ChatGPT, Herbert A. Simon, and Jean Baudrillard
The role of generative AI in the digital-first strategy
I was going to continue on how to design a digital value loop… but today, I will take a detour and talk about ChatGPT and somehow find a way to the idea of a digital value loop. In doing so, I travel through Herbert A. Simon, design, Jean Baudrillard, and ontological reversal.
Generative artificial intelligence (AI) tools like ChatGPT are some of the most powerful tools humanity has ever created. They can augment, extend, and replace our innate abilities to create new future possibilities of our civilization. However, many point out that they still have some problems and challenges, such as generating inaccurate or distorted information. But contrary to the common view, I believe that these aspects of generative AI should be considered features rather than flaws, and that we can use them to design better futures.
Enter Herbert A. Simon
To understand my point of view, we must start with Herbert A. Simon's distinction between natural and artificial sciences.
For Simon, the purpose of natural science is to understand the world as it is and to explain natural laws through empirical observation and experimentation. Natural science aims to explain natural phenomena, establish cause-and-effect relationships, and develop theories and models to predict and explain them.
Artificial science, on the other hand, aims to create new artifacts, systems, and processes that don't already exist. This involves applying existing knowledge to create solutions to problems or to meet user needs. Simon argued that we live in an artificial world, not a natural one, and that we need to understand how humans create artifacts and how we interact with them. Therefore, artificial science aims to create something that is functional, efficient, and effective in achieving its goals, and involves an iterative process of generating, evaluating, selecting, and refining alternatives until the desired outcome is achieved.
Re-presenting vs. Pre-senting
Against this backdrop, computing tools for the natural sciences are used to “re-present" the real world as accurately as possible. Take search engines as an example: they are designed to help us find accurate information about the world as it exists. As such, search engines are a type of "representational tool."
On the other hand, computational tools for artificial science, such as generative AI, are used to “pre-sent" possible worlds that don't yet exist. These tools are not aimed at accurately representing reality, but are used to explore creative possibilities.
Design is, then, at the heart of artificial science. Designers use sketches, prototypes, and mockups to generate and evaluate ideas before bringing them to life. These are examples of generative tools designers use to envision alternative possible futures. These design tools do not represent the world; instead, they present potential future possibilities.
Designing Digital-First with Generative AIs
Generative AI should be seen as an extension of these design tools. Generative AI recombines existing components to create new provisional objects that humans might like. These provisional design objects can inspire human designers or users or offer novel opportunities. What's unique about generative AIs like ChatGPT is that the provisional design objects they create first exist in a digital form before they take on a material form in the physical world. In that sense, tools like ChatGPT are the most vivid examples of "digital first." We are entering an era where digital objects and generative tools are actively shaping the physical world by first constructing digital versions of the real world. I call this the "digital first" ontological reversal.
So, criticizing generative AIs for inaccuracy stems from confusion about their potential roles in value creation. What we need is a way to use generative AI as a part of on-going value creation process. To do that, one must understand their place in the digital-first strategy and a digital value loop.
Enter Jean Baudrillard and Simulacra
The inspiration for this digital-first concept is the blue dot on Google Maps. Google Maps is an ultimate representational computing tool. It shows you the real world and how you move through it in real time and in great detail. When I move, the blue dot moves. But with autonomous vehicles, the story is 180 degrees flipped. Now, the dot moves first and then the car follows. No longer is the technology just representing the world, but it is now creating the world first. What used to be a copy of the real thing is now the real thing. Jean Baudrillard called this a "simulacra," a copy or representation of something for which there is no original or authentic thing.
From digital-first ontological reversal to digital value loop
From this "ontological reversal" comes the idea of the digital value loop. With a strong digital value loop, companies can algorithmically offer users compelling future possibilities, and for this to happen, a successful digital value loop must have generative capabilities. This is where generative AIs can be deployed effectively.
Several practical observations can be drawn from this discussion.
First, a distinction must be made between representational tools, which are designed to sense the world, and generative tools, which are designed to shape the world. And these two types of tools sit at opposite ends of the digital value loop. So we need both because they fulfill different goals.
Second, since representational and generative tools exist for different purposes, we need to evaluate them differently. Representational tools must be accurate; generative tools must be desirable and plausible. When using ChatGPT as a writing tool, the tool should produce text that is desirable and plausible. Determining what is desirable and plausible depends on the context and must be made by a human.
Third, a generative tool will add business value when it is used as part of a complete digital value loop of inferring, predicting, generating, and orchestrating. Even if you have a powerful generative tool, it will not create user value unless it is part of an effective value loop.
Finally, in the future, we are likely to see micro-digital resources interacting with each other, as they constitute a part of a larger digital value loop. Currently, most microservices only perform a subset of a digital value loop. In the future, imagine an online platform with microservices that have their own micro-digital value loops with their own generative AI. Companies will need to figure out how to optimize the interactions between these autonomous microservices to create a desirable user experience.