DCS-T-709-AIES-2024 Main
Link to Lecture Notes

INTRODUCTION

Concepts

Artificial Intelligence	The scientific pursuit of making machines with intelligence. The products (software, machines, ideas) that comes out of AI research.
Data	Result of a measurement.
Information	Measurement.
Correlation	Relation between two or more measurements of information.
Knowledge	Useful information. Information that can be used for various purposes.
Agency	The act of using knowledge to get stuff done.
Agent	The physical locus/embodiment of agency. A system that can sense and act in an environment to do tasks.
Causation	A relation between entities in the physical world that allows us to achieve goals and predict events.
Perception / Percept	A process (perception) and its product (percept) that is part of the cognitive apparatus of intelligent systems. It feeds on measurements.
Goal	A specification of a world sub-state. The resulting state after a successful change.
Task	A problem that is assigned to be solved by an agent.
Environment	The constraints that may interfere with achieving a goal.
Plan	The partial set of actions that an agent assumes will achieve the goal.
Planning	The act of generating a plan.

SCIENTIFIC CONCEPTS

Correlation, Knowledge, Causation

Correlation	Some factors/variables co-vary when changes in one variable are related with changes in the other, negative or positive
Correlation: Powerful tool	Any variables in the world can be measured for correlation. Only two variables are needed (independent and dependent) for doing correlation studies
Main operating principle behind correlation	There is no causation without correlation
Correlation: Pitfall	Correlation does not imply causation between the variables measured!
Quasi-experimental designs	Purpose: Where true experimental design is not possible, approximate it. If direct control over dependent/independent variables is not possible.
How it works	1. One-shot case study (no control group) 2. Single group pre- and post-test (minimal control) 3. ABAB: Single-group repeated measures (slightly less minimal control)
Limitations	Much greater uncertainty as to the internal and external validity of the quasi-experiments than true experimental designs

Empirical Science (or just 'science')

Empiricism	The claim that “the world is its own best model” - that is, no matter what our preconceived notions of how the world works are, the world is the ultimate judge of that.
'Empiricism' the term	The term comes from Greek ('empeiria'), meaning “experience”. Descartes put it best when asking the question “How can I be sure that I exist?” when he answered with the no-famous phrase “I think, therefore I am”.
Empiricism in science	Modern science takes the idea of empiricism to its ultimate conclusion: Let's find the best way to create the most reliable way of creating trustworthy knowledge about the world.
Comparative experiments	…are the key form that this creation has taken, but in fact many other forms of reliable knowledge creation, besides controlled comparative experiments, are in use. The key to them is the systematic categorization and estimation of the uncertainty of the information they produce. By knowing these, the trustworthiness of the knowledge produced can be assessed, and appropriate action taken when the knowledge is used.
The Content of Scientific Knowledge	…is essentially rules about the causal behaviors and relations of things. Causality is a way to extract compact knowledge about any complex system that contains regularities. It works for the physical world because the physical world is highly regular.

The Scientific Method: Classical Description

Identification, description and formalization of phenomenon	1. Observation and description of a phenomenon or group of phenomena.
Hypothesis, null-hypothesis	2. Formulation of an hypothesis to explain the phenomena. In physics, the hypothesis often takes the form of a causal mechanism or a mathematical relation.
Creation of experimental setup to test hypothesis	3. Use of the hypothesis to predict the existence of other phenomena, or to predict quantitatively the results of new observations.
Perform experiment, collect & analyze results	4. Performance of experimental tests of the predictions by several independent experimenters and properly performed experiments. Basic assumption: Repeatability — Can be repeated by anyone anywhere.
Repeatability requires structured framework	Detailed description, clear goals, clear (limited) scope, hence the formalities in their execution.
Key idea: Comparision	Baseline collected in same experimental setup without any other intervention by experimenter
Key way of comparing	Empirical experiments.
Bottom line	Scientific research is a slow and expensive process. But it's the best one we've got (so far). And it's completely worth it.

Scientific Research Concepts / Definitions

Theory (isl. kenning)	“A set of statements or principles devised to explain a group of facts or phenomena, especially one that has been repeatedly tested or is widely accepted and can be used to make predictions about natural phenomena.” REF A theory is a relatively big explanation, covering several phenomena, often through a single principle, or a set of simple principles.
Hypothesis (isl. tilgáta)	Is a prediction about the relationship between a limited set of phenomena, typically formulated as measurable variables, as explained by a particular theory.
Data	Typically “raw numbers” – only contain low-level semantics.
Information	Processed and prepared data. Data organized at more than one level of detail. “Data with a purpose.”

