AI: Tool or Replacement?

You’ve seen this movie; it ends with someone having to go back in time to save the world. Ok, so we’re not quite there yet, in fact, we’ll probably never get there at all. We’ve all heard about the wonders and the possibilities of artificial intelligence (AI). But what are the real-world implications of the technology that is being called AI?

Forgetting any pop culture interpretations and philosophical ramifications inherent in the idea of artificial intelligence, we thought it would be pertinent to have a reality check about this technology that feels like it’s poised to change the way we work and, in some cases, already has.

Is it a replacement or a tool, a boon or a bane? What does AI mean, and how will it help shape working life in the future?

What do we mean when we say AI?

The first thing to get out of the way is that the large language models (LLMs), which are called AI, are not artificial intelligence. What we mean by this is that calling the technology AI is a little bit of sneaky PR.

The awareness, if it is even right to call it that, that programs like ChatGPT and Jasper seem to express is due to something called ‘machine learning’.

Using generative adversarial networks (GANs) and Variational Autoencoders (VAEs), machine learning models are trained on specific sets of data, massive sets of data, to identify patterns and structures within that data.

They are innovative but do not possess ‘intelligence’ in the general definition of the word. They cannot think for themselves nor generate anything truly original.

LLMs and generative algorithms give the appearance of genuine creation. They take information from the internet, be it images, code or text and generate new information, creating a desired output based on human input.

Anyone who has used one of these programs will share the mixture of awe and disappointment as the limitations of the software become apparent very quickly.

What are the current limits of AI?

It can be tempting to look at the revolution in machine learning as a way to shortcut a lot of work. Like with any new innovation, how it’s used by people determines its effectiveness.

The truth of the matter is, people are lazy, most people anyway, and if a tool is created that makes some jobs seem obsolete, especially if it will save money, companies are going to jump all over it, no matter how effective it actually is.

LLMs work by recognising linguistic patterns but they cannot infuse their output with any meaningful beliefs, desires, intentions, or emotions because all the algorithm is doing is looking for pre-existing patterns and then extrapolating information from them.

The apparent patterns we perceive in human behaviour and the world are the key to many people misunderstanding what AI is and what it can do.

It’s only natural, but we tend to think of the world as a series of defined patterns that interact with each other and then create predetermined outcomes. How someone acted yesterday will be how they act tomorrow, right?

In reality, it’s not so simple.

Although there are many things that can be predicted with a degree of accuracy, it is still only guesswork, no matter how sure it may seem. People need patterns; it’s the reason we are so successful as a species, but it is easy to forget that even the most familiar events can sometimes go astray. Patterns we thought of as certain are not, and there is far more randomness in life than we are comfortable admitting, so it goes with the current state of generative AI models and LLMs.

We have a habit of projecting onto them an awareness that isn’t there.

All AI does is look for patterns; there is no understanding or intent behind their output. Because of this appearance of intelligence, many have jumped the gun and gone all-in on AI before it is truly able to deliver what it promises.

Misuse of AI and overpromising on what it can actually achieve has led to a time of uncertainty in many creative fields, including marketing.

When the AI revolution came, we all thought it would remove toil from our lives; what it looks like it’s poised to do instead is devalue and replace the creative arts, not exactly an exciting prospect.

There is also the issue of legality. Generative AI models are trained using writers’ and artists’ output without permission and in some cases, have generated copyright-infringing material. It can be tempting to look at the current crop of AI programs and feel like obsoletion is just around the corner, but this is far from the case.

If used as a tool to enhance and improve workflow processes, the potential for AI is incredible. Rather than look at it as a replacement, we should view it as another tool, one that will only improve with time.

What can you use AI for to get consistent results?

So, what can you actually do with the various AI tools out there?

Let’s take a look at some of the more useful tools, how they work and what they do.

ChatGPT – To help get you out of a rut in terms of creating marketing copy or carrying out market research, this app, which everyone has probably heard of, is an industry leader when it comes to LLMs.

