The course as four chapters divided into two days. Day one will cover "Introduction and Fundamentals" and "Case studies of successful AI application" while day two will cover "Technology maturity" and "Risk considerations"
Chapter 1: Introduction and Fundamentals
In the first part, we will discuss what it means for something to be considered AI. Basic concepts are introduced followed by a short example to solidify the knowledge. We will discuss students’ backgrounds and their expectations from the course together with the learning outcomes.
- Introduction to fundamentals of modern AI
- Classification & Regression (Supervised learning)
Answers to a problem can already be present in a database of previous instances when a similar decision had to be made. Supervised learning can automatize the decision process by leveraging the patterns seen in the past. We will learn how classification works on a simplified spam filter, and how regression works on a problem of real-estate price prediction.
- Clustering (Unsupervised Learning)
Data can hide an unseen structure that may be useful when reasoning about a market. Clustering methods can uncover such structures. We will see how clustering can find customer segments for more effective marketing campaigns.
- AI agents (Reinforcement Learning)
AI promises to replace human agents in highly repetitive and high-frequency tasks. We will get insights into how AI agents work and how the gaming industry uses them to automatize product testing and quality control.
Designing a product or a system requires designers to be creative within bounds defined by some cost metric. Whether it is resource optimization in architecture or scheduling within an organization, AI can deliver superhuman performance while minimizing the resources needed to realize it. We will learn how Autodesk augments product design with its AI services.
- Deep learning
Deep learning and neural networks have garnered extensive media coverage. We will learn what are neural networks and how are they used. We will learn how can neural networks recognize purchasable objects on an image and take customers to relevant stores.
Chapter 2: Case studies of successful AI application
In this part, we will learn which problems companies tackled by deploying AI and how it influenced their business. Using the conceptual vocabulary built in the previous chapter we will be able to organize them into families of problems to help us recognize opportunities in our organizations.
We dive into four cases in this session:
AI used in fraud detection and client creditworthiness
AI used in virtual dressing rooms and Next Best Action agents for customer acquisition
- Food delivery
AI used in courier to order allocation
- Health care
AI used to extract relevant information from medical literature and pathology reports
- Q&A session
Chapter 3: Technology maturity
We will discuss the steps of AI technology maturity. Examples of steps being fundamental AI R&D, emerging technology, early business adoption, media hype stage, technology dead-end, or the final technology maturity. We will discuss at which point are risk minimal in business applications? How to assess if certain technology is ready for your business?
Case-study: Failure of IBM Watson
Case-study: Chatbots - advantages and limitations
Chapter 4: Risk considerations
There is a myriad of potential risks that AI introduces in a business. We will learn about the most common risks and discuss how to assess their cost and mitigate them.
- Data gathering
Certain AI systems need a large volume of data to operate successfully. If data is not present the acquisition can introduce additional costs to AI integration.
- Data cleaning
Historical data often does not exist in format and structure useful for AI applications. Assess whether resources needed to transform data to a usable state are worth the benefits.
- Model interpretability
To avoid unintended consequences of AI automatization, insight is needed into the inner workings of AI models and algorithms. Unfortunately, complex AI systems are often opaque to interpretation introducing unforeseen risks.
- Employee distrust
Complete automation with AI can introduce a sense of losing control and importance over the work process in the employees. Employee distrust can lead to underutilization of the deployed AI systems introducing losses in the work process. Learn how to mitigate the lack of trust among employees and combine their talents with automated solutions.
- Bias and error magnification
AI that depends on big data can be as good as the said data. Poor quality and highly biased data can amplify those biases and errors. Learn what it means for the data to be high quality to avoid perpetuating historical problems within a business.
- Loss of key skills in the workforce
Overreliance on automated systems can lead to the loss of skills among human talent. Hear about concerns that experts expressed that pilots are losing basic flying skills due to cockpit automation. Consider which skills are potentially being lost due to AI automation and are they worth keeping within the employees.
- Q&A session