Chances are you’ve seen at least one Harry Potter by Balenciaga video generated by artificial intelligence (and/or possibly heard of the interviews between dead people). However, beyond creating funny content and other curiosities, generative AI also offers more serious use cases. Specifically, it’s evolving into insanely useful tools available to any business.
Yet, since tools like ChatGPT are still (very) new, their practical usefulness in business may be somewhat shrouded in mystery.
That’s why in this article, we’re going to explain what generative AI is, what it isn’t, and how/if it can augment your operations.
Table of Content
Let’s start with highlighting some important context.
By the end of last year, generative AI took the world by storm. Not without reason. The developments we’re now witnessing are a massive technological leap. In other words – a breakthrough.
Because just as the agricultural revolution established society, and the industrial revolution reshaped it, generative AI has the potential to be the next step in that millenia-spanning journey. And this by challenging the distinctive feature of humankind – high intelligence.
Because just as the name suggests, generative AI is able to generate – or in other words, create. An activity previously reserved for the human brain.
And since that’s the case, let’s put it to the test.
To do that, we’re going to do something that has never been done before*. We’re going to give voice to generative AI itself – and let it explain what it does in its own words.
Here’s what ChatGPT – arguably the most prominent of the generative AI tools – has told us:
Generative AI refers to a type of artificial intelligence that is capable of generating new content, such as images, music, text, or even entire virtual environments, by learning patterns from existing data. It is often used in tasks such as image synthesis, text generation, and video prediction, among others.
Thank you, ChatGPT! Although I must say – this sounds like an incredible understatement. Because, for instance – did you know that ChatGPT (empowered by GPT-4) can transform a wireframe drawn on a napkin into a fully functioning, coded website?
At this point, you might be wondering – how does it even work?
And that’s indeed a fair question. Because nobody knows, exactly.
Of course, the general idea is well-understood. In the words of ChatGPT:
Generative AI uses deep learning neural networks to learn patterns in data. The neural network is trained on a large dataset of examples. Once trained, the network can generate new data that is similar to the training set. This is done by feeding the network some initial input, and allowing it to iteratively generate new data by applying its learned transformations to the input.
So, in plain English, generative AI is pre-trained on existing data to create something new – not just a copy – similar to the input it has previously received.
At this point, it’s important to highlight a key distinction. Just because generative AI is able to come up with something new, doesn't mean it’s in any way “smart” in itself. Or, all the more – sentient.
To give you a practical example, let’s look at the power behind ChatGPT – GPT-4.
GPT-4 is what is called a large language model. What this technically means is – it’s simply a next-word prediction engine. At its most basic level, it only predicts the next best word following the previous one.
However, getting back to the initial statement – how specifically all of that is working, we don’t know. As for now, generative AI models are seen as black boxes.
Vice gives a good example based on ice cream. If asked about your favorite flavor, you might say – vanilla or chocolate. If then asked why, you’d probably say it’s because you like the taste. But why do you like that taste? I suppose you won’t really know.
Similarly, generative AI offers output, but the exact reason why it has given a certain response remains unclear. Generative AI models are mostly assessed in terms of what gets in and what comes out. The reasoning behind certain decisions usually isn’t analyzed in depth.
As to why generative AI suddenly became so popular – now being one of the most discussed topics on the planet – has to do with its swift gain of new powers.
Of course, as we mentioned in the beginning – generative AI is able to create a lot of entertaining content. Of the quality beyond what most of us believed AI would be capable of in our lifetime. However, importantly, AI’s new capabilities also offer strong use cases in business.
And even in government-level operations.
Because for the first time in history, AI is able to competently mimic human creativity, producing content that’s highly realistic and complex. This makes the technology extremely useful in almost any industry.
Specific tools are already available for commercial use – and for anybody.
Let’s take a look at some of the most popular ones.
When describing what generative AI is, we focused on LLMs – large language models – like OpenAI’s GPT-3.5 or GPT-4.
