Gen AI is Great, But is it the Right AI for Your Business?

2023 was the year executives started paying more attention to AI. Many are already thinking about how different AI technologies could redefine business processes. While generative AI (Gen AI) has garnered the most media attention, is it the best choice for your organization, or could other types of AI be a more worthwhile investment?

Gen AI is still in the teething phase

According to the Gartner Hype Cycle for Emerging Technologies, 45% of executives reported that their organization was piloting or experimenting with Gen AI in the second half of 2023, and another 10% had already gone live with solutions. The key words here are experimenting and piloting. Many Gen AI projects are yet to generate real business value and it doesn’t look like 2024 will see much change. According to Gartner, Gen AI has already reached the peak of inflated expectations and we will now enter a two-to-five year “trough of disillusionment.”

As the hype fades and more Gen AI projects are realized, actors in both the private and public sectors will continue to work together in the background to iron out all the issues we saw headline last year: hallucinations, copyright infringements, data privacy, etc. In the meantime, more established AI technologies will gain momentum.

Supervised learning will bring the most business value

In an interview with VentureBeat, AI expert Andrew Ng compared the hype around GenAI to the rise of deep learning five years ago. While it was clear that deep learning would transform different industries, nobody really knew which use cases it would be applicable to transform. We’re in a similar position now with Gen AI.

With ML technologies like supervised learning, however, we’ve already surpassed this hurdle and the use cases for businesses are much clearer. Out of all the AI technologies, it generates the vast majority of financial value today and Ng predicts it will grow much faster than Gen AI in the short term. In his visualization of the projected market share of different types of AI over the next three years, Gen AI looks like Pluto compared to supervised learning.

Supervised learning uses labeled data for classification and predictions, operating on the principle of learning through examples. A human (or even a group of humans, in the case of crowdsourcing) labels a dataset and feeds it to an ML algorithm, which uses the examples to classify or predict new data. It’s the classic type of ML people often think of: give an ML algorithm enough labeled pictures of a cat and it will soon be able to identify them in other images and even predict what cats might look like in the future.

Which AI you need depends on your use case

Though Gen AI and supervised learning are both branches of AI, comparing them technically is like comparing apples and oranges. The use cases and benefits differ vastly. Which AI you choose depends on varying factors like the business problems to solve, the type and quality of available data and the time available for training the model.

In addition to Gen AI and supervised learning, three more AI technologies used in business are unsupervised, reinforcement and deep learning:

Unsupervised learning

One of the serious limitations of supervised learning is the amount of labor involved in labeling all the training data before models can use it to make predictions. Unsupervised learning is more human-like in the sense that it doesn’t need to be pre-trained to analyze something. Taking large swathes of unlabelled datasets, unsupervised learning models cluster data together to identify patterns. However, its use cases are limited to clustering and anomaly detection and it can be complex to implement, with no guaranteed results. This explains why its adoption rate will lag behind supervised learning and Gen AI.

Reinforcement Learning

Reinforcement learning works on the concept of rewarding positive behaviors. In doing so, the agent tries to minimize negative behavior and maximize positive outputs. For example, ChatGPT encourages users to select the best of two responses to a prompt, thereby helping ChatGPT to learn which answers are most helpful. Though the technology may sound simple, it can be used for solving complex tasks in industrial robotics and self-driving cars and is often used in combination with other AI technologies.

Deep Learning

Deep learning is an evolution of ML that works on neural networks consisting of layers of interconnected nodes. The more layers a network has, the better it can learn to solve complex problems, which it achieves in a way not too dissimilar from humans. Deep learning algorithms have revolutionized computer vision and image processing.

Invest in AI — but choose the right one

According to Gartner, business leaders outside of IT plan to spend 6.5% of their functional budget on Gen AI in 2024. Given the current constrained environment, it may be worth first making sure whether other types of AI might make for a better investment in the coming years.

Podcasts