AI Awareness Unlocked

• Separating fact from fiction: Common AI myths and why they persist

• Reality check: capabilities, limitations and evidence in the selected examples.

• Hype cycle analysis: Understanding where we are in AI adoption

• Historical context: How we got from chess computers to conversational AI

• Setting realistic expectations: Avoiding both over-optimism and unnecessary fear

• Narrow/task-focused AI: recommendations, voice assistance and navigation.
• General intelligence as a contested concept: definitions and uncertainty, not a consensus timeline.
• Agentic systems: planning and tool use within configured constraints; agentic behaviour does not by itself establish general intelligence.

Machine Learning fundamentals: Teaching computers to learn from patterns

• The three pillars: supervised, unsupervised, and reinforcement learning
• Data quality, coverage and quantity: why all matter for reliable modelling.

Deep Learning architectures: The brain-inspired approach

• Neural networks explained through everyday analogies
• What multiple layers enable, and what depth alone does not guarantee.

Transformer models revolution: The technology behind ChatGPT and beyond

• Attention mechanisms and their role in transformer models.
• From language to vision to multimodal applications

The training process: From raw data to intelligent systems

• Data pipelines and preprocessing essentials
• Training cycles, validation, and continuous improvement

Why AI decisions can be mysterious: Understanding opacity in complex systems

• The trade-off between performance and interpretability
• When black boxes become problematic in business contexts

Explainable AI (XAI) approaches: Opening the black box

• Techniques for understanding AI decision-making
• Transparency and risk-management frameworks, including voluntary NIST guidance; distinguish standards from legal requirements.

Emergent properties phenomenon: When AI surprises its creators

• Unexpected capabilities that weren't explicitly programmed
• Illustrative unexpected behaviours and implications for evaluation and deployment.

• Illustrative uses in healthcare and drug discovery, finance and fraud detection, manufacturing and maintenance, creative work, and transport/logistics.
• Distinguish a research example, pilot, deployed system and evidence of success; examine cautionary scenarios.
• Agentic workflows and integration challenges.
• AI-assisted cybersecurity and response automation: useful capabilities and failure modes.
• Local/edge AI for mobile and IoT: resource, privacy and performance trade-offs.
• Responsible adoption and changing regulation; investment claims require current evidence.

• Commercial hosted model ecosystems: OpenAI, Anthropic, Google and xAI as examples, not a version ranking.
• Downloadable models and families such as Llama, DeepSeek, Mistral, Falcon, Gemma and Phi: inspect the specific model card, licence and supported use.
• Distinguish open source, open weights, hosted access and custom licence restrictions.
• Compare total cost, customisation, control, infrastructure, data handling and sector requirements.

• Hugging Face model/dataset discovery, LangChain workflows and model-serving tools such as OpenLLM where suitable.
• Cloud versus on-premises hosting, inference costs and resource trade-offs.
• Retrieval-augmented generation, vector search and integration patterns; retrieval does not guarantee correctness.
• Low-code/no-code builders versus custom development: capabilities, limits and access requirements.

Agentic behaviour: goals, tools and bounded autonomy.

• Goal-oriented vs. reactive systems
• Decision-making independence and constraints

Illustrative agent workflows and deployment conditions.

• Customer-service assistance, escalation and reliability considerations.
• Supply chain management and optimization
• Personal productivity assistants and their capabilities

Integration strategies: Bringing agentic AI into existing workflows

• Phased implementation approaches
• Change management and user adoption

Risk assessment and mitigation: The responsible path forward

• Identifying potential failure modes
• Monitoring and control mechanisms
• Building human oversight into autonomous systems

Future directions and uncertainty

• Potential capabilities and limits: separate evidence from speculation.
• Industry-specific applications in development

Core principles

• Fairness: historical bias examples, evaluation across groups, mitigation and limits.
• Transparency and explainability: communicate decisions and trade-offs appropriately.
• Privacy: data minimisation, consent, anonymisation limitations and cross-border questions.
• Human agency: meaningful oversight, approval boundaries and augmentation of judgement.

Governance and legal awareness

• Compare binding laws, voluntary risk frameworks, sector requirements and organisational policies.
• EU AI Act: risk-based obligations and prohibited uses at an awareness level; applicability and requirements depend on the system and legal context.
• United States and Asia-Pacific approaches: federal/state/sector obligations and differing national or voluntary frameworks; no universal compliance recipe.
• Understand the role and limitations of industry initiatives, internal review boards and independent assessments.

From principles to practice

• Assign responsibilities and review processes; assess risks and engage stakeholders.
• Document decisions and evaluation evidence; test fairness, safety and robustness.
• Monitor systems and plan incident response, remediation and vendor due diligence.
• Review data handling and relevant contractual requirements with appropriate specialists.

Emerging ethical issues and culture

• Synthetic media, provenance/detection limits, attribution and creative rights.
• Hiring, surveillance, autonomous decisions and military-use contexts: discuss discrimination, civil liberties, accountability and relevant specialist review.
• Awareness training, concerns/escalation channels and cross-functional collaboration.
• Engage affected communities and choose meaningful measures for responsible use.