When I was asked to do a keynote on AI in 2023 after ChatGPT just took off, there were a lot of words being thrown around. And as all words, some had a shared meaning to people, and so they were used. Other were used more as buzz words, they sounded cool, or people thought they made them sound smart. There also was a lot of confusion as to what was what.
People were thinking — Is AI just smart computers that asnwer questiosn quick than humans and robitic arms? What is the relationships between AI and now this Machine Learning term? Are AI and ML the same thing? Yes, no? What is this thing Deep Learning? I was just figuring out ML. What’s the difference? What are these large language models that everyone is talking about when referring to Chat GPT? Is Chat GPT AI? machine learning? deep learning? neural network? llm?
Artificial intelligence has moved from academic thought experiment to everyday tool in a remarkably short time. But the vocabulary around it – AI, machine learning, deep learning, neural networks – gets used interchangeably which muddies it even more, so I pulled some resources back to the beginning to start putting together some basics for my lectures, workshops, and modeuls. I pulled some key resoures relevant to my work, and I will walk through the core concepts, how they nest inside one another, and where the field stands today. I also, later, discuss how these relate to communication and learning theories.
Back to the original question to get us started — Is Chat GPT machine learning? deep learning? neural network? llm?
Well, yes to all. ChatGPT is an LLM built on deep learning which is a type of neural network, powered by machine learning. People love to describe these at layers likes a Russian nesting doll (I recently bought one at auction; super cute). These terms are not different choices; they are just layers moving broad to specific or the exact technology.
The Nesting Dolls of AI
The simplest way to understand AI, machine learning, deep learning, and neural networks is to picture them as nested systems, largest to smallest, each containing the next.
- Artificial intelligence is the broadest category: machines performing tasks that would normally require human intelligence.
- Machine learning is a subset of AI that can adapt automatically with minimal human interference.
- Deep learning is a subset of machine learning that uses artificial neural networks to mimic how the human brain learns.
- Neural networks are the layered structures of interconnected nodes that make deep learning possible.
Keeping this hierarchy in mind makes the rest of the terminology far easier to navigate.
Where the Conversation Began: Alan Turing
People talk alot about Alan Turing and the Turing Test, so this is a must have piece of knowledge. AI is often linked back to Alan Turing’s 1950 paper, Computing Machinery and Intelligence. Turing, frequently called the “father of computer science,” posed a deceptively simple question: “Can machines think?”
This was actually before the Dartmouth retreat, which I will talk about a bit later. However, I think it is important to note, like with everything new, while everyone tries to and does pinpoint that one moment or event that defined a field, there were other moments, events, and scholars thinking about this same phenomenon and having just as great, if not a greater impact on the advancement of the innovation. I’m not saying that Turing isn’t the guy of AI. I am just saying a lot of people were exploring what “machines” could do that would help people. I’ll talk more about my guy Shannon in a later post.
From Turning’s proposed question came the famous Turing Test, in which a human interrogator tries to distinguish between a computer’s text responses and a human’s. The test has drawn plenty of scrutiny in the decades since, but it remains a landmark in AI history and a living concept in philosophy, drawing as it does on ideas from linguistics.
Defining AI: No Single Answer
There is no universally accepted definition of artificial intelligence. Different thinkers frame it differently:
- John McCarthy (Stanford, 2004) described AI as “the science and engineering of making intelligent machines,” noting that AI need not restrict itself to biologically observable methods.
- Stuart Russell and Peter Norvig, in their widely used textbook Artificial Intelligence: A Modern Approach, center the idea of the intelligent agent—an entity that receives percepts from its environment and takes actions, mapping perceptions to responses.
- Andrew Ng offers a concise version: AI is “the ability of machines to perform tasks that would normally require human intelligence.” He credits AI’s recent surge to the explosion of available data and advances in computing power.
The White House has noted that some define AI loosely as any system exhibiting behavior thought to require intelligence, while others define it as a system that rationally solves complex problems or acts to achieve goals in real-world conditions.