To reiterate, LLMs are part of pre-trained transformer-based models, which are technologies that use information gathered on the web to generate textual content from websites, whitepapers, or press releases.
Other than that model, there are also the widely popular GANs – which stands for Generative Adversarial Networks. These are technologies that can create visual media from textual or imagery input. (They’re also responsible for the funny Harry Potter by Balenciaga videos).
However, these two generative AI technologies are not all there is. To name a few more, there are also variational autoencoders, autoregressive models, Boltzmann Machines, or transformers (and we don’t mean Michael Bay’s robots).
But instead of extensively discussing theory, let’s focus on the practical side.
Now, let’s take a look at specific software solutions – and their potential use cases in business.
We’ve already discussed in detail what it is and what it does, so let’s now look at how it can help you in your business.
The list of ChatGPT’s use cases is exceedingly long. In fact, I’ve asked the bot a few times to give me a list, and every time it provided me with different use cases.
The first one it gave me was this:
Later, it also expanded this list by adding positions like medical diagnosis, creative writing, knowledge management, and more.
I believe that this only shows how broad the possible use cases of ChatGPT are. They can’t be summarized in one comprehensive list. Plus, this list expands regularly. You only have to visit LinkedIn and see how people are finding new creative ways to utilize the tool for business purposes (or leisure, of course).
According to the official website, Bard is a creative collaborator aimed at supercharging your imagination, boosting productivity, and shaping new ideas.
At its core, it’s similar to ChatGPT – it’s also a language model. However, it has a slightly different objective than OpenAI’s solution. While ChatGPT is a general-purpose language model, Bard is specifically focused on enhancing Google's search engine (as a business answer to ChatGPT) and providing automated support for businesses.
Overall, however, the use cases of both solutions are not that different.
DALL-E (in its latest version stylized as DALL·E 2) is another generative AI tool from OpenAI. According to the official website, DALL·E 2 can create original, realistic images and art from a text description. It can combine concepts, attributes, and styles. From a technical perspective, Dall-E is an example of a GAN-based solution.
Overall, DALL-E's capabilities make it a valuable tool for businesses that rely on visual content for marketing, sales, and product development.
The solution’s business sibling, ChatGPT, provided me with such a list of DALL-E’s use cases:
This tool is similar to OpenAI’s DALL-E. It’s also a GAN-type solution, which means it can create unique imagery from short text descriptions.
Compared to other GAN-based tools, MidJourney produces a unique style of art. And maybe most importantly, the tool only works on Discord, the widely popular social app. This makes it extremely easy to use.
Just as in the case of DALL-E, you can rely on MidJourney for creating advertisement/blog images for your brand, producing other marketing materials, and more.
Jasper is a content creation tool marketed as a “AI copywriter”.
Beyond use cases that might be described as “the usual suspects” – meaning writing copy, ads, emails, and more (supposedly with better context than the competition) – Jasper is also able to write code, for example.
The latter means the tool can also support software developers in their everyday work.
➡ To hear more about the use cases of specific generative AI tools, book a free consultation with our Head of Technology.
As you’ve probably summarized by now, relying on generative AI tools in your work can deliver several benefits. In my eyes, they most usually revolve around efficiency.
To prove my point, I’ve asked ChatGPT to help me write this paragraph.
And this is the list of generative AI benefits for businesses the tool put together:
That’s already an impressive list. But considering that generative AI is still a young technology, new use cases are likely to emerge nearly every day.
Not even to mention that all the previously listed AI models are getting increasingly better surprisingly quickly. For instance, the differences between GPT-3 and GPT-4 are positively astonishing. And both versions were launched only months apart.
Ergo, the list of benefits is expected to expand rapidly.
Like any new technology, generative AI also comes with some challenges you should be aware of. Some are general, applying to all kinds of generative AI models, while others are related only to specific AI types.
However, all of them fall under one category: trust issues.
First, there are legal concerns.
