Lecture 1
,Lecture 1
I. Course Overview and Interdisciplinary Approach
1. Structure and Scope
The course is fundamentally interdisciplinary. Although officially part of the law and
technology major, it is not a typical legal course. It introduces students to regulatory and
legal aspects, as well as sociological, philosophical, ethical, and critical computing studies.
The core idea is to obtain different views, theories, and ideas to study AI, focusing on AI not
merely as a technology but as a practice. This practice is implemented in everyday life,
organizations, companies, and education. It is essential to recognize that artificial
intelligence, as a new disruptive phenomenon, cannot be studied from a single viewpoint
but requires different disciplines and gazes.
2. Study Requirements
Active participation is required. Students are suggested to complete the required readings
before attending class or watching lectures. Although lectures are recorded, attendance is
recommended, partly because in-class group exercises help students practice for the exam.
All necessary texts are available on Canvas; there is no required handbook to purchase.
The course provides required readings (needed for the exam) and recommended readings,
which are there to expand scope and interests.
II. Conceptions and Reality of Artificial Intelligence
1. Portrayals in Popular Culture
When considering AI, people often visualize different things, such as computers or computer
code. However, in popular culture and the media, AI is portrayed in various ways:
● Holograms: Providing personalized advertisements (e.g., Blade Runner films, which
are recommended viewing due to their discussion of ethical issues).
● Voices: Such as the AI HAL in Stanley Kubrick’s 2001, a space odyssey, which
steers a spaceship (an old portrayal). Alexa is also a voice AI for the home. Voices
(and robots) are often female, which is striking.
● Robots: Such as Sophia, who obtained citizenship in Saudi Arabia.
● Vehicles: Self-driving cars with consciousness (e.g., the 1980s TV series Night
Rider).
● Other Examples: The 1980s film The Lift featured an elevator with conscious AI that
intentionally murdered people. Cameras embedded with AI for facial recognition and
behavioral recognition are also increasingly common in cities.
,In contemporary media, artificial intelligence is increasingly becoming almost synonymous
with generative AI (e.g., ChatGPT), leading to other types of AI being overlooked.
2. The Mundane Reality of AI
Despite the exciting and fancy portrayals in media and movies, AI is actually very mundane.
It is essentially computer code and statistics. Specifically, AI involves probabilistic
statistics. It is crucial to remember that the current reality is not science fiction, and much of
the perception of AI in media is driven by marketing, particularly concerning generative AI.
III. Applications and Problems of AI
1. Positive Applications
AI is used to solve real issues, not just for marketing. Examples of positive impacts include:
● Protein Folding: AI in the context of chemistry has achieved a serious scientific
breakthrough.
● Medicine and Diagnostics: Medical professionals are highly hopeful that AI will
assist in diagnostics and medicine.
● Non-Western Context: In Sudan, an AI app is being used to identify snakes and
improve bite treatment.
The successful and positive application of AI is highly dependent on the developers'
motivations: whether the AI is designed to solve problems or to create profits. When AI is
implemented within traditional structures, it often merely repeats pre-existing problems and
biases.
2. Problematic Applications
More problematic applications of AI include:
● Facial recognition technology.
● Error-ridden applications: An app providing AI-generated images of mushrooms
and indicating which ones are edible is "full of error," potentially leading to death if
trusted.
● Criminal Use: Criminals are increasingly using AI to commit crimes such as
cybercrime and phishing.
● Deepfakes: AI-generated content and deepfake content used in the context of abuse
are a real problem online, complicated by the difficulty of regulation.
IV. Defining AI: Legal and Disciplinary Perspectives
1. The Legal Definition (AI Act)
There are numerous definitions of AI. Law students are recommended to rely on the
definition found in the AI Act. This definition is a compromised definition resulting from
,many discussions and was narrowed down from earlier versions that included simple
statistical software.
The final definition is:
"A machinebased system that is designed to operate with varying levels of
autonomy and that can for explicit or implicit objectives generate outputs
such as predictions, recommendations or decisions that influence
physical or virtual environments."
The varying levels of autonomy is identified as the crucial element. However, the definition
is still considered quite vague, and its interpretation will likely be determined by future case
law and guidelines.
2. Disciplinary Viewpoints
Different disciplines approach AI differently, leading to varied definitions.
● Computer Science: AI is seen as an old branch of computer science focused on
the effort to make computers think—machines with minds.
Typologies (e.g., Russell & Norvik, 2009) categorize systems based on their
relationship to human intelligence, such as systems that think/act like humans or
systems that think/act rationally.
