Financial Knowledge and Skills for Business: Week 11

Investment Appraisal/Capital Budgeting

• Meaning: Process decision making respect investments fixed assets (non-current assets) that is, should proposed project accepted/rejected.
• Easier evaluate profitable projects, than identify suitable projects + collect necessary data.
• Large amounts resources often involved.
• Often difficult +/ expensive bail out an investment once undertaken.

Cash based Appraisal Techniques

• Payback
• NPV
• IRR

Alternative Appraisal Techniques

• Discounted Payback
• Sustainable NPV
• MIRR

Investment Appraisal/Capital Budgteing Part 2

Sophisticated Techniques

• Net present value 
• Internal rate return
• Modified internal rate return
• Profitability index
• Real options

Non-Sophisticated Techniques

• Payback
• Disocunted payback
• Accounting rate return 
• Return investments
• Return equity

Payback

• Meaning: Number years needed recover initial cash outlay capital budgeting project.
• Decision Rule: Project feasible/desirable if payback period less than/equal firm's maximum desired payback period.
• General, shorter payback period preferred while comparing 2 projects.

Net Present Value

• Meaning: NPV equal present value all future free cash flows less investment's initial outlay.
• Measures net value project today's currency.
• Equation: NPV = PF of Benefits - PV of Costs

Internal Rate of Return 

• Meaning: IRR discount rate equates present value project's future net cash flows project's initial cash outlay.
• Decision Rule: IRR + than Required Rate Return = accept + IRR - than RRR = reject.

Categories of AI

• AI divided several categories, although some overlap between them +  boundaries between categories not always clear.
• Rule-based: Follows set explicit rules/Algorithms solve problems, e.g. Medical diagnosis.
• Expert system: Designed mimic decision-making abilities human expert specific field.
• Machine learning: Involves training computer recognise patterns data + make decisions based patterns.
• Natural language processing: Focused enabling computer understand, interpret + generate human language.
• Neural network: Inspired structure + function human brain, typically used tasks, e.g. image + speech recognition.

Examples of Different Categories of AI

Expert Systems

• Medical diagnosis
• Financial decision-making
• Legal advice
• Technical support

Machine Learning

• Image recognition 
• Speech recognition - voice text
• Language translation
• Fraud detection

Natural Language Processing 

• Voice assistants - Siri, Alexa, Bixby
• Language translation
• Text analysis
• Dialogue systems

Neural Network

• Language translation
• Fraud detection 
• Recommendation systems
• Autonomous vehicles

Expert Systems and Rule-Based Systems

• Use combination AI + expert knowledge reason + make decisions.
• Rules developed using knowledge human experts + encoded into Knowledge Base.
• Facts about domain stored Facts Database.
• Inference Engine uses rules + facts draw conclusions.
• Feature ES ability explain reasoning.

Key Benefits

• Better decision quality 
• Cost reduction 
• Consistent solutions
• Speed
• Relaibility

Expert and Rule-Based Reasoning 

• These systems strict boundaries + used e.g. very specific areas medical diagnosis, finance, human resources, etc.
• Rules typically of If-Then type.

Forward Chaining

• Method starts set known facts + works way towards conclusion.
• Might start fact that patient fever + cough, use rules conclude patient has respiratory infection.

Backward Chaining

• Starts desired goal/conclusion, works backward determine necessary steps/conditions must met order reach goal.

(Artificial) Neural Networks

• Inspired biological neural networks constitute animal brains.
• ANNs consist layers interconnected 'neurons', process + transmit info.
• Connections between neurons modified, allowing network 'learn'.
• Particularly useful solving problems difficult solve using tradtional methods.
• Convolutional neural networks (CNNs) - image + video recognition.
• Recurrent neurla networks (RNNs) - natural language processing + speech recognition.
• Feedforward neural networks - function approximation, regression + classification.
• ChatGPT uses engine called GPT-3 + type RNN.
• Midjourney creates images text decriptions + CNN.

ChatGPT (Generative Pre-training Transformer)

• PArt OpenAI (AI research + development company).
• Trained dataset billions words web text. Text publically available (e.g. Wikipedia, Project Gutenberg).
• GPT language model trained predict next word sequence words, given context words come before it.
• Opened public (beta) late 2022.
• Seemingly provide answers any type instruction, e.g. programming, poetry, text descriptions + essays.

ChatGPT extended...

• God in a Box
• Uses API connect ChatGPT WhatsApp
• Messaging phone number/continuing conversation with instruction provide result.