Despite being a relatively young technology, there are already plenty of examples of generative AI making a significant difference to the way people live and work.
Generative AI, which can draw on its training data to produce content in a specific format or style, carries much potential across many sectors. While some will have first encountered generative AI tools that are text-based, from OpenAI’s ChatGPT to Anthropic's Claude, its potential use cases go far beyond this.
Generative AI is lauded for its potential to help us get work done faster, and achieve more complex outcomes than we might be able to as ‘mere’ humans. But it is also the subject of much discussion around ‘existential threat’ – the potential for AI to go off and make decisions of its own, act on those decisions, and in doing so present a threat to humanity. The debate is exercising national governments, think tanks, international organizations and others.
There’s already evidence that generative AI is driving productivity in certain sectors and interest in the technology is founded in those areas where it’s already producing impressive results.
However, the broad appeal of generative AI can also make it hard for leaders to narrow their focus on use cases that will benefit their business, as Dell CTO John Roese told ITPro in May 2024. As companies look to pick out some of the best uses for generative AI, here are five solid examples of how it’s already being deployed.
Examples of generative AI being put to good use
Text generation
Even as generative AI becomes more sophisticated, text processing continues to be its bread and butter. Large language models (LLMs) were designed to process text inputs using natural language processing (NLP) and output it.
LLMs have come a long way since the early days of GPT-3. The multimodal Gemini Pro 1.5 can draw on up to one million tokens of information per prompt – equivalent to around 700,000 words – for the most detailed text generation experience possible.
It’s not just text output that generative AI can assist users with. Microsoft Copilot and Google Vertex AI Agents, available across the companies’ respective product portfolios, can rewrite text for users to better suit their in-house style or make its tone more appropriate.
Code assistance
As generative AI excels at text generation, it can naturally sidestep into producing code in almost any programming language. Tools such as GitHub Copilot, Code Llama, and Gemini Code Assist can already produce code that conforms to a company’s specifications, with the ability to be grounded in a firm’s codebase.
For more advanced use cases, Gemini Code Assist’s large token window enables it to translate entire codebases from one language to another, particularly useful for applications written in legacy programming languages such as COBOL.
Although many developers are already using AI coding tools, researchers have warned that overreliance on them could produce poor-quality code. Vendors continue to recommend that humans are always in the loop to provide final signoff on code, even as the quality of code output becomes better and better.
Customer services
We are already used to chatbots for customer service, with their ability to answer simple queries like “How do I change my login?” or “What is the returns policy?”
However, Generative AI introduces a whole new set of use cases, and, importantly for customer-facing organizations, can answer more complex questions quickly without necessarily escalating to a human agent. AI models like LLMs can search databases of information to produce bespoke responses and have more conversational interactions with customers than earlier generations of chatbots.
This kind of AI can also take a role behind the scenes, helping human customer service agents through its ability to access and synthesize information more quickly.
Image generation
Generative AI can be used to create images from scratch, which has already become a popular avenue of model development. The technology is capable of producing images for slide decks or similar use cases.
The ethics of AI are important to consider here and there are already AI legal troubles brewing, particularly among those who worry AI could kill art as we know it. For now, image models are easy to access and enterprises are already using them to produce content such as product images.
Accessibility
A somewhat underreported but nevertheless exciting and growing area of generative AI development is accessibility.
Project Astra, the brainchild of Google DeepMind, leans on the firm’s Gemini family of models to achieve a kind of advanced computer vision. The solution is capable of constantly processing video frames, with extremely low latency, alongside speech input to quickly provide answers about the user’s environment.
Astra can cache information to answer questions about the environment it’s seen even when it’s no longer ‘looking’ at the relevant information. For example, Google showed a demo in which a user could show Astra a desk in an office, walk away and continue a conversation with it, then later receive answers on which objects were on the desk.
OpenAI’s GPT-4 has been used to power Be My Eyes, an app which normally pairs people who are blind or have low vision with volunteers who can help the user navigate life, particularly unfamiliar environments such as a travel terminal.
Using GPT-4, Be My Eyes can already answer user questions such as reading signs for them or summarizing web pages.
OpenAI has also targeted similar use cases with its own low-latency, multimodal model GPT-4o. Though limited to providing answers based on text and audio input for now, the developer says that a video-based version will be released for early testing soon.
