Artificial Intelligence Techniques

(ITAIA1B33)
 WEEK_4

•Lesson 7

•Predict Future Developments in AI:

• Discuss potential trends and advancements in AI
technology and applications.
Floridi, in the textbook, however, argues that “the best observables are provided by an analysis of
the nature of

The data used by AI to achieve its performance, and

The problems that AI may be expected to solve.”

Luciano Floridi, The Ethics of Artificial Intelligence Principles, Challenges, and Opportunities. OUP
Oxford. 2023, p31.
Potential trends and advancements in AI technology and applications

The rapid evolution of generative AI, comparing its development to the history of computers.

1. Generative AI's Rise: 2022 marked its explosion into public awareness, and by 2023, it began to be
integrated into the business world.

2. Comparison to Computers: The evolution of generative AI mirrors that of computers, moving from
massive mainframes to smaller, more efficient machines accessible to a broader audience.

3. Current Phase: Generative AI is now in its "hobbyist" phase, similar to early home computers, with
further advancements aiming for greater performance in smaller packages.

4. Open Source Models: 2023 saw the rise of efficient foundation models with open licenses, such as
Meta’s LlaMa, StableLM, Falcon, Mistral, and Llama 2. These models, enhanced by the opensource
community, can outperform many proprietary models.

5. Future Focus: The most impactful developments may be in governance, middleware, training
techniques, and data pipelines, making generative AI more trustworthy, sustainable, and accessible.
Trends and Advancements

Constant Evolution of AI
AI systems are becoming more capable, accessible, and impactful.

Generative AI
• Definition: AI that uses deep learning algorithms to create and enhance various types of
content, such as text, images, videos, and audio, using large and complex datasets.
• Impact: Transforms digital content creation and consumption.
• Examples: GPT3, DALLE (OpenAI).
Conversational AI
• Definition: A branch of AI that uses natural language processing and speech recognition to
enable natural and seamless interaction between humans and machines, using voice or text.
• Impact: Improves customer experience and provides personalized services.
• Examples: Chatbots, virtual assistants (Siri, Alexa, Cortana).
Autonomous AI
• Definition: AI that uses computer vision and robotics to enable machines to move and
interact with the physical world, without human supervision or intervention,
• Impact: Enhances efficiency and safety in various tasks.
• Examples: Selfdriving cars, drones, robots.
Explainable AI
• Definition: Aims to make AI systems more transparent, understandable, and
accountable, by providing explanations and justifications for their decisions and actions,
and by revealing their underlying logic and mechanisms.
• Impact: Builds trust and addresses ethical, social, and legal challenges.
• Focus: Fairness, accountability, transparency, privacy.
Democratized AI
• Definition: Making AI more accessible and inclusive, by providing affordable and user
friendly AI tools and platforms, that enable more people to create, use, and benefit from AI,
without requiring advanced technical skills or knowledge.
• Impact: Empowers individuals and fosters innovation.
• Examples: Nocode/lowcode platforms (e.g., Pictory).
Future Possibilities
• AI has the potential to transform science, technology, and society.
• Must be guided by principles of inclusion, equity, and human dignity.
Matching questions
Solutions

QUESTION 1 C
QUESTION 2 A
QUESTION 3 D
QUESTION 4 E
QUESTION 5 B
New AI Technologies
New AI Technologies

It is also predicted that the following new technologies will emerge to change
the AI landscape as well as the world we live in, even more:

Natural Language Generation

Creates human like text from data.

Speech Recognition
Converts spoken language into text.

Virtual Agents
AI powered assistants that interact with users.

Decision Management
AI systems that make decisions based on data.
Biometrics AI Optimized Hardware
Identifies individuals using physical characteristics. Specialized hardware for AI computations.

Machine Learning AI Generated Art

Algorithms that learn from data to make predictions. Creates art using AI algorithms.

Robotic Process Automation Generative AI Models

Automates repetitive tasks using software robots. Produces new content from existing data.

Peer to Peer Network Democratization of AI Technology

Decentralized network for sharing resources. Makes AI accessible to everyone.

Deep Learning Platforms No Code/Low Code Platforms

Advanced neural networks for complex tasks. Allows creation of applications without coding.
AI Powered Apps Transparency
Applications enhanced with AI capabilities. Clear and open AI processes.

AI Ethics and Fairness Cybersecurity

Ensures ethical use of AI. Protects systems from digital threats.

Bias Removal Edge Computing

Eliminates biases in AI systems. Processes data closer to its source.

Explainable AI Collaboration between Humans and AI

Makes AI decisions understandable. Enhances human capabilities with AI.
 Health & Medical

Auto & Road Safety

Agriculture

Entertainment

Impacted areas Service Industries

Military

Communications

Research and Development

Production

Logistics
Impacted data types
Impacted data types

Another area that will be affected greatly, are the different data types. That is because AI can process
and learn from the various types of data. Depending on which area above we are looking at, different
data types will be affected more or less. Here is a summary of the data types that may be affected the
most:
• Text

• Images

• Audio
Watch this movie in preparation for our next class

Family Buys a Robot To Do Housekeeping But Soon Starts To Develop Feelings (youtube.com)
https://sdgs.un.org/goals

Use Case for AI in the following sectors

1. Marketing
2. Government
3. Telecommunication
4. Education
5. Financial Services
6. Healthcare
7. Human Resources
8. Military
9. Transportation
10. Technological development
https://www.weforum.org/agenda/2021/06/ethical principles for ai/
Lesson 8
By the end of this lesson, you should be able to:

Predict Future Developments in AI:

Identify ethical and societal challenges that may arise with future AI developments.
Ethical and societal challenges because of future AI developments

We are certain to see issues arise from the use of AI and the usage of the data that allows
AI to function, as AI develops.

