Last year generative artificial intelligence (AI) took the world by storm as advancements populated news and social media. Investors were swarming the space as many recognized its potential across industries.  According to IDC, there is already a 26.9% increase in global AI spending compared to 2022. And this number is forecast to exceed $300 billion in 2026.

It’s also caused a shift in how people view AI. Before, people thought of artificial intelligence as an academic, high-tech pursuit. It used to be that the most talked-about example of AI was autonomous vehicles. But even with all the buzz, it had yet to be a widely available and applied form of consumer-grade AI

However, that’s changed. Generative AI has exposed the public to what’s possible. People have become active contributors, making music videos, children’s books and even redesigning creative workflows. Who hasn’t tried Lensa or ChatGPT by now? It has become clear that AI is not just for bloggers or programmers; it’s for everyone. It has the power to evolve education, and it’s an enormous productivity booster that can streamline the modern creative process. 

In 2022, we were allowed a peek at what generative AI can do, but as you may have already guessed, there is much more to come.

Building a solid foundation for future growth 

Most people have heard of generative AI models such as GPT-3, BERT, or DALL-E 2. These are foundational models. OpenAI’s ChatGPT is also built with GPT-3 technology but a slightly enhanced version GPT-3.5. And more recently, ChatGPT-4 was released with greater capabilities, including greater accuracy, more creativity and more collaboration — further proof that AI can and will continue to improve. 

Foundational model is a term coined by the Stanford Institute of Human-Centered Artificial Intelligence to classify a type of tool that can execute simple tasks or outputs. In our case, the task is generating a text or an image. Foundational AI models are typically open-source, meaning they can be used by others or combined with other datasets to serve as core building blocks for large language models (LLMs). 

The foundational models have been instrumental in leading the way for further advancement. They provide a base layer that application players can build on. And that is where the next wave of innovation will take place.

Back to the present

We believe the industry needs to look beyond using generative AI tools for simple outputs. Instead, it’s worth focusing on building computing capabilities and optimizing what is possible for users and large enterprises. Generative AI doesn’t have to mean generic AI. AI solutions are not one-size-fits-all, leaving a need for personalization based on individual needs. In turn, those who opt to implement AI into their workflow have the potential to achieve unique and impactful outputs that resonate with their customers.

Today, many players in the space are making new strides with generative AI’s unlimited possibilities, collectively pushing toward AI maturity. Now, the question is: How do we move forward with the tools that are available?

The road ahead

While 2022 drastically changed the AI narrative, it’s also safe to say that with the current rate of innovation, everything we know about generative AI now is going to radically change within the next 12 to 24 months. Just look at the news the first three months of 2023 have brought us: Google is injecting AI into search, Gmail and Docs, while Microsoft is doing the same with Bing, Edge and Skype.

We believe that there is another approaching wave of breakthroughs that will result from combining foundational models with open-source and user-centric use cases. Bringing all of these together and giving the user exactly what they need at a specific time will be the next big thing. We’re already seeing companies like Snapchat, Notion and Meta implementing generative AI directly into their products to provide services better suited to their users’ needs.

Where many current models fall short is in the attempt to be one-size-fits-all. This approach is prone to factual errors and bias. Customization will lead the way from now on. It offers an opportunity to continue building from open-source models and zero in on segmented needs. Individual customers can refine their own voice within an institution, and enterprise customers can create workflows to be as exact as they need, with the ability to refine over time.

Generative models will perform best when they are implemented in ways that give greater control to users so they can achieve their ideal outcomes. Embracing that ongoing relationship and technical malleability for optimal use case results will be key.

LeackStat 2023