Personal tools

Narrow AI

Madison_Wisconsin_dsc00590
(Madison, Wisconsin - Alvin Wei-Cheng Wong)

 

 Don't be afraid of fail. Be afraid not to try.

 

 

- AI Type-1: Artificial Narrow Intelligence (ANI)

Artificial Narrow Intelligence (ANI) also known as “Weak” AI is the AI that exists in our world today. Narrow AI is AI that is programmed to perform a single task . It is a machine intelligence that is limited to a specific or narrow area. Weak Artificial Intelligence (AI) simulates human cognition and benefits mankind by automating time-consuming tasks and by analyzing data in ways that humans sometimes can’t. Many currently existing systems that claim to use "artificial intelligence" are likely operating as a weak AI focused on a narrowly defined specific problem. A better definition of  weak AI is “augmented intelligence,” which makes it clear that the technology is here to complement the human mind, not replace it.

Narrow (or Weak) AI systems do not have General intelligence; they have specific intelligence. An AI that is an expert at telling you how to drive from point A to point B is usually incapable of challenging you to a game of chess. And an AI that can pretend to speak Chinese with you probably cannot sweep your floors. Weak AI lacks human consciousness, though it may be able to simulate it. However, Weak AI can automate the boring, repetitive parts of most jobs and let the humans take care of the parts that require human care and attention.

Weak AI is both the most limited and the most common of the three types of AI. The idea behind weak AI isn’t to mimic or replicate human intelligence. Rather, it’s to simulate human behaviour. So, it’s nowhere near matching human intelligence, and it isn’t trying to. A common misconception about weak AI is that it’s barely intelligent at all - more like artificial stupidity than AI. But even the smartest seeming AI of today are only weak AI. In reality, then, narrow or weak AI is more like an intelligent specialist. It’s very intelligent at completing the specific tasks it’s programmed to do.

 

- Weak AI Examples

Narrow AI systems are good at performing a single task, or a limited range of tasks. In many cases, they even outperform humans in their specific domains. But as soon as they are presented with a situation that falls outside their problem space, they fail. They also can’t transfer their knowledge from one field to another.

While narrow AI fails at tasks that require human-level intelligence, it has proven its usefulness and found its way into many applications. Examples include Facebook’s news feed, Amazon’s suggested purchases and Apple’s Siri, the iPhone technology that answers users’ spoken questions. Email spam filters are another example of Weak AI where a computer uses an algorithm to learn which messages are likely to be spam, then redirects them from the inbox to the spam folder.

Weak AI helps turn big data into usable information by detecting patterns and making predictions. It’s the volatile rate of data evolution that creates the prerequisite for Narrow AI. Adding to it is the recent avalanche of user-generated content, it’s clear that no organization (or human) can cope without the aid of Narrow AI.

 

MIT_DSC_002
(MIT photo courtesy of Yu-Chih Ko, MIT)

- The Difference between General (Strong) AI and Narrow AI

In contrast to Strong AI, which can learn to do any task a human does, Weak AI (or Narrow AI) is limited to one or few specific tasks. This is the kind of AI that we currently have. In fact, deep learning, which is named after (and often compared to) the human brain, is very limited in its capabilities and is nowhere near to performing the kind of tasks that the mind of a human child can perform. And that’s not a bad thing. 

In fact, Narrow AI can focus on specific tasks and do them much better than humans can. For instance, feed a deep learning algorithm with enough pictures of skin cancer, and it will become better than experienced doctors in spotting skin cancer. This doesn’t mean that deep learning will replace doctors. You need intuition, abstract thinking and a lot more skills to be able to decide what’s best for a patient. But the deep learning algorithms will surely help doctors perform their jobs better, faster and tend to more patients in a shorter amount of time. It will also cut down the time it takes to educate and train professionals in the health care industry.

But while all this talent focuses on finding a way to create Strong AI that can compete with the human brain, we’re missing out on plenty of the opportunities and failing to address the threats that current weak AI technology presents. Some commentators think Weak AI could be dangerous because of this "brittleness" and fail in unpredictable ways. Weak AI could cause disruptions in the electric grid, damage nuclear power plants, cause global economic problems, and misdirect autonomous vehicles. In 2010, Weak AI trading algorithms led to a “flash crash,” causing a temporary but significant dip in the market.

 

- Different Types of Narrow AI Technologies

The narrow AI techniques we have today basically fall into two categories: symbolic AI and machine learning. 

  • Symbolic artificial intelligence, also known as good old-fashioned AI (GOFAI), was the dominant area of research for most of AI’s history. Symbolic AI requires programmers to meticulously define the rules that specify the behavior of an intelligent system. Symbolic AI is suitable for applications where the environment is predictable and the rules are clear-cut. Although symbolic AI has somewhat fallen from grace in the past years, most of the applications we use today are rule-based systems. 
  • Machine learning, the other branch of narrow artificial intelligence, develops intelligent systems through examples. A developer of a machine learning system creates a model and then “trains” it by providing it with many examples. The machine learning algorithm processes the examples and creates a mathematical representation of the data that can perform prediction and classification tasks. 

For instance, a machine-learning algorithm trained on thousands of bank transactions with their outcome (legitimate or fraudulent) will be able to predict if a new bank transaction is fraudulent or not. 
 
 

 
[More to come ...]


Document Actions