As we’ve mentioned, generative AI trains itself on existing data. This data consists of various works (like books, pictures, articles, and more) created by people and protected as intellectual property. This poses the question – does copyright, patent, or trademark infringement apply to AI creations?
Some believe they do.
Multiple claims are already under litigation. One of the most prominent lawsuits involves Getty, an image licensing service, that sued the creators of Stable Diffusion, a generative AI model used to generate images conditioned on text descriptions. Specifically, Getty alleged the improper use of its photos – a copyright and trademark rights violation. Funnily enough, in one of the images generated by Stable Diffusion, Getty’s distorted watermark was still distinguishable.
So, from the perspective of a business owner, it’s good to know that these issues exist.
To learn more, we recommend reading this Harvard Business Review article about generative AI’s intellectual property problem.
Next, it’s important to mention a major problem related to LLM.
As discussed in the beginning, LLMs like ChatGPT are incredibly capable – but at their core, they’re only next-word prediction engines. And this can generate a certain issue – known as AI “hallucinations”.
This phenomenon might happen when the AI lacks the data to answer a query. In that scenario, when predicting the next best word in a sentence, the AI may suggest a word that is no longer factually accurate or relevant to the issue at hand. However, the AI will continue to generate subsequent words based on that initial suggestion, leading to the output of false information.
And the worst part is – the AI won’t even know it’s feeding you with made up facts.
That’s why it’s key to fact-check the output LLM provides you with.
Finally, there’s also the matter of bias.
Since generative AI mimics human activities, it can write (nearly) like a real person and so on, but at the same time, it also displays some “human” shortcomings.
For example, racism. Or sexism.
Because since generative AI is trained on data produced by people, it also inadvertently discovers patterns in human behavior that aren’t necessarily great.
As a result, LLM software has been known to generate inappropriate content. What’s more, a now inactive Twitter bot has become famous for glorifying a famous Austrian painter with questionable morals. Disturbingly, some AI-automated recruitment software tends to prefer white males over other candidates.
However, it’s important to remember that every technology comes with its challenges. In that sense, generative AI is no exception. Importantly, AI solutions get better every day. For instance, ChatPGTs sociopathic remarks were “fixed” quickly – and similar issues are now only rare occurrences.
Ergo, the technology’s current shortcomings should in no way discourage you from using it. But it’s good to be aware of them – to be able to handle them efficiently.
Thanks to the recent technological developments, generative AI is now finally able to offer reliable capabilities that we can use for leisure and business. And since the technology is still very (very) young, it’s only natural to assume that the usefulness of generative AI will only grow.
Experts point out that this growth will be happening quickly.
In the near future, generative AI is expected to advance significantly, resulting in models that produce high-quality, creative content. These models may become more interactive, enabling real-time collaborations with users.
For instance, Gartner predicts that by 2025, 30% of outbound marketing messages from large organizations will be synthetically generated, up from less than 2% in 2022. Or that by 2030, a major blockbuster film will be released with 90% of the film generated by AI (from text to video), from 0% of such in 2022.
What’s more, Gartner also suspects that generative AI will play a key role in the pharmaceutical industry, designing… drugs.
As a result of all of the above, it’s not risky to say that generative AI in business will likely become a market standard.
Generative AI is a technological phenomenon. It writes witty poems, indulges in philosophical disputes, and can even pass the US medical licensing exam. But beyond these curiosities, it’s also useful in business.
Popular generative AI tools like ChatGPT, DALL-E, and MidJourney have various professional use cases, including customer service, content creation, market research, and more. These tools automate tasks, improve accuracy, enable personalization, foster innovation, and offer scalability, thereby providing businesses with increased efficiency, competitive advantage, and cost savings.
And although generative AI also has limitations – including legal concerns related to copyright infringement or AI "hallucinations" – this doesn’t diminish its usefulness. Generative AIs use in business is expected to grow substantially in the following years (or even months).