The Scientific Method: The Comparative Experiment (ísl. samanburðartilraun)

Identification, description and formalization of phenomenon	Observation and description of a phenomenon or group of phenomena.
Hypothesis, null-hypothesis	Formulation of an hypothesis to explain the phenomena. In physics, the hypothesis often takes the form of a causal mechanism or a mathematical relation. Null-hypothesis of a hypothesis is the claim that it is false - i.e. that some relationship that it proposes does not hold.
Creation of experimental setup to test hypothesis	Use of the hypothesis to predict the existence of other phenomena, or to predict quantitatively the results of new observations.
Performance of experiment, collection and analysis of results	Performance of experimental tests of the predictions by several independent experimenters and properly performed experiments. Basic assumption: Repeatability Can be repeated by anyone anywhere
Repeatability requires formal framework	Detailed description, clear goals, clear (limited) scope, hence the formalities in their execution
Key idea: Comparsion	Baseline collected in same experimental setup without any other intervention by experimenter

Causation

What it is	A relation between entities in the physical world that allows us to achieve goals and predict events.
Why it's important	The comparative experiment (“the scientific method”) is based on the assumption that such relations exist, i.e. that the world has regularities.
How it relates to logic	If causal relations are rules, and the world has regularity, then the world is rules-based. Reasoning is the method of following logic when working with rules. It means we can reason about the world.
Types of (basic) causal relations	A→B & A→C : A causes B and C. A→B→C : A causes B and B causes C [A+B]→C : A and B together cause C. A→C, B→C : Both A and B are sufficient to cause B.
Time and Causation	The temporal relation between cause and effect is strict on time: Effects cannot happen before causes.

Scientific Method: Independent of Topic

Phenomenon	The world is filled with “stuff”. Anything is a “thing” - even “nothing” is a thing (a concept in our minds, which is represented as neural patterns and potential for behavior). We can group any arbitrary collection of things and call it a phenomenon. Example: A rock. A mountain. A planet. (If I say that I want to study “thingamajigs” - something you've never heard of - I will first have to list some of the major ways in which thingamajigs can be identified. In fact, this is a good idea anyway, so as to be clear and consistent about what it is that one is studying.)
The scientific method is independent of topic…	One can study any phenomenon with the scientific method, including claims of telepathy; selection of topic is independent of method – there is nothing inherently “unscientific” about studying any subject. (Close-mindedness is, however, very unscientific.) In other words, given that science gets us the most reliable (“best”) knowledge to build on at any time, we should take it seriously. But not so seriously as to exclude the possibility that it's wrong. (Because in fact we already know that all scientific knowledge is wrong – i.e. every scientific theory to date has limits to its scope that we know of.)
How can we trust our knowledge?	The scientific method is a General Way of Producing Trustworthy Knowledge. It is independent of topic. Therefore, it can also be used for AI systems. (In fact, it can easily be argued that something very similar to the scientific method is happening when humans learn cumulatively – with a few caveats that we will carefully cover in this course.)

Theories of the Scientific Method (Philosophy of Science)

A scientific theory predicts	A good scientific theory can be used to predict (known and unknown) results.
A scientific theory is a spotlight	A good scientific theory tells us where to look for interesting things; the more detailed the theory the more detailed should be its predictions and the more narrow its spotlight (specific suggestions for new investigations).
A scientific theory can produce new hypotheses	A good scientific theory helps us do more experiments by being a source of hypothesis creation.
A scientific theory provides control	A good scientific theory gives us control over the phenomenon it addresses that we would otherwise not have.
A scientific theory explains (“tells a coherent story”)	A good scientific theory explains how data is related. The more completely and the more simply it explains things, the better the theory.
A scientific theory gives us the big picture	A good scientific theory relates together, in a coherent way, some part of the world – in general the bigger the part, the better the theory.
Occam's Razor	A good scientific theory cannot be simplified; it is the shortest and most accurate explanation of a phenomenon. Einstein is quoted as saying: “A theory should be as simple as possible, but not simpler”.
A scientific theory can be disproven	A scientific theory or hypothesis is a statement that is disprovable. To count as “scientific” a theory must be disprovable. For this there must exist some measures and actions that are possible (in theory, but better yet, practic) whose results would possibly - should the measurements come out a particular way - disprove the theory. Applying this criterion strictly means that all scientific theories to date have been disproven - i.e. proven incorrect. This is a feature of science (not a bug): Exposing the limits of our theories by demonstrating in which contexts they are incorrect allows us to come up with better theories.

