UPI Study › Courses › Ethics in Technology › Free lessons › Lesson 14.2 Free with a free account
How Machines That Simulate Human Thinking Are Changing Every Industry — Artificial Intelligence, Machine Learning, Deep Learning, and Generative AI
Ethics in Technology · Chapter 14: Tomorrow's Technology: AI, Biotech, IoT, and the Connected FutureSection 14.2 · lesson 90 of 11613 min · free with a free account
Free preview
Create a free account to keep watching
It takes a minute and costs nothing. Your free account unlocks all of Chapter 1 and 3 more lessons of Ethics in Technology.
Lesson complete
Earn the 3 college credits for Ethics in Technology
Enrolling adds a quiz after each lesson, the graded work and the official transcript. Recommended for credit by NCCRS.
Next lesson: Where Biology Meets Technology — The Science of Modifying Living Systems for Human Benefit →← Looking Back at DES — The Encryption Method That Dominated for Decades Before Being ReplacedNext lesson: Where Biology Meets Technology — The Science of Modifying Living Systems for Human Benefit →
In this lesson
Four related terms, practical applications, and responsibility for their effects.
- Distinguish AI, machine learning, deep learning, and generative AI.
- Explain the four industry examples and their human review needs.
- Analyze bias, workforce changes, and accountability.
- Identify meaningful oversight throughout development and use.
Free study notes
Get this lesson's notes by email
The key points and 4 practice questions with answers, in your inbox. One email, free.
One email. Unsubscribe in one click.
Transcript
Everything said in the video, section by section (1,637 words). Read along, or search it.
Tomorrow's Technology: AI, Biotech, IoT, and the Connected Future
AI Across Industries
Artificial intelligence already supports work across industries, including activities that users may never directly notice. This lesson separates four related terms, then connects them with practical applications and ethical questions. AI is the broad category; machine learning and deep learning describe increasingly specific methods. Generative AI describes systems that produce content, often using deep learning in modern implementations. These labels help explain a system’s design. They do not guarantee what it can do reliably in a particular setting. We will examine healthcare, finance, transportation, and retail examples drawn from the course. Each example also raises questions about human review, affected people, and actual performance. Bias, workforce changes, and accountability deserve attention throughout development and use, rather than afterward. By the end, explain both useful capabilities and the responsibilities surrounding their real-world effects.
Four Related Terms
The Vocabulary Describes Scope, Method, and Purpose
Artificial intelligence broadly includes systems performing tasks commonly associated with human intelligence, through varied methods. The term does not require a particular learning algorithm or proof of human-like understanding. Machine learning is an AI approach that learns patterns from data during a training process. It contrasts with specifying every task rule by hand, although people still define important design choices. Deep learning is a specific family of machine learning methods using neural networks with multiple layers. Many modern generative systems use those methods to produce text, images, audio, or code. Generative describes the output-producing purpose, rather than every possible task a neural network performs. These terms therefore identify different aspects of a system. They should not be used interchangeably when explaining its method or purpose. Ask which method is used, which output is intended, and what evidence supports its performance.
Read the Relationship
Deep Learning Is a Specific Method within Machine Learning
Begin with the broad AI field, which includes several possible approaches to performing intelligent tasks. Machine learning sits within that field, while deep learning sits within machine learning. Every deep learning system therefore uses machine learning, but other learning methods need no deep network. Similarly, AI includes approaches besides learning from data, such as systems using specified rules. The diagram’s final connection shows a common route to modern tools that generate content. It does not mean that every conceivable generative model must use a deep neural network. Deep learning can classify an image or predict a value. It need not create content for a user. Generative systems emphasize producing content, even when prediction helps construct that output internally. Keep the strict subset relationships separate from the common implementation choices used in contemporary generative tools.
Learning from Data
Training and Using a Model Are Different Activities
During training, a learning method fits patterns using examples and an objective defined for the task. People choose the data, representations, training approach, and the criteria used to judge the result. The trained model then produces outputs for supplied inputs during inference, or ordinary use. These activities differ: responding to a new input need not retrain the deployed model automatically. A system does not become more reliable simply because it has processed another user request. Evaluation checks the model against relevant evidence, including conditions it will encounter when deployed. Training data can differ from current inputs. Successful development results may therefore not transfer unchanged to the deployed setting. Large datasets can support useful learning, but quality and relevance matter alongside their size. This distinction helps explain why ongoing monitoring is necessary after a system starts doing real work.