A Common Taxonomy
One popular framework sorts AI into four types:
- Systems that think like humans — cognitive architectures and neural networks
- Systems that act like humans — passing the Turing Test through natural language processing, knowledge representation, and learning
- Systems that think rationally — logic solvers, inference, and optimization
- Systems that act rationally — intelligent software agents and robots that perceive, plan, reason, and act
Other thinkers add their own lenses. Venture capitalist Frank Chen breaks the field into five problem areas: logical reasoning, knowledge representation, planning and navigation, natural language processing, and perception. Researcher Pedro Domingos groups AI researchers into five “tribes”—symbolists, connectionists, evolutionaries, Bayesians, and analogizers—based on the methods they favor.
The Three Capability Levels
In tech circles, AI is often classified by capability:
- Artificial Narrow Intelligence (ANI) — every AI that exists today, including the most advanced systems ever built. It excels at specific tasks but nothing beyond them.
- Artificial General Intelligence (AGI) — a hypothetical system matching human-level ability across domains.
- Artificial Superintelligence (ASI) — a hypothetical system surpassing human intelligence entirely.
A related classification describes AI by how it handles memory and experience:
- Reactive machines — the most basic systems, unable to form memories or draw on past experience.
- Limited memory — systems that learn from historical data to make decisions. Nearly every AI in use today, from chatbots to self-driving cars, falls here. An image-recognition model, for instance, trains on thousands of labeled pictures and then draws on that stored experience to identify new images with growing accuracy.
- Theory of mind and self-aware AI — still theoretical stages, involving machines that understand emotions and, eventually, possess consciousness.
Machine Learning: Learning from Data
At its core, machine learning is an automated process that extracts patterns from data. Supervised machine learning, in particular, builds predictive models by learning the relationship between descriptive features (inputs) and target features (outputs) from historical examples.
The workflow generally involves:
- Data exploration — examining data quality, missing values, and outliers
- Model search — testing many models to find the one that best captures the relationship in the data
- Learning types — including information-based learning (such as decision trees), similarity-based learning, and probability-based learning
A guiding principle: the more descriptive data available, the better, because richer data offers a more complete picture of the problem.
Deep Learning and Neural Networks
Deep learning is an evolution of machine learning that layers algorithms and computing units, neurons, into an artificial neural network. These networks take inspiration from the structure of the human brain, passing data through webs of interconnected algorithms in a non-linear fashion, much the way our brains process information.
A neural network, then, is a machine learning method that teaches computers to process data in a brain-inspired way, using interconnected nodes arranged in layers. Because they can model complex, non-linear relationships between inputs and outputs, neural networks let computers make intelligent decisions with limited human assistance.
Natural Language Processing
Natural language processing (NLP) is the branch of AI focused on giving computers the ability to understand text and spoken language much as people do. It relies on machine learning to analyze text and speech data, and it powers many tools we use daily: spam detection, machine translation, virtual assistants like Siri and Alexa, social media sentiment analysis, and automatic summarization.
Under the hood, NLP breaks language down into a series of tasks:
- Speech recognition — converting voice into text, a challenge complicated by accents, slang, and imperfect grammar
- Part-of-speech tagging — identifying whether “make,” for example, is a verb or a noun based on context
- Word sense disambiguation — choosing a word’s intended meaning, as in “make the grade” versus “make a bet”
- Named entity recognition — flagging useful entities, such as recognizing “Kentucky” as a location
- Co-reference resolution — determining when two words refer to the same thing, like matching “she” to “Mary”
- Sentiment analysis — extracting attitudes and emotions from text
- Natural language generation — the reverse of speech recognition, turning structured data into human language
Generative AI: Creating Something New
Generative AI represents the next evolution of the field. Sometimes called Strong AI or Creative AI, it can produce text, images, video, and other content. Tools like ChatGPT, Bard, and DALL-E are all examples.
The key difference from traditional AI is simple but profound: generative AI creates something new rather than merely classifying or predicting. These models are trained on vast quantities of existing content, learning the underlying patterns and probability distributions in the data. When given a prompt, they generate new outputs that follow those patterns. As with all machine learning, this depends on enormous volumes of training data.
Emotion AI: Machines That Read Feelings
A distinct branch known as emotion AI—also called affective computing or artificial emotional intelligence—measures, interprets, simulates, and responds to human emotions. The field dates back at least to 1995, when MIT Media Lab professor Rosalind Picard published Affective Computing.