Footnote: Fields of Research

Philosophy	A systematic investigation into any phenomenon. Fundamental motivation: Deepened understanding of our place in the universe. Fundamental driving principle: Human reasoning and creativity. Fundamental organizing principle: Schools of thought; methods of reasoning. When you hardly know anything about a phenomenon, yet insist on getting to the bottom of it, philosophizing gets you started.
Induction	The cognitive act of generalizing from experience. Example: Socrates is a man. Socrates died. Hence, all men will die.
Science	A systematic investigation into phenomena in the natural world susceptible to physical experimentation. Fundamental motivation: Reliable knowledge of the world. Fundamental driving principle: Induction. Fundamental organizing principle: Controlled comparative experiment. When you embark on improving your understanding of a phenomenon with measurable/quantifiable variables, through comparative experiments, you are applying the scientific method (“doing science”).
Engineering	Effort to construct things using relevant knowledge (often state-of-the-art scientific models/theories - see below), systematic methods, and relevant technology. Fundamental motivation: Control of human environment. Fundamental principle: Design. Fundamental organizing principle: Methodical application of known procedures and methods. When you embark on changing or improving any aspect of your environment, working towards the implementation of a well-defined end product, through an application of best known practices, you are doing engineering.
Technology	Tools and techniques for getting things done. Fundamental principle: Composition, design, engineering.
Deduction	The cognitive act of following preconceived rules to their inevitable implication. Example: All men are mortal. Socrates is a man. Hence, Socrates is mortal.
Mathematics	A systematic study of quantity, numbers, patterns, and their relationships. Fundamental principle: Deduction. When you embark on clarifying the behavior and nature of quantifiable domains, using axiomatic rules and proofs, you are doing mathematics.
Causal Relations	A relation between two entities that makes one predictable from the other. Examples: If I flip the light switch, the lights will shine. (Deduction) If I want the light to shine, I can flip the light switch. (How to achieve a goal)

INTELLIGENCE RESEARCH

Intelligence: A Natural Phenomenon

Intelligence	A phenomenon encountered in nature. Intelligence is a phenomenon with good examples in the natural world, but may have more forms than the examples from nature. The only one that everyone agrees on to call 'intelligent': humans.
Natural Intelligence	Intelligence as it appears in nature. Some kinds of animals are considered “intelligent”, or at least some behavior of some individuals of an animal species other than humans are deemed indicators of intelligence.
Cognitive Science	The study of natural intelligence, in particular human (and that found in nature).
Artificial Intelligence	The study of how to make intelligent machines.
Intelligent Machines	Systems created by humans intended to display (some - but not all?) features required of beings encountered in nature.
How to define 'intelligence'	Many definitions have been proposed. See e.g.: A Collection of Definitions of Intelligence by Legg & Hutter.
Definitions: a word of caution	We must be careful when it comes to definitions – for any complex system there is a world of difference between a decent definition and good accurate appropriate definition.
Related quote	Aaron Sloman says: “Some readers may hope for definitions of terms like information processing, mental process, consciousness, emotion, love. However, each of these denotes a large and ill-defined collection of capabilities or features. There is no definite collection of nec- essary or sufficient conditions (nor any disjunction of conjunctions) that can be used to define such terms.” (From Architectural Requirements for Human-like Agents Both Natural and Artificial by A. Sloman)

Researching Intelligence

What is “intelligence”?	It is important to know what you're studying and researching! …A researcher selects their methods from their (premature) understanding of the phenomenon they'd like to 'figure out'. Without a good working definition, questions, methods and approach may be flawed.
BUT: Beware of Premature Definitions	You cannot define something precisely until you understand it! Premature precise definitions may be much worse than loose definitions or even bad-but-rough definitions: You are very likely to end up researching something other than what you set out to research.
What Can We Do?	List the requirements. Even a partial list will go a long way towards helping steer the research. Engineers use requirements to guide their building of artifacts. If the artifact doesn't meet the requirements it is not a valid member of the category that was targeted. In science it is not customary to use requirements to guide research questions, but it works just the same (and equally well!): List the features of the phenomenon you are researching and group them into essential, important but non-essential, and other. Then use these to guide the kinds of questions you try to answer.
Before Requirements, Look At Examples	To get to a good list it may be necessary to explore the boundaries of your phenomenon.
Create a Working Definition	It's called a “working definition” because it is supposed to be subject to (asap) scrutiny and revision. A good working definition avoids the problem of entrenchment, which, in the worst case, may result in a whole field being re-defined around something that was supposed to be temporary. One great example of that: The Turing Test.
Two Key Ingredients	Well, this is a simplification, but it is a useful one. Intelligence can be thought of as being composed of classification and control. The former concept is all the rage in contemporary AI research, the latter sits by the wayside. No system with real intelligence can be built without both.