● Lawyers: Are primarily interested in how the law deals with AI, how it is defined, and
how the law can be enforced regarding risks.
● Social Scientists (Science and Technology Studies): Are more interested in the
social technical practice of AI: how it is implemented, the issues that arise, and the
interaction between technology and humans.
The question of whether AI should be defined broadly or narrowly creates different legal
implications. The definition in the AI Act was narrowed specifically because a broad
definition would have categorized too many things as AI, making it problematic.
V. Historical Context and Philosophical Debates
1. Early Conceptualizations
The ideas and development of AI are quite old, and claims of "extremely new technology"
are often part of a marketing strategy.
● Ada Lovelace (19th century): Worked with Charles Babbage, who created a pre-
computer calculating machine. Lovelace conceptualized the analytic engine and
speculated that the machine might act upon things besides numbers,
preconceptualizing the computer.
2. The Turing Test and Criticism
, ● Alan Turing: The most famous person in the history of AI. He is known for creating
the Enigma machine during WWII. His article, Can machines think, explored whether
machines could possess human intelligence and is recommended reading.
● Imitation Game (Turing Test): A test devised by Turing to determine if a machine
has intelligence. If an interrogator (Person C) cannot identify that A is a computer
(and B is a person) based on their responses, then A is considered intelligent.
● Eliza: The first chatbot, developed in 1966, functioned similarly to modern chatbots
and had a psychological therapist function.
● The Chinese Room: A classic philosophical theory proposed by John Searle in
1980. This thought experiment criticizes the Turing Test by arguing that although a
computer program can convince a human they are communicating with another
human, it only simulates knowledge. The machine, like ChatGPT, does not truly
understand or possess intelligence, even if it is convincing.
3. The Debate on Intelligence
Many individuals claim AI is intelligent and can reason like humans, especially regarding
chatbots. The impression that one is talking to a human is deliberately created, and chatbot
answers are often improved by humans.
However, scientific research, particularly from a computer science perspective, generally
concludes that AI is not intelligent; it is just predictions. AI possesses an
"uncomprehensible ability of predicting and generating text that comes from the data". It is
essential to be critical, as scientific claims require several studies to show the same results
for causality, unlike the claims often made in the media based on only one study.
VI. AI Terminology and Classifications
Based on European policy reports and developments, a distinction is often made between
predictive and generative AI.
Classification Definition and Examples/Context
Characteristics
Predictive AI Traditional AI, using machine Classification systems, automated
learning and statistical decision-making systems deployed in
analysis on historical data to the public sector (e.g., credit scoring,
identify patterns, forecast hiring). This is a continuum of
events, and anticipate algorithmic risk assessments. Note:
outcomes. It provides data- Fully automated decisions (e.g., early
driven decisions. parole in the US) are not allowed in
the EU; using the algorithmic decision
to help human decision-making
(human in the loop) is allowed.
, Generative AI AI systems used to create or ChatGPT, Grock, DALL-E (for
generate media such as text, pictures), Midjourney, and newer
images, sound, or video. developments involving video creation.
Models are trained on large
data sets to learn patterns and
structures, then generating
new synthetic content.
Agentic AI Uses a digital ecosystem Self-driving cars and virtual assistants.
(LLMs, ML, NLP) to perform
autonomous tasks on behalf
of the user. There is no
human in the loop checking
the decision.
General A term from the AI Act. Generative AI falls under this
Purpose AI Intended by the provider to classification, but GPAIS is broader.
System perform generally applicable Used in the AI Act for systems that are
(GPAIS) functions (image/speech generative but not high risk.
recognition, translation,
audio/video generation,
pattern detection, question
answering). It can be used in
multiple contexts and
integrated into other AI
systems.
Artificial The concept of singularity— Requirements for AGI include the
General reaching human-level ability to solve puzzles, make strategy,
Intelligence intelligence. AGI requires the represent knowledge, plan, learn,
(AGI) / Strong AI to independently reason. communicate in natural language, and
AI integrate all these skills to solve
problems.
VII. The AI Hype and Critical Studies
1. The Hype Phenomenon (Singularity)
The philosophical theory of singularity addresses the future point at which AI exceeds
human intelligence and subsequently makes itself rapidly smarter, reaching a superhuman
level. Prominent voices, including Elon Musk, Stephen Hawking, Ray Kurzweil, and Sam
Altman (Open AI), claim that this will happen.
The widespread proclamation that humanity is close to building digital super intelligence is
known as the AI hype.
