AI Cheat Sheet

AI is no more and no less the drive to create robots with human minds so they can do everything we do and more.

There is narrow AI—inventions like digital assistants—and general AI that's the more ambitious challenge to create thoroughly intelligent robots. The latter seems out of our reach, but experts say that goal could be accomplished by 2075. If achieved, robots would either change our world for the better or destroy it.

As a topic that generates a lot of discussion and interest from the C2C community, particularly when they’re getting started with bringing AI or ML into their stacks, we set out to create this cheat sheet to make it easier. 

AI Types

• Narrow AI (weak AI): Smart machines that are the most well-known and can perform a single task or set of closely related jobs. For example, digital assistants like Alexa or Google Search, image recognition software, and self-driving cars.
• General AI (strong AI): The drive to create robots that simulate the human mind and even supersede it. In these sci-fi Star Trek-type scenarios, machines can think strategically, abstractly, and creatively, as well as effortlessly handle complex tasks. They have ambition, emotions, social skills, and conscience, just as we.

AI Fields

For scientists to accomplish their lofty ideal of recreating the human mind in robots, they had to split AI into various fields. The most common are:

• Speech Recognition, where robots process and transcribe human speech into writing.
• Reasoning/Problem-Solving/Planning, where robots replicate the function and processing of the human brain. One computational model is artificial neural networks (ANNs) that loosely model our biological neural networks. 
• Memory/Remembering the Past: We achieve that through programming robots with recurrent neural networks (RNN).
• Theorem Proving (or mathematical capacity): Machines have yet to invent new theorems.
• Natural Language Processing: A robot's ability to intuitively write and read text just as we do.
• Computer Vision (Machine Vision): This is the ability to process just as humans do visually.
• Image Processing: It's not just enough to process visually; robots also have to discern images and recognize their shapes and features just as we.
• Pattern Recognition: Robots learn to recognize patterns and regularities in data.
• Image Scanning: We scan images left to right or up/ down. Scientists program robots to simulate that ability through the computational model of convolution neural networks (CNN).
• Motion and Manipulation/Robotics: This is the ability to fluidly understand, manipulate, and move around in environments.
• Social Robotics: Robots learn to interact with humans in a natural interpersonal manner.

Related Topics

• Machine Learning: Often confused with AI, ML is teaching machines to analyze, predict, and advise possible outcomes. In contrast, AI is the process and the state of having robots cogitate like humans.
• Deep Learning: To teach robots how to master complicated things like reasoning, planning, decision-making, etc., without human assistance.
• Data Mining: This is the process of extracting specific data from big data. When it comes to ML, some of the most common data mining kinds that engineers do are Bayesian networks, decision trees, and neural networks.
• Internet of Things (IoT): IoT deals with devices that interact using the Internet. AI enables these devices to communicate, collect, and exchange information.

AI Approaches

Our human cognitive processing is automatic. Since computers function through binary digits, scientists have to convert our mental models into computer language so that machines can “think” like us. The most common methods used to teach machines are the following:

• Symbolic Learning: Symbolic AI, also known as Good Old Fashioned AI (GOFAI), feeds labeled data of real-world entities or concepts they want the machine to memorize. Computational algorithms include AI programming languages and explicit symbolic programming, both are easier to control and debug, and best for abstract problems.
• Non-symbolic Learning: Scientists feed raw data to the machine and expect it to form patterns and high-dimensionality representations of that data. Computational algorithms include Bayesian learning, deep learning, neural nets, and connectionism and are used for big data and best for perceptual problems.
• Sub-symbolic Learning: A hybrid of symbolic and non-symbolic methods, used in speech recognition, as an example.

AI Knowledge-Based Systems

For robots to cogitate like humans, they must excel in three meta areas: 

• Knowledge (rules, objects, metaknowledge): Useful for controlling and monitoring systems, such as data surveillance. Also beneficial for design features, such as innovating products or helping humans augment VLSI chips.
• Interpretation (inference, control, strategy): Robots have to acquire if-then probabilistic thinking, which leads to inference. The AI skill predicts the future benefits industries like the economy, politics, meteorology, finance, and more. 
• Facts (data, solutions): The ability to remember, act on, and provide insights leading to diagnostic capabilities used in fields like medicine and machine failure.

AI Reasoning Mechanisms

Scientists developed various models that simulate human decision-making, reasoning, and cognition and programmed with these models. The most important ones are:

• Approximate and Fuzzy Reasoning: Fuzzy logic that approximates our rough and imperfect way of decision-making. Such a model enables us to include unpredictable elements that accompany everyday life.
• Model-Based Reasoning: These apply the psychological processes used when making if-then inferences from a given set of premises.
• Constraint-Based Reasoning: Scientists use algorithms to insert limits in the number of items fed to machines to constrain their output.

AI Stages

The complexity of AI progresses from our most common AI to an abstract futuristic level.

• Reactive AI:  Machines react to stimuli. They absorb external information and respond accordingly.
• Limited Memory AI: Machines have limited memory of their programmed data. They act on that memory.
• Theory of Mind AI: Machines interact in a social context, responding to queries and guiding humans. They lack the capacities of imagination as well as understandings of human emotions and context.
• Self-Aware AI: These robots have the same sentient emotions, consciousness, and imagination as humans.

AI Applications

Below are some key ways industries use AI:

• Real-time Recommendations: Business websites create individualized recommendations for customers based on their browsing activity, among other factors.
• Virus and Spam Prevention: Virus and spam detection software uses AI to detect new viruses and spam.
• Automated Stock Trading:  Using AI, robo-advisers crunch millions of data points and recommend which stock to invest in types.
• Household Robots: These are robots that help you with all aspects of your house (and business), from making beds to cooking and cleaning.
• Autopilot Technology: Advanced AI equips self-driving cars with cameras, radar, and LiDAR sensors.

AI Ethics

Since AI-powered machines can enormously improve or harm our lives, most governments impose regulations for trustworthy AI strategies. Obligations include:

• Data Security: Enterprises must ensure and demonstrate that data is collected, used, managed, and stored safely and responsibly.
• Bias in Decision-Making: Computer models cannot think objectively because subjective humans program them. Results lead to discrimination in areas such as criminal justice, health care, and hiring.

• The Threat to Human Dignity: Machines’ evolution does not displace humans nor alienate or devalue them.

• Machine Ethics (or machine morality): This is the field that occupies itself with designing Artificial Moral Agents (AMAs), or robots that are programmed to behave morally with the injunction not to harm.

Extra Credit

• Journal of Artificial Intelligence Research
• Cognitive Science
• AI Magazine
• IEEE Computational Intelligence Magazine

Let's Connect

There is so much more to discuss, so connect with us and share your Google Cloud story. You might get featured in the next installment! Get in touch with Content Manager Sabina Bhasin at sabina.bhasin@c2cgobal.com if you’re interested.

Rather chat with your peers? Join our C2C Connect chat rooms!