The goal is more natural human-machine interaction. As MIT researcher Javier Hernandez puts it, humans constantly read one another’s faces and body language and adjust accordingly; a machine that can’t sense your emotional state is limited in how effectively it can communicate with you.
Building Trustworthy AI
Powerful models are only as good as the data and safeguards behind them. As IBM emphasizes, identifying the right datasets from the start is essential to using quality data effectively, and organizations increasingly need AI-ready architectures that can draw on data wherever it lives—mainframes, data centers, private and public clouds, and the edge.
Just as important, AI must be trustworthy. Models that mislead, carry bias, or hallucinate can damage a company’s reputation, invite regulatory penalties, and erode customer trust and privacy. The goal is AI that is explainable, fair, and transparent.
The Takeaway
Artificial intelligence is not one technology but a layered family of them, from the broad ambition of “thinking machines” down to the specific mechanics of a neural network. Understanding the vocabulary and the hierarchy behind it is the first step to thinking clearly about where the technology is now and where it may be headed.
Sources: IBM, Forbes, Coursera, MIT Sloan, The Conversation, Stanford University, and Andrew Ng’s coursework on artificial intelligence.
References
Sources consulted for Understanding Artificial Intelligence: From Turing to Generative AI.
Foundational Works
Turing, A. M. (1950). Computing Machinery and Intelligence. — Origin of the “Can machines think?” question and the Turing Test.
McCarthy, J. (2004). What Is Artificial Intelligence? Computer Science Department, Stanford University.
Russell, S., & Norvig, P. Artificial Intelligence: A Modern Approach. — Introduces the intelligent-agent framework.
Picard, R. (1995). Affective Computing. MIT Media Lab. — Foundational text for emotion AI.
Definitions, Types, and Taxonomies
IBM. Artificial Intelligence. https://www.ibm.com/topics/artificial-intelligence
Forbes. 7 Types of Artificial Intelligence (June 19, 2019). https://www.forbes.com/sites/cognitiveworld/2019/06/19/7-types-of-artificial-intelligence/
Hintze, A. Understanding the Four Types of AI, from Reactive Robots to Self-Aware Beings. The Conversation. https://theconversation.com/understanding-the-four-types-of-ai-from-reactive-robots-to-self-aware-beings-67616
Banafa, A. Narrow AI vs. General AI vs. Super AI. LinkedIn. https://www.linkedin.com/pulse/narrow-ai-vs-general-super-ahmed-banafa/
Andreessen Horowitz (a16z). AI, Deep Learning, and Machine Learning: A Primer. https://a16z.com/ai-deep-learning-and-machine-learning-a-primer/
Machine Learning, Deep Learning, and Neural Networks
Coursera. AI vs. Machine Learning vs. Deep Learning: A Beginner’s Guide. https://www.coursera.org/articles/ai-vs-deep-learning-vs-machine-learning-beginners-guide
AWS. What Is a Neural Network? https://aws.amazon.com/what-is/neural-network/
Htun, P. Z. Learning AI for Everybody: My Experience with Andrew Ng’s Course. LinkedIn. https://www.linkedin.com/pulse/learning-ai-everybody-my-experience-andrew-ngs-course-paing-zin-htun/
Natural Language Processing
IBM. Natural Language Processing. https://www.ibm.com/topics/natural-language-processing
Generative AI
U.S. Chamber of Commerce. Traditional AI vs. Generative AI. https://www.uschamber.com/co/run/technology/traditional-ai-vs-generative-ai
Investopedia. — On how generative models train on existing content to produce new outputs.
Emotion AI
Somers, M. Emotion AI, Explained. MIT Sloan. https://mitsloan.mit.edu/ideas-made-to-matter/emotion-ai-explained
Allerin. Artificial Emotional Intelligence: The Future of AI. https://www.allerin.com/blog/artificial-emotional-intelligence-the-future-of-ai
AI in Robotics
Intel. Artificial Intelligence in Robotics. https://www.intel.com/content/www/us/en/robotics/artificial-intelligence-robotics.html