AI for Business Leaders Flashcards
4 main areas where AI is powerful
Scale - processing large amounts of data
Pattern - finding patterns and optimums
Grouping - data with similarities
Extracting features from unstructured data
4 main branches of AI
Machine Learning
Natural Language Processing
Machine Vision
Robotics
Name 5 skill sets that are hardest to automate:
Building relationships Empathy Critical thinking Creativity Storytelling
3 main reasons for on-going AI revolution:
Computing power up, and costs down
Storage capacity up, and costs down
Data transport costs down
‘Data that has no defined organizational structure’ =
Unstructured Data
Give 2 reasons why data is growing so fast:
- Dramatic cost reduction in devices (e.g. cameras, microphones, sensors)
- Massive user content generation (social media)
Name of Anil Kumble’s company?
Spektacom
Is often used by corporates
to be a focal point for innovation for the organization
Often is a dedicated space that can also be used to showcase ideas to the outside world
Can be both on-site (tend to have a business unit focus) or off-site (more disruptive focus)
=
An Innovation Lab
Typically no fixed schedule
Typically sponsored by VC funds or corporates
Typically don’t provide upfront capital
Often have a specific industry or market focus
Example: Idealab
=
An Incubator
Used to accelerate growth for a company that already exists
Typically a fixed time schedule (e.g. 3-months)
Benefits for the startup may include:
– Small seed investment
– Access to a mentor network
Examples include: Y Combinator, Techstars
=
An Accelerator
Training a machine to identify patterns and/or predict outcomes
Often used to find relationships between variables
Contains many subsets of AI methodologies (Neural Network with Deep Learning, Evolutionary Algorithm etc.)
=
Machine Learning
4 questions machine learning asks:
- How can we learn from data?
- Can we find new patterns?
- Can we predict outcomes better?
- Can we automate decision making at scale?
5 techniques used in machine learning:
- supervised learning
- unsupervised learning
- ensemble learning
- neural networks and deep learning
- reinforcement learning
Prediction of output results from specific input data
– e.g. predict apartment price based on size, location, etc.
=
Supervised Learning
– No output results specified, only input data is given to the system
– Used to typically find insights or patterns from data
– e.g. McDonald’s can identify different customer groups based on their food preference
=
Unsupervised Learning
– Allows a system to “learn” through trial and error by utilizing feedback generated from past actions
– e.g. OpenAI Multi agent-hide and seek
=
Reinforcement Learning
Supervised learning, unsupervised learning, or reinforcement learning?
- optimization
- correlation analysis
- clustering
- classification
- learn objectives from real-life behavior
- regression
Supervised learning: classification, regression
Unsupervised learning: correlation analysis, clustering
Reinforcement learning: optimization, learn objectives from real-life behavior
‘Automated platforms that seek to understand your investment objectives and risk tolerance, select an asset allocation and investments, and then monitor the portfolio and make any necessary changes.’
=
Robo-Advisors
Name 3 robo-advisors based in Hong Kong:
Kristal
Aqumon
YF Financial
– Inspired by biological networks in the brain; shares insights from cognitive neuroscience
– Use layers of processing to derive an output
– Each node is connected to other nodes and each connection can be weighted differently
Neural Networks
3 components of a neural network:
- Input Layer
- Hidden (computation) Layer
- Output Layer
Cycle through combining and re-evaluating possible solutions to come up with the best one
Are a subset of Machine Learning
Draws inspiration from Darwin’s theory of evolution
Evolutionary Algorithms
Machine learning explores permutations, generates design alternatives, picks optimal design
– Learns from each iteration what works and what doesn’t
– Identifies & mitigates clashes between models generated by individual teams (early warning)
Generative Design
– Training a machine to collect pictures and derive meaning
– Hardware required varies considerably based on use case
– subsets includes Optical Character Recognition (OCR)
Machine Vision