Specifically, from an ethical and societal perspectives.

It is important to note that there are several proposed ethical frameworks, for different
areas.

Floridi (in the prescribed textbook) was able to align all the proposed frameworks to the
framework being used in bioethics, which has 4 basic principles.

He then had to add one additional principle that is very AI specific and is required by both
the developers as well as the users of AI.
 Principles for a Framework of Ethics for AI

The five principles therefore form an ethical framework within which policies, best practices, and other
recommendations may be made
Beneficence: Promoting Well Being, Preserving Dignity, and Sustaining the Planet

Montreal & IEEE:

Montreal: AI should promote the well being of all sentient creatures.
IEEE: Prioritize human well being in all system designs.
AIUK & Asilomar:
AIUK: AI should be developed for the common good and benefit of humanity.
Asilomar: Similar emphasis on the common good.
Partnership on AI:
Ensure AI technologies benefit and empower as many people as possible.
European Group on Ethics (EGE):
Emphasizes human dignity and sustainability.
AI must ensure basic preconditions for life, continued prosperity, and a good environment for future
generations.
 Non maleficence in AI: AI systems should avoid causing harm. This includes preventing misuse and
mitigating risks
General Principle:
Distinct from beneficence; focuses on avoiding harm.
Privacy Concerns:
Prevention of personal privacy infringements emphasized in five of the six documents.
Misuse of AI:
Asilomar Principles: Warn against AI arms race and recursive self improvement.
Partnership on AI: Importance of AI operating within secure constraints.
IEEE: Need to avoid misuse.
Montreal Declaration: Developers should work against risks from their innovations.

Overall Principle:
Unclear whether the responsibility lies with AI developers or the technology itself.
Central issue: Autonomy and responsibility in preventing harm.
 Autonomy in AI: Promote human autonomy while restricting and making AI autonomy reversible.

General Principle:
Balance between human decision making and delegation to AI.
Risk: Growth in AI autonomy may undermine human autonomy.
Explicit Statements:
Montreal Declaration: AI should promote human autonomy.
EGE: Autonomous systems must not impair human freedom to set standards and norms.
AIUK: AI should not have the power to hurt, destroy, or deceive humans.
Asilomar: Humans should choose how and whether to delegate decisions to AI.
Meta Autonomy:
Concept of 'decide to delegate': Humans retain the power to decide which decisions to delegate.
Delegation should be overridable to protect or re establish human autonomy (e.g., pilot turning off
autopilot).
 Justice in AI: AI should promote fairness and equality, avoiding biases and ensuring equitable access
and treatment.
General Principle:
Addresses disparities in autonomy and promotes fairness.
Explicit Statements:
Montreal Declaration: AI should promote justice and eliminate discrimination.
Asilomar Principles: Emphasize shared benefit and shared prosperity from AI.
EGE: AI should contribute to global justice and equal access to benefits.
Warns against bias in AI training datasets.
Defends against threats to solidarity (e.g., social insurance, healthcare).
Other Meanings:
Use of AI to correct past wrongs (e.g., eliminating unfair discrimination, promoting diversity).
Preventing new threats to justice.

Diverse characterizations of justice highlight the complexity of AI's role in society.

Raises questions about whether humans are the recipients or controllers of AI's 'smart agency'.
Explicability in AI: Explicability is crucial for understanding and managing AI's impact on society.

General Principle:
Understanding and holding AI decision making processes to account.
Explicit Statements:
Asilomar & EGE: Emphasize "transparency".
IEEE: Emphasize both "transparency" and "accountability".
AIUK: Emphasize "intelligibility".
Partnership on AI: Emphasize "understandable and interpretable".
Novel Aspect of AI:
AI workings often invisible or unintelligible to most people.

Explicability Principle:
Combines epistemological "intelligibility" (how it works) and ethical "accountability" (who is responsible).
Complements other principles:
Beneficence & Non maleficence: Understand AI's impact on society.
Autonomy: Informed decisions about delegating to AI.
Justice: Knowing who to hold accountable for negative outcomes.
Ethical Benefits

AI offers several ethical benefits, which can be understood through its capabilities and specific
applications:

1. Analytical Prowess:
- AI can process and analyze vast amounts of data quickly and accurately, leading to better decision-
making and insights.

2. Consistency:
- AI systems can perform tasks with a high degree of consistency, reducing human error and ensuring
reliable outcomes.

3. Adaptation to Changing Inputs:

- AI can adapt to new information and changing conditions, making it flexible and responsive to
dynamic environments.

4. Freeing Humans from Tedious Tasks:

- AI can automate repetitive and mundane tasks, allowing humans to focus on more complex and
creative activities.
Specific Attempts at Using AI for Good:

- United Nations Sustainable Development Goals (SDGs):

- AI can help achieve SDGs by optimizing resource use, improving healthcare, enhancing education, and
promoting sustainable practices.

- Human Rights:
- AI can support human rights by monitoring and reporting violations, providing access to justice, and
ensuring fair treatment.
Ethical Benefits Measured Against Goals/Rights:

- Promoting Equality:
- AI can help reduce inequalities by providing equal access to services and opportunities.

- Enhancing Quality of Life:

- AI applications in healthcare, education, and other sectors can improve the quality of life for
individuals and communities.

- Sustainability:
- AI can contribute to environmental sustainability by optimizing energy use, reducing waste, and
supporting conservation efforts.

These benefits highlight how AI, when used ethically, can significantly contribute to societal well-being
and global development goals.
Ethical issues

3 types of ethical issues that are distinguishable:

Issues arising from Machine Learning

Issues arising from living in a digital world

Metaphysical issues/questions