top of page
Search

Generative Artificial Intelligence: The Creative Side of Machine Learning

For centuries, humans were thought to have a monopoly on creativity. However, as artificial intelligence (AI) makes ever greater advances, this long-cherished belief may soon be history. While AI may not yet be creating great works of art, it can already generate astonishingly realistic content in a variety of formats. This generative AI is the focus of this month’s blog. Read on to find out more about the tech, its applications, and its benefits.


An AI Advance that Promises Great Things


Today’s artificial intelligence is getting smarter and smarter. And now that some AI is actually capable of evolving largely independently of human intervention, the tech can help us by creating new ways of performing a range of increasingly complex tasks.


One flavor of artificial intelligence that’s already delivering impressive results in various areas is generative AI. In fact, MIT Technology Review has characterized this tech as one of the most promising AI advances in recent decades.


Generative AI delivers economies of scale by generating large volumes of highly realistic synthetic data or helping humans to do so more quickly. A recent Gartner study also predicts great things for the technology – stating that, by 2025, it will account for 10% of all data produced, up from just 1% today.


What Exactly Is Generative AI and How Does It Work?


At the most basic level, generative AI denotes programs that can leverage real-world content – whether text, audio, or images – to create plausible synthetic content. It does this by detecting underlying patterns in the input. These patterns then serve as the basis for creating similar content, which should be so realistic as to be indistinguishable from human-created artifacts.

To achieve this, generative AI deploys what are known as generative adversarial networks (GANS). These consist of a generative network and a discriminative network, which work together to create new content.


Generate, Discriminate: Two Steps to Plausible Synthetic Content


As its name suggests, the generative network generates realistic new data. It does so by first learning the common properties of its source data (original content) and then creating artifacts with similar properties. In a next step, the generative network passes on this data to the discriminative network.


The role of the discriminative network is to compare and differentiate source and generated data in order to determine whether the latter meets the required plausibility standards. As you might expect, any data that strays too far from the source is rejected as fake. But any data that’s too perfect also fails the test – because it will appear unnatural, making it unfit for purpose.


The two networks are trained in alternating cycles. During this process, the generative network learns to create more realistic data while the discriminative network gets better at telling fake from source. As a result, with each iteration, the synthetic data becomes increasingly similar to the real data.


A Host of Real-World Applications...


This tech talk is all fine and good – but what, you might ask, are the practical applications of this emerging AI technology? The answer is provided by a growing number of enterprises in various sectors that are already embracing generative AI.


For simplicity’s sake, let’s consider only use cases involving image data. In healthcare, for example, generative AI can draw on real medical imaging to create synthetic images that realistically depict the future development of a disease. What’s more, the tech can be leveraged to design made-to-measure artificial limbs based on patients’ distinctive behavior and movement patterns, captured via wearable IoT devices.


...from the Cosmetics Counter to the Big Screen


E-commerce is another sector where the tech is being deployed to great effect. Here, it’s used to generate ultra-realistic images of customers and products – allowing online shoppers to “try out” cosmetics and apparel in the comfort of their homes, reducing the risk of disappointment and returns farther down the line.


Generative AI is even making major inroads into the creative sector. In the photography and film industries, for instance, it’s currently being used to enhance old images and movies by upscaling them to 4K and beyond, eliminating noise, and adding color.


Upsides and Downsides


As generative AI continues to evolve, its benefits are clear. The improvements in the quality of synthetic data produced by the tech are one obvious plus. In addition, generative AI offers a cost-effective design solution – reducing the time and effort required to come up with new products and variants while increasing the number of designs to choose from.


But like almost any cutting-edge technology, generative AI also has its disadvantages. For one thing, the sheer volume of data needed to train a solution of this kind poses a serious hurdle. And perhaps more worryingly, generative AI sometimes yields synthetic data that doesn’t match expectations – though it can be difficult to figure out just why.



Want to Learn More About AI?

Generative AI is just one flavor of artificial intelligence. If you’re interested in finding out more about AI solutions and what they have to offer for your business, feel free to contact me. And if you have your own thoughts about where and how (generative) AI can add value, please share them in the comments section.

1 view0 comments

コメント


bottom of page