Subsymbolic AI (machine learning)
Topics
- How machine learning works at a general level
- Learning paradigms
- Supervised learning
- Unsupervised learning
- Reinforcement learning
- Neural networks
- Neuro-symbolic AI
Weekly Discussion: Identifying Machine Learning in Everyday Contexts
Please submit all graded work via Canvas. Participation requirements and grading details are provided in Canvas.
Step 1: Choose a familiar context
Select one real-world example you have encountered from your work or daily life. It may already use machine learning, or it may be a system you believe could benefit from it.
Examples include:
- Email spam filtering
- Fraud detection in banking
- Predictive text or autocomplete
- Social media feed ranking
- ……
Briefly describe:
- What the system or situation is
- What problem it is trying to solve
- What kinds of decisions or predictions are involved
Step 2: Why machine learning fits this problem
Explain why this scenario is suitable for a machine learning approach.
You may consider:
- Are there patterns that could be learned from past data?
- Is the task about prediction, classification, ranking, or grouping?
- Would writing explicit rules be difficult or impractical?
- Does the system need to adapt over time?
Clarify whether the task would likely involve:
- Supervised learning
- Unsupervised learning
- Reinforcement learning
No technical detail is needed, but your reasoning should be clear.
Step 3: How the system would be adapted for machine learning
Think concretely about how this problem would be operationalized.
You may address:
- What would count as the training data?
- What are the inputs (features)?
- What is the output or prediction?
- How would performance be evaluated?
- What risks of overfitting or bias might arise?
Focus on the logic of learning from data, not on programming details.
Step 4: Limits and considerations
Reflect on potential limitations or risks.
- Could the system make confident but incorrect predictions?
- What kinds of bias might appear in training data?
- Would transparency or interpretability matter?
- Would a rule-based approach be preferable in some cases?
Consider both benefits and trade-offs.
Expected outcome
Your post should present a clear analysis of one scenario, showing:
- why machine learning is appropriate (or not),
- how it would learn from data,
- and what practical or ethical considerations might arise.