team of experts who would love to talk to you.Get in touch
Want more? Here are some other blog posts you might be interested in.