The Use of Generative AI in Business

Introduction to generative AI

The best thing since sliced bread?

Or, almost anything else.

Ever since OpenAI launched the first version of ChatGPT sometime in 2022, generative AI has caught people’s imagination and is being put to an increasing set of applications, both personal and work-related.

Simply put, generative AI refers to the capability of an engine, powered by some intelligence, referred to as artificial intelligence (AI), to generate content that could be considered original by some definitions. In this context, it is similar to human intelligence. Like humans produce fresh content based on their experience, learning, context, and other variables, so does generative AI. It creates content based on the “training” it has received.

There are at least a couple of strong reasons for its growing popularity:

• It enhances efficiency. What might take a human several hours to draw or write can be created with a prompt.
• Provides creative options. A blank page is what many writers and artists struggle with, unable to find a starting point. Generative AI presents a starting as well as a completion point. They can use it for inspiration or as an initial draft to make changes to. 

Downsides and risks

Are there any downsides of using generative AI that we should be aware of?

Data processing inaccuracy

Desired output often requires the processing of data. While a human mind may understand its limitations and express an inability to produce content, AI models may be less aware and may end up producing inaccurate content. This is also referred to as “hallucination.”

Output could also be in violation of regulatory requirements, especially in areas where regulations might change often, creating operational and legal challenges.

Bias

The output is based on the training provided to the model. Biases, if any, in the training content are likely to influence the output, since the tool does not have the human capacity to instinctively identify inaccuracies and biases.

Data privacy violation

Vast amounts of data are required to train AI models. Often, this data includes personal, identifiable information, as well as sensitive information such as medical history. Loopholes in the collection, processing, and storage of this information could manifest themselves in the form of data privacy violations.

A related issue is that of the training data being collected and used without the knowledge of the owners of the personal information within it.

Intellectual Property (IP) Concerns

The generation process comes with legal risks pertaining to IP infringement. AI-generated outputs may inadvertently replicate copyrighted content, raising questions around ownership, plagiarism, and IP infringement.

An HBR article also highlights that “in many cases, it also poses legal questions that are still being resolved. For example, do copyright, patent, and trademark infringement apply to AI creations? Is it clear who owns the content that generative AI platforms create for you, or your customers?”

These issues can result in costly legal disputes if not properly managed.

Human skill degradation

As humans rely more on software such as AI to handle tasks, the degeneration of their own skills in those spheres could result.

Managing the risks

Managing risks is an ongoing process. Generative AI is no different.

Human in the loop

The responsibility eventually lies with humans, or their designated vehicles, such as the joint stock corporation.

While blind faith in the output produced by generative AI may seem difficult to resist because of the seemingly endless productivity push it can provide, the potential issues that could arise call for a sensible approach to be followed. “Eyeballing” of the output by a human is often necessary to ensure that the basic requirements of accuracy, bias, etc., are not being violated.

Relevant training for employees who use generative AI for business should also be considered.

Use proven AI models

Though this could be considered as discriminatory against upstarts and entrepreneurial providers in the space, from a user perspective, it might be a good idea to trust models built by companies that stand to lose a lot if their model is found guilty of issues such as data privacy, IP infringement, etc. This is especially true for small-scale users who may not have the experience and knowledge to evaluate providers.

Establish governance models and SOPs

It is an established good practice to create SOPs and process flows that cover at least the risk-prone parts of the business, such as generative AI. Having policies around data sourcing, storage, and access will provide clarity to employees.

You should also consider the adoption of one of several AI standards to ensure systematic evaluation and mitigation of AI risks. This is likely to include an emergency response action plan should an exposure arise.

People are the key

With its untold and yet unexplored promise, the use of generative AI is bound to rise rapidly. Responsible organizations must practice use with responsibility and ensure the message is communicated across functions and levels.

Ushankk has been in the business of providing the best AI brains to corporations, developing the models that all of us use.

As users of generative AI models, you should consider investing in people who understand the technology and can use it effectively, while avoiding many of the risks that often come with the early adoption of technology.

Ushankk can help.