The definition of artificial intelligence

? & ! with Lee Coulter

Add bookmark
Lee Coulter
Lee Coulter
06/28/2019

Artificial General Intelligence

HAL 2000. That’s the answer most people think of when the term AI is used. The notion of Artificial General Intelligence (AGI) is so expansively large, it will likely not be humans that create it. For all practical purposes, we are a long way from this (read: decades). The arrival of AGI is often referred to as the “singularity”, because once a system is sufficiently capable of general intelligence, it no longer needs humans to write it. In fact, it is extraordinarily likely that a narrow or weak AI designed to write software will actually write the successive iterations of software that expand the capabilities of intelligent systems.

Next Best Action Machine Learning

As you know, it irks me a LOT that people use these terms without understanding what they mean or what the data requirements are to make such systems useful. As we look to the immediate future in Intelligent Automation, we are really not seeking weak AI, or strong AI and certainly not AGI. What will actually provide significant business value in the near term is the injection of Next Best Action Machine Learning (what I am calling NBAML) systems into task automation systems to produce an autonomic system (defined below). It is in an autonomic system that we are seeking to manage end to end processes such that it produces digital labor (also defined below).

Narrow artificial intelligence: Narrow AI

Today, people use AI when talking about OCR, ICR, and NLP. These are infact narrow AI systems, with EXTRAORDINARILY NARROW knowledge domains. For example, the knowledge domain of OCR is a set of 50 or so shapes we use in common English language (note, in Chinese, this knowledge domain is identical  -- a set of shapes human use to convey meaning. The challenges of OCR in different language are different but not necessarily more difficult to solve).

My definition of artificial intelligence

So if you ask ME, what AI is for practical use in business today, I offer definitions below. If you ask me what generally others mean when they use the term, there are two populations: one that really has no idea and thinks AGI is either terminator or Wall-E, and the other is a group of people that are taking the most basic NBA ML “orchestrators” to execute tasks in the right order. This may include the addition of ultra narrow knowledge domain AI systems such as OCR. Can you say the system overall is AI? No. By and large, they are systems that are getting better and better at making good decisions on which order to execute tasks in such that a meaningfully valuable business result is produced (digital labor).

Here again, the width of the knowledge domain of the prediction/prescription system is operating is infinitesimal. They are purpose built to literally predict and prescribe the correct set of finite actions in a single business process.

DEFINITIONS

Intelligent Automation is RPA + AI (or NBAML) to create systems that can fulfill a role as digital labor in the enterprise. I have been hearing this notion of “integrated automation”, a term I used some years ago to describe full-stack automation that is sufficiently smart, it can complete a significant portion of an end to end business process without much human supervision in the wild. Yet another marketing term to create separation from the workhorse of today’s task automation. Let’s keep our eyes focused on the next three years rather than three decades.

digital labor: Digital automation of information technology systems and/or business processes that successfully delivers work output previously performed by human labor or new work output that would typically or alternatively have been performed by human labor. Syn: dFTE, eFTE See also: automation, business process.

Intelligent Process Automation: Preconfigured software instance that combines business rules, experience-based context determination logic, and decision criteria to initiate and execute multiple inter-related human and automated processes in a dynamic context. The goal is to complete the execution of a combination of processes, activities, and tasks in one or more unrelated software systems that deliver a result or service with minimal or no human intervention. See also: activities, automation, business rule, decision, process, service, task.

artificial intelligence (AI): The combination of cognitive automation, machine learning, reasoning, hypothesis generation and analysis, natural language processing and intentional algorithm mutation producing insights and analytics at or above human capability.  See also:  cognitive automation, machine learning.

NOTE—This is distinct from artificial general intelligence (AGI)

machine learning: Detection, correlation, and pattern recognition generated through machine-based observation of human operation of software systems along with ongoing self-informing regression algorithms for machine based determination of successful operation leading to useful Predictive Analytic or Prescriptive Analytic capability. See also:predictive analytics, prescriptive analytics.

narrow artificial intelligence (Narrow AI or Weak AI): Complex computational AI system capable of providing descriptive, discovery, predictive and prescriptive analytics with relevance and accuracy equal to or exceeding a human expert in a specific knowledge domain. (Also known as Weak AI). See: artificial intelligence (AI), descriptive analytics, predictive analytics, prescriptive analytics.

autonomic: A monitor-analyze-plan-execute (MAPE) computer system capable of environmental sensing, policy interpretation, access to knowledge (data→information→ knowledge), decision making, and initiating dynamically assembled routines of choreographed activity to both complete a process as well as update the set of environmental variables that enables the autonomic system to self-manage its own operation as well as the processes it oversees. See also: choreography, knowledge, process.

An autonomic system is identified by eight characteristics:

  1. a)Knows what resources it has access to, what its capabilities and limitations are and how and why it is connected to other systems.
  2. b)Is able to configure and reconfigure itself depending on the changing computing environment.
  3. c)Is able to optimize its performance to ensure the most efficient computing process.
  4. d)Is able to work around encountered problems either by repairing itself or routing functions away from the trouble.
  5. e)Is able to detect, identify and protect itself against various types of attacks to maintain overall system security and integrity.
  6. f)Is able to adapt to its environment as it changes, interacting with neighboring systems and establishing communication protocols.
  7. g)Relies on open standards and requires access to proprietary environments to achieve full performance.
  8. h)Is able to anticipate the demand on its resources transparently to users.

For the complete IEEE 2755 – 2017 Standard please visit here for a copy.


RECOMMENDED