The meaning of Generative AI in business can be blurred by buzzwords and overwhelming marketing hype. Yet global business can use this technology to reinvent itself, and generative AI algorithms could be rocket fuel for small businesses and enterprises.
According to Salesforce data, nearly half of the US population uses generative AI solutions daily. Also, 6 out of 10 believe they are on their way to mastering this technology.
Companies are also interested in following this tech trend. The salesforce report shows that 57% of the IT experts surveyed by the company consider generative AI as a “game changer” regarding the current technological landscape.
But the term itself requires a little clarification.
What is generative AI?
Generative Artificial Intelligence is taking the news headlines and business discussions by storm. Business use cases are soaring, with a thriving startup scene built around new features and an increasing number of companies willing to adopt hot tech to support their daily operations.
Generative AI definition
Generative AI is a set of algorithms, tools, and sometimes services that enable users to generate their desired content. OpenAI puts this matter simply:
“The desired text” can apply to emails, utility texts, summaries, poems, messages, stories, or any other written content. The same goes for AI-generated images, be it a fantastical artwork, a logotype, or a schematic drawing.
For example, a generative AI can write a desired text, generate an image or a film, or provide the user with sound.
What makes generative AI special is the flexibility of the technology. The neural network is not trained to deliver a particular type of text – it is rather about gnawing through immense amounts of data to deliver the result one desires, no matter the context or previous interactions.
Why generative AI is a hot topic now
As mentioned above, generative AI is way more flexible than more narrow AI technologies. As such, the typical approach to delivering AI-based solutions was reversed. With ChatGPT, for example, it was not about finding use and then training a narrow neural network but rather about delivering a super-flexible network and then letting people play around with it.
The effects are breathtaking. According to a McKinsey report, over 60% of organizations are using generative AI, and 40% of those using it report they will increase their involvement.
Generative AI basics – how does it work
Generative AI is similar to what is commonly understood as modern, machine learning-based AI. The neural network needs to process gargantuan amounts of data to spot patterns and recognize internal relationships within a dataset. Later, armed with this knowledge, the system can deliver the desired output. Regarding the most common generative AI-based solutions, the natural language interface is the key to its popularity.
The art of prompting
Natural language-based instructions for an AI system are called “prompts.” Writing precise and desired instructions is challenging and requires experience. This fact has resulted in the rise of prompt engineering – a new trade that focuses on building and improving prompts used in AI systems.
What’s even more interesting is that the prompt can be a mixture of media; for example, one can provide the system with an image and instructions on how to modify it. Or the system can be fed numeric data and later asked to deliver an analysis. Last but not least, the system can work on code and can be accessed via API to deliver results based on input from other systems.
Types of generative AI models
Modern neural networks can deliver multiple types of content and media with varying efficiency. Considering that, the only logical and understandable approach is to divide generative AI systems into content-based types.
- Text – probably the most apparent and top-of-the-mind type of content to create. Generative AI-based solutions like ChatGPT, Google Bard, or Bing Chat can deliver text as the result of their work. These can include official texts, emails, or speeches. The publication of ChatGPT in November 2022 ignited the current popularity of GenAI technology.
- Music and sound generation – this type of generative AI aims to deliver sounds and other audio data from prompts or initial instructions delivered by the creator. These can be used to deliver a credibly-sounding speech, new pieces of music, or any other sound one can imagine. One of the early yet exciting examples of this is the Relentless Doppelganger, a neural network that delivers AI-generated death metal in an endless stream.
- Image creation – these systems aim to deliver an image from a prompt, either using a previously existing image or creating one out of the prompt itself. The system needs to see subtle differences between words and be able to deliver an image befitting the vision of the artist – or at least as fitting with the prompt as possible. The StableDiffusion model, with all its plugins and enhancements, is currently the state-of-the-art solution when it comes to generating images from texts. The model itself can be run on commodity hardware and was released as open-source software. There are also two popular online services – Midjourney and Dall-e, where proprietary models are used.
- Code – Large Language Models appear to be effective not only in generating output in natural languages but also in artificial ones, including programming languages. Depending on the system, they can generate code from scratch, suggest code during writing, or review it in search of bugs and mistakes. Copilot, released by GitHub, proves to be one of the most popular solutions, yet ChatGPT has proven to be nearly equally useful when generating code.
- Video – combining multiple sounds and multiple images can result in a video – and that’s basically the process of creating videos by generative AI – the system derives the content from a prompt.
Considering what has been listed above, generative AI is currently suitable to work with nearly any type of data and can support a variety of business processes in a company.
How to use generative AI in business
Generative AI is a powerful tool that can be used to enhance those workflows currently limited by the amount of manual work required. This may include information management, summaries, communication duties, and multiple marketing activities, among others.
Considering all that, there are at least three paths to follow when thinking about introducing generative AI in a company.
The company can build a system from scratch to deliver new features or reinvent its current approach. This requires extensive knowledge and a good deal of funds, but there are also great gains to be won. As an early adopter, the company will gather knowledge faster than the competition and will use the unique information for its benefit.
On the other hand, generative AI can be used only to enrich and streamline existing processes. This is radically different from the approach described above. The company does not need to design new procedures or build new use cases.
An interesting example of this approach can be seen in McKinsey, where the company launched Lilli – an AI-based assistant that scrapes through the company’s internal resources and supports consultants in launching new projects faster and with better outcomes.
This is about spotting an element that can be replaced with an AI-powered system and replacing it. This may include multiple steps in the countless processes across a company – for example, accounting, Human Resources, marketing, or business operations.
The key challenge is to spot and replace the key element without wreaking havoc in the company. There may be unbelievable interdependencies, so mapping the process first and forging a conscious approach is crucial in avoiding a painful failure.
A good example can be seen when tools like Todoist or Notion added generative AI-based solutions to their usual offers as plugins or additional features. The new element itself is not revolutionizing the system itself – the feature offers support to the user but is not transforming the service.
Last but not least, the company may run GenAI tools not to be used in a single use case but rather to enhance the work one does. This may include automated assistants or copilots that are not incorporated into a particular workflow but rather support the daily tasks of the employees.
This approach is seen in the launching of ChatGPT – the system was only a language model available to users to play around with, with little to no suggestions on what to use it for. It was the user base’s role to discover the tool’s use cases – and according to the Salesforce data shown in the intro – freeing user creativity was the best approach possible.
Generative AI solutions are both praised and feared. Companies like OpenAI show a vision of the future where people can be creative and innovative, transferring the heavy lifting to machines. On the other hand, intellectuals like Yuval Noah Harari show that computers’ ability to communicate like human beings is an existential threat to the whole civilization.
Similar fears have been seen during all revolutions of the past. With the introduction of the steam-powered industry, Luddists gathered in defense of their workplaces in workshops. Even the game-changing invention of writing met with resistance from the brightest minds of its era – Socrates condemned writing as a way to build forgetfulness.
Considering that this new tech will transform the current business landscape, resistance may be futile.