2 Mapping out the modern AI landscape

As a PM, you will be more successful if you build products based on a solid technical understanding of the technologies used in your product. For example, if you are using Large Language Models, you should understand the benefits and limitations of your specific model and know how it differs from comparable models. With this understanding, you can confidently navigate the space of technological options, maximize the value of AI in your application, and manage related risks and limitations. In this chapter, we will introduce a basic categorization of the AI space, which we can widely use to characterize specific AI systems. Think of it as a “mind map” that allows you to situate specific applications and delimit your potential solution space. In subsequent chapters, we will be using this basis to elaborate on further technical details of AI applications in each of the categories.

2.1 Some important categories of AI

2.1.1 Assisted, augmented, and autonomous intelligence

2.1.2 Analytical, generative, and action AI

2.1.3 Symbolic vs. neural AI

2.2 Understanding AI modalities

2.2.1 Numerical modality

2.2.2 Textual modality

2.2.3 Visual modality

2.2.4 Auditory modality (incl. speech)

2.2.5 Sensorimotor modality

2.2.6 Emotional modality

2.2.7 Code modality

2.2.8 Multimodal AI: Combining modalities

2.3 Your AI toolbox

2.3.1 Data

2.3.2 Models and algorithms

2.3.3 Output validation

2.3.4 Plug-ins and integrations

2.3.5 LLM frameworks

2.3.6 Computing infrastructure

2.3.7 MLOps and LLMOps

2.4 Summary