us clarify this definition:
Being able to successfully perform work: This implies that the entity or artifact can perform work, and its success is determined in accordance with the goal set for the entity. Work implies the activities conducted by humans that add value for human life and survival.
Goal-directed behavior: Goal-directed behavior implies that the entity is operating with a sense of purpose and not just randomly. It has a goal. This goal may have been given by a human or it may have assigned the goal to itself.
Uncertainty: Uncertainty implies the ability to navigate through situations where more than one choice exists. Of course, the greater the uncertainty, the more intelligence is needed to navigate. In a simple system, intelligent decision could be as simple as making a binary choice between two options. For example, a simple regulator can start the air conditioning or not. In a more complicated system, more choices are available. For example, in a game of chess, depending on the position, a certain number of moves are possible. Another example will be of your spellchecker or text recommender to fill in words as you type. In even more uncertain situations, for example, autonomous cars navigating and driving on a busy street, the number of choices is unimaginably large.
But performing or accomplishing work requires more than intelligence. It requires us to take actions.
ACTIONS
An intelligent entity must interact with its environment. A thinker who loses the capacity to interact with the environment will only have thoughts in his or her head but will not be able to perform any action. Also, this thinker will not be able to entertain any new input or sense any new information since it is disconnected with the environment. A thought trapped in a mind incapable of any interaction with the environment is not helpful to accomplish any work task. To perform a work task, therefore, an intelligent entity needs to interact with its environment.
Perhaps it will be helpful to make a distinction between thinking and acting. When we act, we create a change in the environment. Whether action is composed of movement or transferring data or sending a notification, it is something that materializes in the environment in which an intelligent artifact is operating. In other words, it does not happen in the mind of the artifact as it necessarily requires some type of expression that is based on interaction with the environment. As we make the distinction between thinking and action, we can view intelligence (thinking) and action as interdependent and interconnected. Thinking drives action—for example, you decide to make a right turn while driving—but action also drives thinking. For instance, as you start turning the steering wheel, the feedback from the car, the steering wheel that you are holding, and the surroundings send signals to your brain and the brain helps you to keep the car on the road as you perform the action of turning the car by turning the steering wheel. This hand and mind coordination can be viewed as constant back and forth rapid transition between states of thinking and acting. A thinking system does not think in isolation; it interacts with its environment.
AUTOMATION
Automation is the ability of a synthetic entity to perform work. Since work requires both intelligence and actions, automation involves both.
Thinking automation: Automation automates the thinking part where a synthetic entity can be intelligent (refer to the definition of intelligence presented above) autonomously. Here autonomously implies that it can make decisions to navigate through uncertainty on its own.
Action automation: Automation automates actions where non-thinking parts of work are automated. For instance, a car automates movement on land, an airplane automates mobility in the sky, a non-thinking computer automates work tasks such as spreadsheets, word processing, and others. Instead of walking, you ride in a car. Instead of flying (not sure how a human would fly, perhaps jump or fall is a better comparison), a human can fly in an airplane. A human driving a car, flying an airplane, or using a computer is benefiting from the automation in these artifacts even though he or she is using his or her own cognitive skills to operate these machines. These machines are not intelligent, but they are automated. That automation is the automation of action where an artifact can enable human work that requires interaction with the environment.
INTELLIGENCE ACTION CHAIN AND SEQUENCE
From a business process perspective, one can view a process as composed of intelligence-action sequences that accomplish work. In fact, it can be viewed as state transitions between thinking (decision making) to acting and doing them successively till the work sequence is finalized and the work goal is achieved. The action itself, or at least certain part of it, does not require any thinking. Going back to our example of you deciding to turn your car right, the “decision” in this case is to turn right, but an action follows. The action itself can be broken down into states of action to turn, as brain and hand coordination maintains what turning is. Thus, decision-oriented intelligence decides to turn right, and then brain, eyes, hands, and other sensors work collaboratively to help us make the right turn. We receive feedback, and our brain processes this feedback to turn the wheel in accordance with the feedback to keep the car on the road, accident free, and make the right turn. The think-act sequence itself is the second kind of decision—i.e., doing what needs to be done or monitoring an action-neural response (feedback) sequence once an action has begun.
Clearly, the automation from artificial intelligence requires the convergence of the two types of automations: automation of thinking (synthetic, or machine, thinking) and automation of action. It is a merger of the two where artifacts can think autonomously and take actions.
The acts of action can be many. Any time information is extracted from the environment and goes into a system (artifact, entity) as input and then when the environment is acted upon by the system as some form of output, these are actions. Thus, when the machine receives market data, it can be viewed as specific steps where no thinking is required by the machine. Then the machine thinks and makes a trading decision. The decision when communicated back to the environment where a trade is made is again an action and requires execution and not just thinking.
ENTERPRISE SOFTWARE
The enterprise software, then, is composed of a combination of “thinking” and “acting” software. The thinking-acting sequences imply integration of AI software with non-AI software to build work-task sequences. But these task sequences are not built around automating human-centric processes. In other words, automation requires rethinking the business models, and processes need to be built around machine work. Machines work differently than humans. In the next chapter, we will cover the design principles of AI-centric designs. At this stage, it is important to recognize that designing a modern investment management firm requires building an integrated software architecture of non-intelligent (legacy or traditional software) and intelligent (machine learning, rules-based) software.
DATA
Data is the lifeblood of machine learning. Without data, machine learning models fail to learn. In fact, not only do we need to have plenty of data, its quality needs to be good for the learning to be consistent with the goals of developing the artifact. Since each firm has its own data, the potential for each firm to perform in the AI era will be different.
Here are some of the capabilities needed for an AI-centric transformation:
Data Management Expertise
What data do you have? What data do you not have but is needed? What data is needed but you cannot have? What data is essential for your core operation?
Having data is one thing,
