1.6. Real-world product categories #

There are a few categories of products where these models are already being used in.

The leading category in this space is AI assistants, which have evolved significantly beyond earlier versions like Apple’s Siri or Amazon’s Alexa. The key distinction lies in how interactions are programmed. Earlier assistants relied on hand-coded responses, excelling in narrow tasks like weather reporting but falling short in complex assignments. In contrast, LLMs can easily draft a creative homework assignment, but are, in exchange, prone to factual inaccuracies due to a phenomenon known as “hallucinations.” We’ll discuss it in the next chapter. So while these models represent the most comprehensive aggregation of human knowledge to date, they lack an inherent understanding of truth.

AI assistants are typically designed for general-purpose use, capable of performing a wide range of tasks. However, a burgeoning subcategory focuses on AI companionship. These are specialized AI personas that serve as your digital friends, peers, or acquaintances. For example, apps like Character allow you to engage in conversations with characters from your favorite books and movies just for entertainment. Meanwhile, platforms like Facebook are developing AI personas intended to act as, say, your personal trainer, motivating you to hit the gym.

Another category includes AI utilities, which can be integrated either into the front end or the back end of existing products. On the front end, these utilities could include features like document summarization in platforms like Google Docs, voice transcription in Zoom, or automated thumbnail generation for WordPress articles. On the back end, large language models can be employed for tasks like optical character recognition or tagging unstructured text data at scale. While these capabilities offer substantial utility, they usually need to be incorporated into broader products to be truly effective, as they provide limited standalone value.

The most innovative yet least validated category in the realm of generative AI consists of autonomous agents. While it’s debatable whether these models can truly think like humans, they have demonstrated the ability to mimic certain aspects of human thought. For instance, they can tackle logical problems, ranging from simple to complex, and some can even ace university exams.

This has led to their use in creating task-solving loops. Essentially, you can instruct a model to devise its own action plan, then have it execute the steps it outlined. A rudimentary example would involve asking a large language model to generate a table of contents for a book, then having it create subsequent chapters based on that framework until the book is complete. While potentially revolutionary for multiple industries, these autonomous agents are still in early stages, delivering results that are too inconsistent for reliable production use.