2. Criticism of the Hype
,Lecture 1
I. Course Overview and Interdisciplinary Approach
1. Structure and Scope
The course is fundamentally interdisciplinary. Although officially part of the law and
technology major, it is not a typical legal course. It introduces students to regulatory and
legal aspects, as well as sociological, philosophical, ethical, and critical computing studies.
The core idea is to obtain different views, theories, and ideas to study AI, focusing on AI not
merely as a technology but as a practice. This practice is implemented in everyday life,
organizations, companies, and education. It is essential to recognize that artificial
intelligence, as a new disruptive phenomenon, cannot be studied from a single viewpoint
but requires different disciplines and gazes.
2. Study Requirements
Active participation is required. Students are suggested to complete the required readings
before attending class or watching lectures. Although lectures are recorded, attendance is
recommended, partly because in-class group exercises help students practice for the exam.
All necessary texts are available on Canvas; there is no required handbook to purchase.
The course provides required readings (needed for the exam) and recommended readings,
which are there to expand scope and interests.
II. Conceptions and Reality of Artificial Intelligence
1. Portrayals in Popular Culture
When considering AI, people often visualize different things, such as computers or computer
code. However, in popular culture and the media, AI is portrayed in various ways:
● Holograms: Providing personalized advertisements (e.g., Blade Runner films, which
are recommended viewing due to their discussion of ethical issues).
● Voices: Such as the AI HAL in Stanley Kubrick’s 2001, a space odyssey, which
steers a spaceship (an old portrayal). Alexa is also a voice AI for the home. Voices
(and robots) are often female, which is striking.
● Robots: Such as Sophia, who obtained citizenship in Saudi Arabia.
● Vehicles: Self-driving cars with consciousness (e.g., the 1980s TV series Night
Rider).
● Other Examples: The 1980s film The Lift featured an elevator with conscious AI that
intentionally murdered people. Cameras embedded with AI for facial recognition and
behavioral recognition are also increasingly common in cities.
,In contemporary media, artificial intelligence is increasingly becoming almost synonymous
with generative AI (e.g., ChatGPT), leading to other types of AI being overlooked.
2. The Mundane Reality of AI
Despite the exciting and fancy portrayals in media and movies, AI is actually very mundane.
It is essentially computer code and statistics. Specifically, AI involves probabilistic
statistics. It is crucial to remember that the current reality is not science fiction, and much of
the perception of AI in media is driven by marketing, particularly concerning generative AI.
III. Applications and Problems of AI
1. Positive Applications
AI is used to solve real issues, not just for marketing. Examples of positive impacts include:
● Protein Folding: AI in the context of chemistry has achieved a serious scientific
breakthrough.
● Medicine and Diagnostics: Medical professionals are highly hopeful that AI will
assist in diagnostics and medicine.
● Non-Western Context: In Sudan, an AI app is being used to identify snakes and
improve bite treatment.
The successful and positive application of AI is highly dependent on the developers'
motivations: whether the AI is designed to solve problems or to create profits. When AI is
implemented within traditional structures, it often merely repeats pre-existing problems and
biases.
2. Problematic Applications
More problematic applications of AI include:
● Facial recognition technology.
● Error-ridden applications: An app providing AI-generated images of mushrooms
and indicating which ones are edible is "full of error," potentially leading to death if
trusted.
● Criminal Use: Criminals are increasingly using AI to commit crimes such as
cybercrime and phishing.
● Deepfakes: AI-generated content and deepfake content used in the context of abuse
are a real problem online, complicated by the difficulty of regulation.
IV. Defining AI: Legal and Disciplinary Perspectives
1. The Legal Definition (AI Act)
There are numerous definitions of AI. Law students are recommended to rely on the
definition found in the AI Act. This definition is a compromised definition resulting from
,many discussions and was narrowed down from earlier versions that included simple
statistical software.
The final definition is:
"A machinebased system that is designed to operate with varying levels of
autonomy and that can for explicit or implicit objectives generate outputs
such as predictions, recommendations or decisions that influence
physical or virtual environments."
The varying levels of autonomy is identified as the crucial element. However, the definition
is still considered quite vague, and its interpretation will likely be determined by future case
law and guidelines.
2. Disciplinary Viewpoints
Different disciplines approach AI differently, leading to varied definitions.
● Computer Science: AI is seen as an old branch of computer science focused on
the effort to make computers think—machines with minds.
Typologies (e.g., Russell & Norvik, 2009) categorize systems based on their
relationship to human intelligence, such as systems that think/act like humans or
systems that think/act rationally.
