Is Artificial Intelligence Truly Intelligent?

Preparing for the Singularity

In the news recently there has been some really valuable discourse about artificial intelligence by some of the most influential representatives in the industry. If you ignore the media over-sensationalization of the discourse, you'll actually find that there are two views associated with artificial intelligence and its potential impact on humanity. However, it's important that we first understand where we are with artificial intelligence.

 Some of my data science colleagues get rather riled up when they hear the term "artificial intelligence" thrown about so wildly. Some regard artificial intelligence as a description for a field of computer science that has yielded a manufactured approximation of human intelligence. Without getting into what human intelligence actually is, its commonly identified as the situational application learning, creativity, ingenuity combined with context and "common sense".  

Weak AI 

 Weak AI can be described as the focusing of a subset of AI capabilities on a single task. For example, the ability to learn data from a set of labeled data that is provided to an algorithm and answer questions such as the following:

• Are these things the same or different?
• Is this an outlier (is it weird)?
• Can I predict an outcome with specificity?
• What is this thing?
• What does this phrase mean?

In each case, we are asking the system to learn and apply its capabilities to new situations, but only within the frame that the machine learning model has been trained on. If you ask a computer vision model that has been trained on a data set consisting only of pictures of cars to differentiate between a car and a sandwich it won't be able to.

In essence, this is the application of algorithmic machine learning to narrow problems to answer specific sharp questions. There is no sentience or general common sense awareness or creativity involved in the application of AI in this sense. It is in this branch of AI that we see the most progress. The ability of vendors to supply configurable machine learning models that answer these questions with increasingly less training and ownership tasks is accelerating. These enable smart digital experiences, but are not yet close to rocketing humanity towards the singularity (when AI gains sentience). Yet.

Strong AI

Strong AI, also known as an Artificial General Intelligence (AGI) is what every movie since the mid -80s to now has taught us to be wary of. An AGI will most likely take the form of a system that can incubate and integrate learning, creativity, ingenuity and other cognitive capabilities. In this situation, an AGI can consume Weak AI and add it to its capabilities to create a seeing, hearing and thinking machine. This machine could leverage common sense, weigh outcomes and develop opinions. Strong AI might be able to do all the things Weak AI provides as well as the following:

•  Disregard petabytes of training data in order to consider new information and change decision making processes
• Formulate an opinion and have that opinion shape its future behavior
• Question its environment and use Weak and Strong AI capabilities to acquire knowledge and information and eventually take action in the real world

 Ultimately, it can be regarded as a system of intelligence that can perform the same cognitive tasks as a human. A valid observation might be that the rapid pace of Weak AI development will eventually summon the manifestation of Strong AI. This happenstance is farther out on the horizon. 

 So What Does This Mean?

 Just like science communicators, artificial intelligence communicators must be specific and frame their discussions. Those that are warning against being responsible and perhaps even afraid of the impact of AI generally may mean Strong AI. This is of significant importance because when the singularity does happen, humanity will need to be prepared and ethical. These viewpoints may extend to the rapid development of Weak AI as well. The rapid development of Weak AI will give Strong AI significant capabilities (seeing, hearing, communicating, judging, measuring, etc.). While the focus of the warnings on Strong AI may be farther out on the horizon, humanity must anticipate and respect the potential and be prepared for it.

 Those that are indicating that we may not have to worry for a while are most likely focused on the development and inclusion of Weak AI into digital experiences that assist and augment human capabilities. This development includes folding Weak AI into apps, devices, appliances and other elements of human life that will have narrow focus to begin with. As these Weak AI capabilities increase in their efficiency, management and capability, the inclusion of AI into our world and our experiences will arguably get better, thereby improving our quality of life. Examples include autonomous driving, digital assistants, news feeds that anticipate our interests, entertainment and programming that conforms to our desires. 

 Regardless of Strong AI or Weak AI, the warnings are timely. We should pay attention to inclusivity, privacy, governance, and transparency when utilizing AI to benefit humanity. We should adhere to principles and make sure we prepare for the singularity and determine what to do when AGIs become possible and pervasive.

.