Layered Neural Networks
Deep Learning Uses Multiple Stages of Learned Processing
Deep learning uses multiple layers of processing within a neural network trained for a task. Inputs may include an image, a passage of text, recorded sound, or another information representation. The layers transform that information through learned operations, building representations used by later stages. The resulting output might classify an input, predict an outcome, or support a generative application. The description of layers distinguishes this approach from other forms of machine learning. Neural-network terminology draws on a limited brain analogy. The design does not reproduce the whole biological brain. It does not establish human-like thought, understanding, intention, or consciousness merely from the system’s name. Claims about capability should therefore be assessed through performance evidence for the particular task. Use the vocabulary to explain a method, while avoiding assumptions that ordinary human traits automatically apply.
Generating Content
Creating an Output Is Different from Merely Assigning a Category
A generative tool produces content, rather than only returning a category or an outcome prediction. It may draft text in response to a prompt, including a passage for someone to review. Image generation can produce a synthetic visual, while audio generation can produce speech or other sounds. Code generation proposes a program or function, which still needs review before being relied upon. These examples differ in output format, but share the purpose of constructing a content result. A generated result can look plausible to its audience. It may still contain errors, unfair patterns, or unsuitable material. Creating new output does not guarantee accuracy, originality, or that the intended audience will benefit. Prediction may help construct the content, so generation and underlying prediction are not mutually exclusive operations. Evaluate the actual output and use context, rather than treating the generative label as an assurance.
The Four Industry Examples
Pattern Recognition Supports Different Practical Tasks
In healthcare, AI can flag possible tumors in medical scans for a radiologist to review. The flag supports attention and analysis, rather than establishing a diagnosis without appropriate clinical assessment. In finance, unusual transaction patterns can be identified as possible signs of fraud needing investigation. Unusual activity is not automatically fraudulent, so the organization remains responsible for how flags are used. Transportation applications include driver-assistance features and predictions about when a vehicle may need maintenance. Their intended operating limits and safety requirements matter alongside the ability to recognize patterns. Retail applications personalize product recommendations and forecast demand, influencing both shoppers and organizational planning. These examples share data-based processing at scale. Their consequences and review requirements differ according to the task and setting. Faster or more consistent output still needs evidence that it works reliably for the actual task.
Harmful Bias
Unfair Patterns Can Survive Behind an Apparently Consistent Output
Harmful bias can arise when historical or unrepresentative data influences the patterns a system learns. Past unfairness may therefore be repeated in outputs, even without an intention to discriminate. However, data is not the only influence on harmful bias. Design, measurement, deployment, and human decisions also matter. An objective or evaluation method can overlook differences that are important to affected groups. A seemingly ordinary output may conceal a harmful pattern when compared with actual real-world outcomes. Overall accuracy can also hide disparities, if performance is not examined for relevant populations. That helps explain why bias sometimes remains unnoticed until people have already experienced harm. Review the outcomes, the affected groups, and the setting, rather than assuming consistent outputs are fair. Oversight should look for those patterns throughout use, with a practical route to investigate and respond.
Jobs and Tasks
New Opportunities Do Not Automatically Replace Lost Livelihoods
Automation changes tasks within jobs, and can also reduce or remove some existing roles. Those changes affect livelihoods, rather than only the technical efficiency of the organization’s workflow. The same technology may create new opportunities, but those opportunities need not match the losses. A newly created role may require different skills from the ones held by displaced workers. Its location, timing, accessibility, or total number of positions may also differ from what disappeared. Consequently, new work elsewhere does not automatically compensate an affected workforce for a particular loss. Exposure to a technology is also different from a guarantee that an entire job will disappear. Many roles contain tasks that still require human input, so transformation is another possible outcome. Examine actual changes and support affected workers, instead of assuming either universal replacement or automatic benefit.
