The AI Elephant In the Room: “What is AI?”, and how to navigate a Pandora’s Box landscape.
- oscarggstreet
- Jun 10
- 4 min read
“What is AI?” one may ask, having heard the word hundreds of times by now.
AI chatbots, AI assistants, AI content, AI agents, AI premium features, or perhaps the controversial "AI slop" - you've heard it all, and yet you’re still not sure what the term really means, and whether or not it’s actually “intelligent”.
Whether you use ChatGPT on a regular basis, or AI simply isn't your cup of tea, you will have a more insightful and nuanced understanding of "AI" by the end of this blog.
The “LLM”
For simplicity's sake, the focus here will be on text-based AI models, and not image or video (Albeit all AI that is 'generative' will have things common amongst each other).
Going off this section’s header, you may be wondering what an “LLM” has to do with “AI”? A lot, actually.
An AI such as ChatGPT works as an 'autocompletion' machine: The AI is trained on a model known as an LLM (Large Language Model), essentially a dataset of text-based content, and the LLM strings together words by predicting what the next 'token' (A parameter from the dataset) will be.
For example, by prompting "The quick brown fox jumps over the" ... the LLM will mathematically 'guess' the next token based on probability and pattern, so the next word could either be "lazy", "fence", "moon", and so forth (With "lazy" being the most probable). Another example in this case can be checking for spelling/grammar or rephrasing a sentence.
It is because of statistical probability combined with the large datasets requiring great amounts of computation that an LLM works so efficiently to generate content with a single prompt, and it is because of the speed and quantity that makes 'generative AI' a very accelerative tool.
The AI Cause
We refer to an LLM as an “AI” because it appears to have intelligence, and having an artificial intelligence in your pocket is a very appealing endeavour in an age where you can find a solution to your problem within seconds.
LLMs have been around for a long time, and they have now skyrocketed greatly thanks to the advent of 'generative AI' with platforms such as OpenAI's ChatGPT and Google's Gemini, alongside the development of image and video AI models.
Not only does generative AI possess great speed and quantity, but it is also widely accessible and free to use, even without the requirement of a subscription. A user may prompt ChatGPT to generate content such as general advice, copywriting, ideas for social media posts, business documents, and so much more that is limited only by the imagination and what the AI can do.
The landscape has now gone from businesses piloting the integration of AI tools into daily work, to massive amounts of AI-generated content that can be distributed online.
As you can imagine, this is a lot of output. A lot.
The AI Effect
Being realistic here, this is a Pandora’s Box situation.
AI-generated content can be produced in massive quantities, with perfect grammar and structuring. Remember, however, that this is still an autocompletion machine working off predictability, and not necessarily an intelligence that can reason and discern which information is 'correct'.
This dictation based on pattern can result in 'hallucinations', which refers to generated content that has inaccurate/outdated content due to the LLM’s probability patterns. Without human verification, this mass generation of AI content can risk in becoming homogenised 'spam' as it collapses the value of said content - hence the coinage of the term "AI slop".
This does not mean all AI technology and content is inherently bad, and this blog is not aiming to explore further controversy and debate, but rather provide a nuanced, present-minded and knowledgeable sense of awareness to the AI era of the internet.
What this does mean, however, is that AI is not going away. So, how do we handle it?
The Director’s Chair
In any large-scale project, having a director is vital for coordination, steering towards strategic goals and, well, directing. A good director will get involved and ensure their vision is aligned and achieved, and a bad director sits back and does nothing.
It is the same with integrating generative AI, and why you must become the ‘AI director’ in this instance.
Learning how to engineer an “AI prompt” is not a job in itself. However, the way you tailor a prompt to determine context, instruction and technicality will dramatically improve the output of an AI-generated response. Generally, you want to go to your respective chatbot’s settings and load it with custom instructions so that you can enforce that same context for every single time you send a prompt.
For brands, tailoring your AI does wonders for maintaining brand voice and consistency, as it reduces the time spent amending the output for, say, copywriting your company’s next social media post.
And, for those who still wish to remain observant, you will at least begin to recognise where AI output has been utilised in online content.
Some additional pointers:
Treat the AI is a collaborator rather than a content factory and ensure that human curation and quality assurance is always at the forefront.
Research how to write efficient prompts, as well as getting the most out of a single conversation thread with the AI.
Identifying ‘AI language’ and the tone of voice often found in AI-generated outputs to avoid homogenisation.
Does all of this seem overwhelming? That’s where Mothership can help.
Mothership’s Response
At Mothership Branding, we take a refined and nuanced approach to emerging technologies such as generative AI, understanding both the benefits and pitfalls of these tools and the easy/best means to integrate them into creative and digital workflows.
If you're wanting to learn how to navigate the AI landscape and get ahead of the curve, please get in touch with us for consultancy and coaching with assured peace of mind and time-saving advice.
Comments