The chatbot is best used as a means to explore different avenues of thought and ways of formatting copy for sales emails and marketing content.

Grammarly – When it comes to checking your work, having as many tools at your disposal as possible is advantageous. The more a piece of content, sales copy or even creative writing is checked for errors, the better it will be.

Utilise programs like Grammarly to quickly see any issues or errors that have crept into your writing. The program features suggestions for clarity, engagement, and so on.

Like with all AI assisting tools, the results can sometimes be a bit off, so use your own judgement with the corrections it suggests.

DALL-E 3 – The task of writing AI image prompts is becoming a skill in itself. This program was developed by the same people that made ChatGPT and is their image generation software.

The results are impressive, although the images that are generated depend entirely on the quality of the prompt. And it still only gets fingers right about half of the time!

Good for inspiring compositions and depending on the artist, can be used to generate the backgrounds for images.

Fireflies – If you need to take notes during a meeting, this app, which integrates with Zoom, Teams and others, records the video and audio of a meeting and then produces transcriptions automatically.

A good example of AI being used to improve efficiency and productivity, helpfully providing notes for record keeping and later use. You could even combine the output generated from this program and then create a voiceover using an AI text-to-speech software if you desire.

Clockwise – A scheduling assistant used to optimise meetings for an entire team. One of a number of productivity-enhancing AI programs that give you the opportunity to automate your team’s calendars, optimising their schedules any way you prefer to work.

Helps to avoid conflicts between your team when organising meetings and has many customisation options to get it working in the way you want.

Murf – If you need to create a voiceover for a presentation, training video, or text-to-speech, this program has an impressive array of voice options to choose from to turn your script into a voiceover in minutes.

Like with any AI tool, your results will vary depending on your input and even though the results are often impressive, even professional sounding, you may regularly hear mispronounced words and inflection in the incorrect places.

These voiceovers can’t match a human voiceover in terms of quality just yet. Depending on your aims, this may or may not be a problem. The ability to choose from various voices allows you to tweak the tone to your liking and if you need to create voiceover in bulk, then this may be an ideal option.

Jasper – Like many of these AI tools, they empower those who do not specialise in those fields to output material on par with that of an experienced individual. This is no more apparent than with the slew of content generation tools available using LLMs and generative algorithms.

Unlike ChatGPT, programs like Jasper have been created to try to capture the intent behind a specific audience. Use for short-form content like emails, product descriptions, and snappy headlines; when it comes to anything with substance or length, the choice is up to you; results for long-form content are often filled with errors and require extensive re-writing.

If you need help creating content, if it’s not your forte, then tools like this will help you to capture your audience’s voice, and the provided templates give you an excellent starting point for content generation.

When using any content generation software, always proof the content it generates and fact-check where necessary.

AlphaCode – Developed by DeepMind, this software is an AI system designed to write high-quality code for programmers. Every machine learning program excels at combing through vast amounts of data, specifically code, to learn and generate content that is already optimised.

This AI is best used as a way to aid programmers in coding rather than creating code wholesale.

New versions of these systems are being created all the time. New and improved iterations are regularly released that further hone their abilities.

It’s not all doom and gloom either, like with almost every invention in the history of the human race, it’s not the thing itself that has some inherently undesirable attributes; it is the way it’s used by people that determines its effect on work and society at large.

No matter which tool you opt to use, it will not fully automate the processes of your work; there will always be a need for further refinement and guidance from a human hand, at least for now.

Embracing AI and utilising it as a way to improve productivity, creativity, and efficiency in human-created output is the only way forward.

Rejecting AI outright isn’t an option and fearing change is always pointless. Nothing beats human judgement (until they come up with an AI that mimics that function as well), and the initial rush of enthusiasm for AI LLMs has waned since its heyday as the reality of what these tools are capable of became apparent.

Using any of these programs can improve efficiency, and as with any tool, it’s only as useful as its user.