● Lawyers: Are primarily interested in how the law deals with AI, how it is defined, and
how the law can be enforced regarding risks.
● Social Scientists (Science and Technology Studies): Are more interested in the
social technical practice of AI: how it is implemented, the issues that arise, and the
interaction between technology and humans.
The question of whether AI should be defined broadly or narrowly creates different legal
implications. The definition in the AI Act was narrowed specifically because a broad
definition would have categorized too many things as AI, making it problematic.
V. Historical Context and Philosophical Debates
1. Early Conceptualizations
The ideas and development of AI are quite old, and claims of "extremely new technology"
are often part of a marketing strategy.
● Ada Lovelace (19th century): Worked with Charles Babbage, who created a pre-
computer calculating machine. Lovelace conceptualized the analytic engine and
speculated that the machine might act upon things besides numbers,
preconceptualizing the computer.
2. The Turing Test and Criticism
, ● Alan Turing: The most famous person in the history of AI. He is known for creating
the Enigma machine during WWII. His article, Can machines think, explored whether
machines could possess human intelligence and is recommended reading.
● Imitation Game (Turing Test): A test devised by Turing to determine if a machine
has intelligence. If an interrogator (Person C) cannot identify that A is a computer
(and B is a person) based on their responses, then A is considered intelligent.
● Eliza: The first chatbot, developed in 1966, functioned similarly to modern chatbots
and had a psychological therapist function.
● The Chinese Room: A classic philosophical theory proposed by John Searle in
1980. This thought experiment criticizes the Turing Test by arguing that although a
computer program can convince a human they are communicating with another
human, it only simulates knowledge. The machine, like ChatGPT, does not truly
understand or possess intelligence, even if it is convincing.
3. The Debate on Intelligence
Many individuals claim AI is intelligent and can reason like humans, especially regarding
chatbots. The impression that one is talking to a human is deliberately created, and chatbot
answers are often improved by humans.
However, scientific research, particularly from a computer science perspective, generally
concludes that AI is not intelligent; it is just predictions. AI possesses an
"uncomprehensible ability of predicting and generating text that comes from the data". It is
essential to be critical, as scientific claims require several studies to show the same results
for causality, unlike the claims often made in the media based on only one study.
VI. AI Terminology and Classifications
Based on European policy reports and developments, a distinction is often made between
predictive and generative AI.
Classification Definition and Examples/Context
Characteristics
Predictive AI Traditional AI, using machine Classification systems, automated
learning and statistical decision-making systems deployed in
analysis on historical data to the public sector (e.g., credit scoring,
identify patterns, forecast hiring). This is a continuum of
events, and anticipate algorithmic risk assessments. Note:
outcomes. It provides data- Fully automated decisions (e.g., early
driven decisions. parole in the US) are not allowed in
the EU; using the algorithmic decision
to help human decision-making
(human in the loop) is allowed.
, Generative AI AI systems used to create or ChatGPT, Grock, DALL-E (for
generate media such as text, pictures), Midjourney, and newer
images, sound, or video. developments involving video creation.
Models are trained on large
data sets to learn patterns and
structures, then generating
new synthetic content.
Agentic AI Uses a digital ecosystem Self-driving cars and virtual assistants.
(LLMs, ML, NLP) to perform
autonomous tasks on behalf
of the user. There is no
human in the loop checking
the decision.
General A term from the AI Act. Generative AI falls under this
Purpose AI Intended by the provider to classification, but GPAIS is broader.
System perform generally applicable Used in the AI Act for systems that are
(GPAIS) functions (image/speech generative but not high risk.
recognition, translation,
audio/video generation,
pattern detection, question
answering). It can be used in
multiple contexts and
integrated into other AI
systems.
Artificial The concept of singularity— Requirements for AGI include the
General reaching human-level ability to solve puzzles, make strategy,
Intelligence intelligence. AGI requires the represent knowledge, plan, learn,
(AGI) / Strong AI to independently reason. communicate in natural language, and
AI integrate all these skills to solve
problems.
VII. The AI Hype and Critical Studies
1. The Hype Phenomenon (Singularity)
The philosophical theory of singularity addresses the future point at which AI exceeds
human intelligence and subsequently makes itself rapidly smarter, reaching a superhuman
level. Prominent voices, including Elon Musk, Stephen Hawking, Ray Kurzweil, and Sam
Altman (Open AI), claim that this will happen.
The widespread proclamation that humanity is close to building digital super intelligence is
known as the AI hype.
2. Criticism of the Hype
