Module 4 Flashcards
What is Artificial Intelligence?
An attempt at building a computer system that thinks and acts like humans.
Grand vision: Presumes that computer systems are as smart as humans
Realistic vision: views AI as systems that take data inputs, process them, and produce outputs, while performing complex tasks, that can be too difficult for humans.
Expert systems:
Captures knowledge in a very specific domain of human expertise. They look at knowledge as a set of rules and perform limited tasks.
An example is when an engineer constructs rules from which the AI can diagnose malfunction, then the IT expert constructs
A user interface: The user’s query is received and sent to the inference engine
Inference engine: contains rules to solve the problem received.
Knowledge base: contains all the rules based on which the decisions are made.
Machine Learning:
Starts with a very large data set and enables the computer programs to learn from this data. Learns how to recognize patterns, and utilize this prior learning. An example is Facebook Ads and Netflix recommendations.
Neural Networks:
Inspired by the structure and functioning of the human brain. Their function is to find patterns and relationships in a large amount of data. The AI “learns” patterns by searching for relationships, building models for them, and correcting them continuously. Meanwhile, humans “train” by feeding data with known outputs.
For example, credit card users’ behavior is analyzed and flagged if considered unusual.
Limitations of Neural Networks and Machine Learning:
1)Both require very large datasets
2)Some patterns may require human judgment
3)Often large datasets are not available, as the decision does not have many inputs
4)It is difficult to understand how the system arrives to the solutions
5)AI has no sense for ethics
Genetic Algorithms:
The main goal of this AI is to find the optimal solution for a problem by examining a large number of solutions.
Solves optimization problems by changing variables and arriving at the best solution fast, for example, cost minimization, scheduling, and designs.
Furthermore searches for solution variables based on evolutionary processes.
Natural language processing:
This AI has the ability to understand, speak, read in natural language, and translate it. Examples are Google Translate, Siri, etc.
Useful in a limited amount of fields, however, for example, helps communicate with car’s heating system.
Computer vision systems:
Digital image systems which create a map of an image, which then allows to recognize these images in a large database. For example, Facebook DeepFace, which identifies faces across systems.
Robotics:
Involves designing, constructing, and operating machines that can substitute for humans. In-home context vacuums substitute certain actions, in factory level welding machines and assembly.
Intelligent agents:
Works without human intervention and their function is to carry out repetitive tasks. Limited knowledge base, which is learned or built-in. For example, chatbots that delete junk emails.
Use of AI at work:
1)Engage in task coordination - for example, for uber the AI assigns rides and communicates with customers
2)Exercise soft surveillance: through data collection Uber drivers are tracked, to monitor working hours and behavior.
3)Reward or punish the workforce.
Risks:
It can nudge workers to behave in a certain way, which may have bad repercussions. It can be done through surge prices or psychological levers. AI can be biased also in recruitment if it learns from past data. AI also lacks transparency, thus decisions are hard to interpret. And it also involves security risks and ethical concerns.
AI in recruitment:
1) Resume screening and sorting
2) Scheduling interviews and answering common questions.
3)Using predictive analytics for analyzing past decisions.
4)Can analyze video interviews - facial expressions and tone.
5)Matching candidates with jobs.
Name major types of AI
- Expert systems
- Machine learning
- Neural networks
- Deep learning
- Genetic algorithms
- Natural-language processing
- Robotics
- Intelligent agents
