Unit 6.3 - AI

branch of computer science dealing with the simulation of intelligent behaviours by computers

-narrow -general -strong

machine has superior performance to a human for one task (can’t think or feel like humans)

machine has similar performance to human doing a specific task (can multi-task like humans)

machine has superior performance to a human for multiple tasks

-news generation -smart home devices (Alexa, Siri) -chatbots (Chat GPT) -autonomous cars -facial expression recognition

-collection of data -collection of rules for using that data -ability to reason (-ability to learn and adapt)

-facial recognition -chatbots -alphago

sub-set of AI that can mimic human knowledge, abilities, and experiences. (They use expert knowledge and inference to solve problems)

-diagnosis of patients’ illness -tax and financial calculations -strategy games (chess) -identification of living things and scientific compounds -oil and mineral prospecting

-high accuracy (gives probability of their suggestion being correct) -consistent results -can store many ideas and facts -fast response times -unbiased reporting and analysis of facts

-set up and maintenance is expensive -expert system needs lots of training -cold responses (might not be appropriate always) -only as good as facts in the system -can be wrong

-user interface -inference engine -explanation system -knowledge base -rules base

method of interaction with user (dialogue boxes, command prompts) –> questions asked are multiple choice / yes or no –> questions are based on previous answers

-processing element -examines knowledge base for information that matches users’ answers -gathers information from user by asking questions based on previous answers -uses inference rules in rules base -uses rules base to find a match to the user

-respository of facts -collection of objects and attributes -stores knowledge about an area of expertise, obtained from expert resources

-set of inference rules -so that (used by) inference engine to draw conclusions

1) information gathered from human experts / written sources / internet 2) gathered information populates knowledge base (first created) 3) rules base is created (series of inference rules) 4) inference engine is set up 5) user interface is developed for communication 6) fully tested by running the system with known outcomes (compare results, changes made)

subset of AI where a program has the ability to automatically adapt its own processes and/or data due to their past experiences

Search engines –>if user finds desired website on page 1, success. If user goes to page 2 onwards, it is a failure and changes should be made Categorising email as spam 1) data is collected about emails (content, headers, senders name, email address, etc..) 2) stop words are removed (the, a, and) 2) frequently used words / phrases indicate that the email is likely spam (lottery, earn, full-refund) 3) machine learning model is built 4) training data set is used to train the model and make it learn using past emails that are spam 5) model is tested live Recognising user buying history –>If customer A is interested in same things as customer B, might recommend some other customer A’s shopping things to customer B Detection of fraudulent activity 1) Data is gathered by survey or web scraping (shopping habits and personal data) 2) unnecessary data is removed 3) algorithm is trained through real examples of customer purchasing behaviour –> If a customer goes out of buying habits (suddenly buying too expensive things) then high chance of fraudulent activity taking place 4) model is built based on the learning from the training data 5) model is fully tested with known data and outcomes and model is modified if needed

-AI represents simulated intelligence in machines, ML is machines making decisions without being programmed to -AI aims to build machines capable of human-like thinking, ML aims to make machines that learn through data acquisition, to solve new problems