Life Cycle of a Firm and Stakeholder Goals

Life Cycle of a Firm:

• The foundation phase entails choice of location and provision of initial capital.
• The operational phase involves performance, utilization, and potential restructuring.
• The liquidation phase includes insolvency and liquidation.

Economic, Social, and Ecological Goals:

• Economic goals focus on long-term profit maximization, profitability, and corporate growth, primarily benefiting equity providers.
• Social goals emphasize fair wages, job security, and motivating working conditions for employees.
• Ecological goals involve avoiding waste, recycling waste, and limiting harmful emissions, benefiting the general public.

Stakeholder Approach:

• Equity providers (owners, stockholders) claim an increase of invested capital and contribute equity capital to the company.
• Debt capital providers (lenders) claim repayment and interest on invested capital and contribute debt capital.
• Employees claim adequate wages, motivating working conditions, and job security. Employees contribute labor, executive activities.
• Management claims salary, power, influence, and prestige. Managment contributes labor, dispositive activities.
• Customers claim reasonably-priced, high-quality goods and supply the purchase of goods and services.
• Suppliers claim reliable payments and long-term supply relationships. Suppliers contribute the supply for goods and services.
• The general public claims tax payments, compliance with legal regulations, and respectful treatment of the environment. The general public contributes necessary infrastructure, legal order, and environmental assets.

Artificial Intelligence (AI) - Basic Definition:

• AI's fundamental principle is chain of prediction of the next word, learned from internet texts.
• AI is limited to the data it has been trained on and cannot imagine things outside of its horizon, similar to humans.
• AI can perform any task that can be learned.

Defining Intelligence:

• Intelligence is used when interacting with the environment.
• Intelligence enables success and profit relative to specific goals and objectives.
• Adaptability is critical for intelligence because it depends on the ability to adapt to different objectives and environments.
• Intelligence measures an agent's ability to achieve goals in a wide range of environments.

AI Types Based on Capability:

• Narrow AI is task-specific, performing a specific task using virtual assistants, recommendation algorithms, and facial recognition. It lacks general intelligence.
• General AI possesses human-level intelligence, capable of learning, reasoning, and problem-solving across various fields without pre-programming, but this currently doesn't exist.
• Superintelligent AI surpasses human intelligence in all aspects but is not yet in use raises ethical concerns around control and safety.

AI Functionality:

• Reactive machines have no memory, reacting only to specific inputs without using past experiences. Chess-playing AI is an example.
• Limited memory AI learns from past experiences and historical data but cannot store it permanently. Specialized and task-specific, it's rule-based as self-driving cars use data on traffic patterns.
• Theory of mind AI understands emotions, intentions, and human mental states, improving complex decision-making potentially as chatbots learning from past conversations but it isn't fully developed.
• Self-aware AI possesses hypothetical consciousness, self-awareness, and humanlike emotions, raising questions about its role in society but it does not exist, yet.

AI Approaches:

• Rule-based AI follows predefined rules and requires explicit instructions. This AI is limited in adaptibility, however early expert systems in finance are use if-then rules.
• Machine learning-based AI learns from data, improves, identifies patterns, and makes predictions/decisions without explicit programming, used in spam filters, recommendation systems.
• Deep learning-based AI uses artificial neural networks to learn from large datasets, effective for image recognition, for usage check AlphaGo from Google which is an image classification program.
• Natural language processing AI enables machines to understand, interpret, and respond to human language, as Chatbots and translation tools.

AI Learning Types:

• Supervised learning uses labelled datasets to train algorithms for classifying data and predicting outcomes, models learn over time, and are used for known results in spam detection and weather forecasting.
• Unsupervised learning uses machine learning to analyze and cluster unlabelled data, discovering hidden patterns, outputs need human validation, for detecting anomalies, using recommendation engines, and medical imaging.

Reinforcement Learning AI:

• Reinforcement Learning models learn constantly, minimizing errors in the long run. Various complex problem-solving models are possible with high accuracy.
• Reinforcement Learning is inconvenient for simpler problems, requires huge processing power and space, extensive data for accuracy, and high maintenance costs.

Types of Algorithms:

• Symbolic AI/Rule-Based AI encodes information in symbolic form, like facts, rules, and logical structures. Expert systems in medical diagnosis showcase this. Easy to interpret and explain, but not suited for complex, unfiltered data.
• Machine learning algorithms learn from data, predict patterns. They are Data-driven, not pre-programmed. Supervised Learning uses labeled examples whilst Unsupervised Learning finds hidden patters using any labeled outputs. Spam and Customer Filters can be used with Supervised learning Machine advantages include adaptability and handling large amounts of data, but needs lots of data and sometimes is difficult to provide interpretable data.
• Deep Learning is is a subset of machine learning, and uses Neural Networks to process data. Two types of Neural Netwoks are Convolutional and Recurrent. Convolutational specializes video or image analysis, were Image recognition and classification can occur. Recurrent specializes sequential data like natural language. Despite being able to preform difficult task, they are hard to interpret.

Alternative Types of AI Algorithms:

• Evolutionary Algorithms evolve solutions to optimization problems by generating and testing variations.
• Bayesian Inference uses probability theory for inferences under uncertainity. Graphical models representing probabilistic relationships, as well as Hidden Markov Models are part of this system. It an be used in email spam detection.
• Reinforcement learning has agents learn to make decisions by Interacting with an Enviornment by receiving feedback.

Components of Policy Gradient Methods:

• The method can directly learns a policy for action without utilizing a value function. These systems can be used in autonomous robots that learn to interact with environment through trial & error,
• Their abilities to learn is a unique advantage to their counterparts.
• Disadvantages include high use of trail and error, and long intensive training.

Functions of Goals in Computing:

• Cognitive computing aims to mimic human thought by simulating brain and decision-making. Systems take context into account when making decisions, as IBM Watson,
• Natural language processing can understand and create context from Human Language/

Capabilities and Drawbacks of combining approaches in AI:

• Hybrid approaches combine techniques, with the advantageous use-case of creating systems with multiple Al, for uses cases like in robotics.
• Potential disadvantages also arise, such as its reliance on computing resources , and its complexity to design and implement

Real world applications of AI:

• In the instance of Car insuranace ckaisms, services like Ping An is capable of retreiving Informatiom by voice recognition, as well as using video to assess vehicle danage
• In health car, Ping an goof offers both consultation and medicine delivery
• AI also has agriculture with farm data collection.

Ethical considerations when using AI:

• Historical biases and societal inequalities in data may be present when systems are being trained. There are also privacy concerns with the collection and use of use data.
• Transparancy and accountability concerns arise as the system make complex judgements. This is especially concerning in the Justice System.
• Job displacemebt may occur as tasks become automated.