Artificial intelligence (AI) is automated technology capable of performing functions previously dependent on the human mind. At a basic level, artificially intelligent systems can perform tasks commonly associated with human cognition, like interpreting speech, playing games, and identifying patterns. AI systems process massive amounts of data to model their decision-making. Humans often supervise an AI’s learning process, reinforcing good and discouraging bad decisions. Some AI systems are designed to learn without supervision.
Perspective
AI is nothing new. It’s been around for a long time. A better way to think about AI is more automated automation. As computer programming has evolved, it has become possible to store larger and larger data sets inside a computer’s memory. The natural human tendency is to anthropomorphize technology and make it seem human. Rest assured; there is nothing human about AI. It is simply a way to get things done faster than ever before. AI has the potential to supplement and enhance human activities like:
- Data Entry
- Customer Service / Chat / Call Center
- Proofreading & Grammar
- Paralegal Services
- Bookkeeping & Records Management
- Translation
- Transcription & Copywriting
- Market Research
- Email & Social Media Management & Marketing
- Appointment Scheduling
- Telemarketing Services
- Virtual Assistants
- News Aggregation & Reporting
- Travel Agency Services
- Education, Tutoring & e-Learning
- Technical Support
- Content Moderation & Governance
Machine Learning
A machine learning algorithm is fed data by a computer. It uses statistical techniques to help it “learn” how to get progressively better at a task without necessarily being programmed explicitly for it. Instead, ML algorithms use historical data as input to predict new output values. To that end, ML consists of supervised learning (where the expected output for the information is known thanks to labeled data sets) and unsupervised learning (where the desired results are unknown due to unlabeled data sets).
Deep Learning
Deep learning is a type of machine learning that runs inputs through a biologically inspired neural network architecture (modeled after how we think the human brain works). Neural networks contain several layers of data processing. This allows the machine to dive “deep” into its learning to make connections and weigh input to achieve the best result or outcome.
Four Types of AI
AI is divided into four categories based on the type and complexity of a task the system can perform:
- Reactive machines
- Limited memory
- Theory of mind
- Self-awareness
Reactive AI
Reactive artificial intelligence is the most basic form of AI. It is programmed to provide a predictable output based on the input received. Reactive machines will always respond to situations in the same way, every time. They cannot learn actions or conceive of the past or future.
Examples of reactive AI:
- Deep Blue – chess-playing supercomputer. It beat world champion Gary Kasparov.
- Spam filters – email technology that keeps promotions, mass-email, and phishing attempts out of your inbox.
- Recommendation engines – like the one that recommends movies to you when you watch Netflix.
At a basic level, reactive AI was a significant step forward in AI development. However, this type of technology can’t function beyond the basic tasks it was initially designed for. Think of AI as the basic building block and foundation of automation technology.
Limited Memory AI
Systems that can learn from the past and build experimental knowledge by observing actions or data are called limited memory AI systems. These systems store, aggregate, and analyze data in combination with pre-programmed information and instructions to make predictions and perform complex classification tasks. It is the AI most widely used today.
Self-driving vehicles use limited memory AI to observe cars and objects on the road to help the vehicle “read” the road, interpret what is happening, and predict what is about to happen. A good example is “smart cruise control” in many vehicles today. The AI system can recognize when to apply the brakes to prevent a crash, thus making it safer on the road.
As the name implies, limited AI is still limited. The information a self-driving vehicle works with is fleeting and not saved in the car’s long-term memory.
Theory of Mind AI
At some point in the future, machines will acquire decision-making capabilities similar to humans. Systems containing theory of mind AI will be able to understand and remember emotions and adjust behavior and reactions based on those emotions as they interact with people. With this type of AI, holding a meaningful conversation with an “emotionally intelligent” robot that looks and sounds like a real human being will become possible.
The current hurdle to achieving the theory of mind AI is that rapidly shifting emotions are fluid in human communication. Stated simply: it is difficult (I think it’s impossible) to mimic the human heart.
Practical applications could be achieved with theory of mind AI:
- Human cognitive processing and learning (educational, classroom, certification)
- Emotional capacity and resilience (modal talk therapy, psychoanalysis)
- Logic and debate
- Willpower and mindset
- Idea development and brainstorming
- Game theory and outcome modeling
Humans have been making progress on theory of mind AI. The Kismet robot head was developed in the 1990s by Professor Cynthia Breazeal. It is a system designed to recognize emotional signals on human faces and replicate these emotions on its own face. A more recent example, developed in 2016, is the Sophia robot developed by Hanson Robotics.
Self-aware AI
We haven’t developed this type of sophisticated AI yet, and don’t have the hardware or algorithms to support it. Like humans, systems with self-aware AI will be aware of their internal emotions and mental states. They would make inferences and statements like, “I’m feeling happy because someone gave me a smile,” or, “I’m feeling angry because someone interrupted me while I was speaking.” When machines can be aware of their own emotions and the feelings of others around them, they will have a certain level of consciousness that is similar to humans.
What Does the Future Hold?
The question and concern around AI technology (and a critical thing to remember) is that humans are at the root of the technology. Humans developed IT. We are responsible for governing IT. We must be accountable for (and with) IT. I envision a future where humans and their “AI counterparts” can coexist and co-evolve, making each other more vital, resilient, and intelligent every step of the way. I am a cynical optimist in general, especially regarding technology.
What do you think?