Key Concepts in Control

Sensor	A transducer that changes one type of energy to another type.
Transducer	A device that changes one type of energy to another, typically amplifying and/or dampening the energy in the process.
Actuator	A physical (or virtual) transduction mechanism that implements an action that a controller has committed to.
Control Connection	Predefined causal connection between a measured variable <m>v</m> and a controllable variable <m>v_c</m> where <m>v = f(v_c)</m>.
Mechanical Controller	Fuses control mechanism with measurement mechanism via mechanical coupling. Adaptation would require mechanical structure to change. Makes adaptation very difficult to implement.
Digital Controller	Separates the stages of measurement, analysis, and control. Makes adaptive control in machines feasible.
Feedback	For a variable <m>v</m>, information of its value at time <m>t_1</m> is transmitted back to the controller through a feedback mechanism as <m>v{prime}</m>, where <m>v{prime}(t) > v(t)</m> that is, there is a latency in the transmission, which is a function of the speed of transmission (encoding (measurement) time + transmission time + decoding (read-back) time).
Latency	A measure for the size of the difference between <m>v</m> and <m>v{prime}</m>.
Jitter	The change in Latency over time. Second-order latency.

Key Concepts in Classification

What it is	The separation of one “thing” from “another thing” through measurement.
Why it matters	No intelligence can be brought about without the ability to separate something from something else.
Background / Foreground	The 'central things' and 'contextual' or 'unimportant' things.
Dimensions	The number of variables with measurable values that can help separate one thing from another.
Classification Methods	AI is replete with classification methods. This is the bulk of AI research for the first 70 years of its history.
Examples	Artificial neural networks, decision trees, nearest neighbor, support vector machines, autocorrelation, …

Historical Concepts

AI	In 1956 there was a workshop at Dartmouth College in the US where many of the field's founding fathers agreed on the term to user for their field, and outlined various topics to be studied within the field.
GOFAI	“Good old-fashioned AI” is a term used nowadays to describe the first 20-30 years of research in the field.
Cybernetics	Going back to WWI the field of cybernetics claimed a scope that could easily be said to subsume AI. Many of the ideas associated with information technology came out of this melting pot, including ideas by von Neumann. However, cybernetics has since all but disappeared. Why?
GMI or AGI	“Artificial general intelligence” or “general machine intelligence”: What we call the machine that we hope to build that could potentially surpass human intelligence at some point in the future – a more holistic take on the phenomenon of intelligence than the present mainstream AI research would indicate. Will we succeed? Only time will tell.

AI is a Broad Field of Research

A Scientific Discipline	As an empirical scientific discipline, AI aims to figure out the general principles of how to implement the phenomenon of intelligence, including any subset thereof. Does empiricism mean there is no theory? Absolutely not! Physics is an empirical science, yet is the scientific field that has the most advanced, accomplished theoretical foundation. What would a proper theory of intelligence contain, if it existed? A general theory of intelligence would allow us to implement intelligence, or any subset thereof, successfully on first try, in a variety of formats meeting a variety of constraints.
An Engineering Discipline	What separates AI from e.g. cognitive science? AI explicitly aims at figuring out how to build machines that are intelligent. Note that CogSci may also build intelligent machines, but the goal is not the machine itself, or the principles of its construction, but rather its role as a tool to figure out how intelligence works. The outcome from such work is unlikely overlap much due to the difference in working constraints,* although in the long term the two are likely to converge – and the two different approaches should be able to help each other.
AI spans many fields	Psychology, mathematics and computation, neurology, philosophy.
Alternative View	Psychology, mathematics & computation, neurology, and philosophy all were sooner than AI to address concepts of high relevance to the study of intelligence.
Is AI a subfield of computer science?	Yes and No. Yes, because this is the field that has the best and most tools for studying it as a phenomenon. No, because the field does not address important concepts and features of intelligence.
Isn't intelligence 'almost solved'?	Short answer: No! If it's almost solved it's been “almost solved” for over 60 years. And yet we still don't have machines with real intelligence.
Should we fear AI?	Short answer: No! Or, no more than genetic engineering. The threat lies with humans, not with machines – human abuse of knowledge goes back to the stone age.
Is the Singularity near?	Short answer: Who's to say? Predictions are difficult, especially wrt the future.