Accountability
People and Organizations Need Clear Responsibility for Harm
Accountability concerns who must answer for a system’s use and the consequences of its decisions. Several participants can influence the result, which makes responsibilities harder to trace when harm occurs. Developers and suppliers shape design choices, training processes, limitations, and the evidence they communicate. The deploying organization decides whether the system is suitable, and what safeguards accompany its use. Operators and decision makers configure it, interpret outputs, escalate problems, and take resulting actions. Calling the outcome an automated decision does not make these human and organizational choices disappear. The applicable legal duties depend on the setting, rather than a single rule for every AI application. Operational responsibilities should nevertheless specify review authority, records, and ways to challenge harmful decisions. Do not treat the machine itself as an explanation that removes the need for accountable people.
Meaningful Human Oversight
Build Review into the Process and Keep It Active
Meaningful oversight begins before use, with a clear task and an assessment of affected people. Decision makers need evidence about performance and limitations, together with assigned responsibilities for the application. During use, check outputs and their consequences, rather than merely accepting whatever the system produces. Reviewers need information, time, competence, and authority to question or stop unsuitable decisions. A routine approval by someone unable to intervene supplies little protection against harmful outcomes. As conditions change, investigate incidents and update safeguards before continuing to assume the original assessment holds. Monitoring should include bias, workforce effects, and accountability, alongside whether the technical task succeeds. These practices do not require rejecting every useful application; they make its deployment more carefully assessable. Human review works best when it is an active part of the process throughout operation.
Review and Apply
Explain the Distinctions and Who Bears the Consequences
For the first review question, identify deep learning as the layered-neural-network subset of machine learning. Machine learning also includes other approaches, so the two terms do not describe identical categories. For the second, explain that generative AI produces content, such as text, images, audio, or code. For the third, explain how ordinary-looking outputs or overall averages can hide unfair real-world effects. Historical data may contribute, while design and deployment choices can create or amplify harmful patterns. For the fourth, compare displaced and new work through skills, numbers, timing, and practical access. For the fifth, identify the different participants whose choices shape an automated decision and its consequences. Define their duties and review authority, rather than using automation as a reason to leave responsibility unclear. Precise vocabulary and ongoing evidence help evaluate useful applications together with their effects on people.
About Ethics in Technology at UPI Study
Ethics in Technology (Computer Science 195) is a self-paced online course of 116 video lessons in 15 chapters, from cybersecurity basics, privacy and surveillance to big data, hacking, malware, network defense, intellectual property, cybercrime, encryption, emerging technology and digital citizenship. It carries a NCCRS credit recommendation for 3 semester credits.
Every lesson transcript is free on pages like this one, and the videos are free with a free UPI Study account. Enrolling adds a practice quiz after each lesson, the graded work and final exam that earn the credit, lifetime access, and an official transcript on completion. UPI Study has helped over 130,000 students with college credits. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.
Enroll now → Course details Download the syllabus All Computer Science courses
Compare: Cybersecurity vs Ethics in Technology · Which ethics course?
Why take it here
Why students take Ethics in Technology with UPI Study
Where the credit goes
Universities that work with NCCRS
Ethics in Technology carries a NCCRS credit recommendation. These universities work with NCCRS; some are UPI Study's signed partners. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.
Thomas Edison State UniversitySigned partner
University of the PeopleSigned partner
Charter Oak State CollegeSigned partner
Strayer UniversitySigned partner
Columbia Southern UniversitySigned partner
Indiana TechSigned partner
Penn StateWorks with NCCRS
Southern New Hampshire UniversityWorks with NCCRS
SUNYWorks with NCCRS
+many more cooperating universitiesFind my college →Find my college → Not listed? Ask Athena in the corner of this page, or book a 15-minute call.
FAQ
Questions students ask us
About How Machines That Simulate Human Thinking Are Changing Every Industry — Artificial Intelligence, Machine Learning, Deep Learning, and Generative AI
What is the difference between AI, machine learning, and deep learning?
What is generative AI, and can its output be trusted?
How does AI become biased?
Who is responsible when an AI system makes a harmful decision?
About the Ethics in Technology course
How many college credits is Ethics in Technology?
How much does Ethics in Technology cost and what is included?
Which universities accept UPI Study credits?
How long does Ethics in Technology take?
How are the exams done?
How do my credits reach my college?
Ready for the full course? Earn 3 credits in Ethics in Technology.
116 video lessons, a quiz after each one, and the graded work that earns the credit. Recommended for credit by NCCRS.
3 credits · NCCRSEnroll →