The foundation of all successful AI is high-quality, findable, secure data!
The FAIR principles are guidelines for ensuring that data is
- Findable,
- Accessible,
- Interoperable, and
- Reusable.
I think it should be FAIRS for:
- Findable,
- Accessible,
- Interoperable,
- Reusable, and
- Secure.
Tip: If your organisation's data is in a good FAIRS state, AI projects are far more likely to succeed.
1. Artificial Intelligence (AI)
- AI is attempting to make computers use and behave using human intelligence.
- Teach PC to do tasks for us - automation.
- PC predicts using patterns and can act. And good at looking for anomalies.
- PC uses camera/photos to look for patterns.
- Engage in valuable conversations, use multiple sources of knowledge.
2. Machine Learning (ML)
- Train PC's to see patterns and see patterns, and look for anomalies.
- Example. Predict stock prices by looking at factors that affect stock prices.
- Anomaly Detection - Detects unusual patterns, e.g. CC used in Asia when typically in Europe, but transactions 10 min apart. Therefore, it is likely to be fraudulent. Sort rubbish.
- Predictive models by finding relationships. Give model data and train the model.
- Example: flowers have features/characteristics, e.g. colour, size, no petals, ...
- Using data to teach a machine
- Supervise ML - need quality data, including labels. Avg humidity, hrs sunshine, rainfall, temp, month of year (features), ice creams sold(label/class), so we feed in temp is Regression ML. Patient has features (weight, sex, age, bmi,...) give value btwn 0-1 of the person developing diabetes. is Classification ML
- Unsupervised ML - data is not labelled. Just the feature provided will group into clusters. Pulls data out and figures out its own criteria is Clustering ML. Useful for fraud detection.
- Training - good data based on a training set and a validation set. Train model, with most data, then check with the remaining - allows us to see how close to what happened. Service tries to figure out relationships. Model is used by test data - see how close/useful the model is.
3. Compute Vision
- Self-driving cars, sorting. Sort rubbish.
- Facial recognition, object recognition,..
- How do computers see? The picture is cut up into pixels, data is pulled and used to find possible answers.
- Some types on Azure: object detection, i.e. car, bike, bus. Image classification, i.e., horse, car. Semantic segmentation, i.e. Teams blur background. Image analysis context by bringing various tougher. Face detection. OCR - reads the image and converts it to text.
4. Natural Language Processing (NLP)
- interpret, e.g Grammarly, Spam Check, Alexia,
- Knowledge Mining - Extract info from knowledge and gain insights, e.g. social media marketing.
Principals:
- Fairness - ensure bias is excluded, e.g., based on gender.
- Reliable & Safety - need high confidence and in specific systems cannot fail, e.g. health systems, autonomous cars.
- Privacy & Security - Ensure data is protected and not giving away sensitive data.
- Inclusiveness - should be fair, i.e. VI users
- Transparency - what is the model based on, what could be an issue?
- Accountability - who is liable for AI decisions
Azure:
- Scalability & Reliability
- AI Resources: sit in an Azure Resource Group
AI Services in Azure:
- Azure Machine Learning - Developers to train, test and deploy ML models. Within a subscription, create a Azure ML Workspace (consists of: compute, data, jobs, models) can then publish as a service. Azure ML Designer, used for creating an ML pipeline, data to train the model. Automated Machine Learning user only needs to provide the data and select the model to use; the service figures it out.
- Cognitive Services - vision, speak, language, decision. REST API endpoints have already been trained; choose the model. Can deploy multiple parts individually or together.
4. Azure Cognitive Search - Data extraction & enrichment for indexing. Makes data searchable.
Anomaly detector resource - wizard to set up - Add Keys and endpoints to allow access.
Create a new Azure Machine Learning Service, which will create a Workspace. Use multiple Azure services such as Key Vault, AI, and storage accounts.
- Launch Studio
- Add Compute Cluster
- Add Data (csv, spreadsheets, nearly any form,...)
- AutomateML (figure it out without me) and run the job
- Will show trends
- Deploy the model (i.e. to a web service)
- Shows "Endpoints" - get URL and a test rig.