* Like the difference between constructing brick walls to study the stability of rock formations in nature versus the engineering principles of building brick walls: If the principles are well understood (weight distribution and stability), you should be able to build walls out of many materials.

Reasoning

What is Reasoning?	A systematic way of thinking about implications.
How is it done?	Via processes that observe rules.
What are the main types?	Deduction: All men are mortal. Socrates is a man. Hence, Socrates is mortal Abduction: How did this come about? (Sherlock Holmes) Induction: What is the general rule? Analogy: 'This' is like 'that' (in 'this' way).
How is it used in science?	In empirical science to unearth the “rules of the universe”. In mathematics as axioms. In philosophy as a way to construct arguments. In computer science to write code.
Why does it work?	Because the world's behavior is/seems rule-based, some of which we have figured out to date (a very small part!).
How is Reasoning relevant to AI & Ethics?	To manage rules requires reasoning. Intelligence makes models of the rules of the physical world. Therefore, reasoning is a necessary part of any intelligence. AI seeks to create machines that contain intelligence, thus AI must seek methods for machines that reason. Ethics can only make sense if it can be encoded in rules.

Models in Science, Engineering & Math

What it is	A model is a “cartoon” of a phenomenon – an information structure that captures the most important (preferably all the important) aspects of a phenomenon in question.
The universe: Nothing is given	How do we know that the sun will come up tomorrow? What evidence do we have? Can we prove it mathematically that the sun will come up tomorrow? The only thing we know for sure is that we can perceive things in the world and that “I am here now”. (This principle is most famously captured by Rene Descartes who wrote “I think, therefore I am”.) But since that perception is provided/generated by the same universe that we want to claim “exists” through those senses, using those grey cells, we cannot possibly know for sure what that really is, and hence whether it can be trusted. Therefore, the universe is (and cannot be anything but) non-axiomatic.
All scientific theories seek a model	No matter how explicit or implicit, all scientific theories are models of the world. Good scientific theories tie a lot of knots, leaving few loose ends. Best known example: E=mc^2
Science vs. Mathematics	Mathematics is axiomatic: Some a-priori premises are (and must be) known and complete. In contrast, science is non-axiomatic: We do not know the full set of rules that govern the universe, and we will never know.
Science vs. Engineering	Science sees an undifferentiated world and looks for the model. In engineering the model is given; the world is modified to behave like the model.
Science + Math	We strive to make scientific theories (models of the world) mathematical because of the compactness, precision, and specificity this can give us. However, it is not guaranteed solely through the use of math because a theory must detail how it maps to the thing it is a model of. If this is not done properly the math provides no benefits. Mapping a model to its reference: A good scientist does it properly; a bad scientist does it sloppily; the wannabe ignores it happily. Bottom line: Being mathematical is no guarantee for good science - it is neither necessary nor sufficient.
Science + Engineering + Math: The Holy Trinity	The three fields so defined support each other: Building better scientific models helps us engineer better; engineering better helps us build new tools for doing science better. Both are bootstrapped by philosophy and clarified through math.
Computer Science	A creative mix of empirical science, engineering and mathematics. Direct testing of applications and programs; user studies. Models and simulations. Logical and mathematical proofs.

Footnote: Terminology

Terminology is important!	The terms we use for phenomena must be shared to work as an effective means of communication. Obsessing about the definition of terms is a good thing!
Beware of Definitions!	Obsessing over precise, definitive definitions of terms should not extend to the phenomena that the research targets: These are by definition not well understood. It is impossible to define something that is not understood! So beware of those who insist on such things.
Overloaded Terms	Many key terms in AI tend to be overloaded. Others are very unclear. Examples of the latter include: intelligence, agent, concept, thought. Many terms have multiple meanings, e.g. reasoning, learning, complexity, generality, task, solution, proof. Yet others are both unclear and polysemous, e.g. consciousness. One source of the multiple meanings for terms is the tendency, in the beginning of a new research field, for founders to use common terms that originally refer to general concepts in nature, and which they intend to study, about the results of their own work. As time passes those concepts then begin to reference work done in the field, instead of their counterpart in nature. Examples include reinforcement learning (originally studied by Pavlov, Skinner, and others in psychology and biology), machine learning (learning in nature different from 'machine learning' in many ways), neural nets (artificial neural nets bear almost no relation to biological neural networks). Needless to say, this regularly makes for some lively but more or less pointless debates on many subjects within the field of AI (and others, in fact, but especially AI).

EOF

Center for Analysis and Design of Intelligent Agents

Table of Contents