By Amal Joby for Learning Hub
If you use a smartphone, you probably have come across narrow AI.
It’s the level of artificial intelligence we currently have access to, and probably the only one we will have for a few years or decades more.
But don’t be deceived by the term narrow. Although this AI has a limited set of abilities and merely tries to mimic the human brain, it’s pretty adept when it comes to performing the single task it’s designed to do.
It makes self-driving cars a reality, recommends the products you’re more likely to buy and movies you’ll be interested in watching, and even keeps your email inbox free of spam messages.
Narrow AI systems can analyze and interpret data with remarkable accuracy and better timing than humans. It can help us make better data-driven decisions, and more importantly, relieve us from numerous monotonous tasks.
Although this type of machine intelligence lacks consciousness or the ability to reason, it strives to improve all aspects of human life with automation and is an indispensable, ingenious technology of the 21st century.
Simply put, ANI operates within a pre-defined range and can’t think for itself. It performs just the specific tasks it’s designed to do and never attempts to achieve anything beyond them. But this won’t be the case when AI technologies progress and we create machines with human-level intelligence.
General AI vs. narrow AI
Artificial general intelligence, or general AI, is an AI system capable of learning, comprehending, and functioning just like human beings. Such an AI agent will have artificial consciousness (the state of being aware of and responsive to one’s surroundings) and be able to solve unfamiliar problems.
General AI is also known as full AI or strong AI and will have human intelligence. However, we’re years away from creating such artificial intelligence systems, and their estimated time of arrival ranges from decades to centuries, depending on the AI researcher you ask.
One of the most significant differences between narrow AI and general AI is that the former lacks consciousness or the ability to reason. Additionally, general AI has a wide range of cognitive abilities (brain-based skills which are essential for communicating, acquiring knowledge, and reasoning) – similar to humans – while weak AI has none.
The core components of narrow AI, such as machine learning, natural language processing, artificial neural networks, and deep learning, would still be leveraged by AGI, but probably their advanced versions or along with soon-to-be discovered technologies.
AGI systems can perform virtually any intellectual task humans could ever do. As mentioned earlier, they can think and act like humans and can most likely beat us at our own game as they don’t feel tiredness or emotions like fear or sadness unless they’re programmed to do so.
Narrow artificial intelligence systems like Siri, Google Assistant, or Cortana will stutter if you ask questions such as “how would the universe end?” or “what is the meaning of life?” – unless there’s an article on the internet that explains the same. These virtual assistants could be seen as well-performing natural language processing robots as they process our speech and input them into a search engine.
However, a strong AI would be able to come up with plausible answers, just like humans, and can put their imagination into play. They could even ask questions we’ve never heard before, and would probably have the ability to lie. This also means that during the tests for assessing an AI like the Turing Test, the machine could intentionally act dumb to not reveal its intelligence.
General AI could also be our key to achieving artificial super intelligence (ASI), which is the third and ultimate level of AI and could surpass human intelligence and decision-making abilities by a million times. Achieving ASI – let alone AGI – is viewed as an existential threat by many, including Elon Musk and Stephen Hawking.
But there’s little to no reason to be worried right now as we’re years away from achieving either, and some experts claim we might never achieve them. That’s because it’s virtually impossible to model the human brain, which is critical for attaining features of the human mind, including consciousness and sentience.
However, it would be entirely pessimistic and naive to think that there’s not even the slightest chance as the pace of technological progress is tremendous, and we’re unlocking and utilizing new technologies like quantum computing.
Examples of narrow AI
It’s easier to spot narrow AI examples as the technology has impacted almost all spheres of human life. Here are a few of the real-world applications of narrow AI that you probably have come across, especially if you’re hooked on the internet.
Alexa, Siri, and Google Assistant
Hours rarely pass without you interacting with your smartphone’s voice assistant. As you might have guessed, voice assistants like Siri and Google Assistant are the most common examples of narrow AI.
However, they aren’t the best examples as the majority of their tasks are speech recognition-related.
Driven by machine learning and AI technologies, IBM Watson is a question-answering machine used extensively in the healthcare industry. When doctors may take weeks or months to look through documents, Watson can do the same in seconds. Interestingly, Watson was initially developed to answer questions on the television show Jeopardy!.
Watson can also assist companies with risk and fraud management and significantly lowering business costs. It can also help in identifying areas of a business requiring process improvement by analyzing vast amounts of unstructured data sets.
AlphaGo is an AI-based computer program developed by DeepMind – an artificial intelligence company acquired by Google in 2014. It’s the first computer program to beat a professional human Go player and was initially a research project aimed at testing the competence of a neural network at Go.
AlphaGo has numerous advanced versions, including AlphaStar and AlphaZero – the latter being succeeded by an advanced version called MuZero, which can learn without being taught the rules. Creating AI systems that can improve without much training or assistance from humans is critical for achieving AGI, and so, AlphaGo is a huge win for the AI community.
AlphaGo uses deep learning and artificial neural networks to identify the best moves with the highest winning percentages. In terms of competitiveness, AlphaZero is incredibly superior to the earlier versions of AlphaGo and is currently one of the world’s top players in Go and chess.
Narrow artificial intelligence makes it possible for self-driving cars to navigate through traffic, sense obstacles in a lane, and make sure passengers and pedestrians are safe. The enormous volumes of data generated by the cameras, sensors, and GPS fitted on the vehicle are analyzed and processed with the help of ANI.
More precisely, weak AI enables self-driving cars to see, hear, and think. But do note that self-driving cars aren’t made possible by a single ANI – rather through a collection of ANI systems. Although fully autonomous vehicles are still in their infancy, we’re only a couple of years away from being driven around by drivers who don’t get tired.
One of the biggest reasons you feel the urge to check out Netflix or YouTube is their recommendation systems. Such a system is powered by AI algorithms and suggests the shows and videos with impressive accuracy to keep you engaged and entertained.
The system continually learns from your responses to each piece of recommended content. In a way, these algorithms understand your preferences more than anyone else, including yourself.
Chatbots are software applications powered by narrow AI. They’re capable of simulating conversations with humans and can give customers the impression that a company’s customer support is online 24/7.
Chatbots learn more as they communicate with customers and can solve basic issues without requiring human assistance.
Narrow AI: Solving one problem at a time
Narrow AI robots aren’t the ones you could talk with for hours on end, nor can they perform tasks beyond their scope. But they can outperform humans at the specific tasks they’re designed to do and can eliminate numerous tedious tasks. Narrow AI is where we are right now, and strong AI is our next destination.
Visit our friends over at TranscriptionGear to get the rest of what you need! From headsets to foot pedals, they have you covered.