Industry executives and experts share their predictions for 2024.  Read them in this 16th annual VMblog.com series exclusive.

2024 is the year “AI building AI” will become mainstream

By Colin Priest, Chief Evangelist at FeatureByte

Over the past year, AI has become more and more mainstream among businesses and the general public. For example, ChatGPT, released in late 2022, has already grown to more than 100 million users. However, while the demand is growing, many don’t realize that the current generation of AIs require a great deal of manual work to build and use. Popular AI models such as ChatGPT and Stable Diffusion require dozens of specialized data scientists to prepare data and write code, and this work can take several months.

Sometimes, rather than replacing or enhancing human work, generative AI’s pedantic input requirements have created more work for humans. In 2023, “prompt engineers” emerged as a career option. These professionals specialize in optimizing the inputs or prompts given to AI systems like ChatGPT to ensure the best possible outputs. They fine-tune the questions and requests sent to an AI system so that it can produce the most valuable responses.

Another challenge with current AI models is their focus. While they’re adept at handling text and images, they falter when dealing with tabular business data. As a result, corporate data scientists often find themselves involved in repetitive tasks like data preparation and coding to make this information AI-compatible.

What if we used AI to automate these mundane tasks? For example, what if you could tell an AI what you want an image to look like, and it could automatically generate optimal (AI) prompts for image generation? Or, what if you could tell AI to automatically write Python code to transform data and train machine learning models? Finally, what if you could tell AI to document and tag the semantics of your business data automatically, and it would suggest the key metrics for your use case?

While this may seem far away, all these things are happening now. You can ask ChatGPT to suggest image generation prompts, based upon your natural language description of your requirements.  Researchers have published a paper detailing how they used AI to successfully auto-generate code to prepare data and train machine learning models for a popular data science competition. And, one software startup is using Large Language Models (LLMs) to intelligently tag and interpret data, then suggest and rank the key metrics required to implement many popular business use cases, providing insights that previously required human expertise.

The implications of these advancements are profound. If AI systems can take on the mantle of developing or assisting in the development of other AI systems, we’re looking at a future where efficiency scales exponentially. This self-reliant model of AI development could reshape industries, making AI not just a tool but an active participant in its evolution.

For end-users, this means an enhanced experience where they can interact with AI using everyday language. The barrier of needing specialized knowledge to communicate with, or even develop, AI models would diminish. This democratization of AI access can lead to a wider adoption across sectors, from small businesses to academic research.

Moreover, the expertise of data scientists and developers would be freed from routine tasks. Their skills could be redirected towards innovation, research, and areas where human intuition and creativity are paramount. We’re going to see much more of this automation going forward…so here is my prediction….

I predict that in 2024 AI-building AIs will become mainstream tools. And as a result, we will see an explosion in productivity, because people will use natural language to interact with AI systems and they will build AIs without coding skills. The end result is that all of this automation will free up people’s time to do more human-facing, value-added tasks instead of routine, mundane tasks.

##

ABOUT THE AUTHOR

Colin Priest is Chief Evangelist at FeatureByte. With a focus on data science initiatives, he has held several CEO and general management roles, while also serving as a business consultant, data scientist, thought leader, behavioral scientist, and educator. He has over 30 years of experience across various industries, including finance, healthcare, security, oil and gas, government, telecommunications, and marketing.