AF7620 Chap.1 AI for Creatives: Tools and Shifting Creative Roles
AI for Creatives: Tools and Shifting Creative Roles
Start with the words, because the tools change monthly
The opening lecture lays out the vocabulary before it shows a single tool, and that order is deliberate. Artificial intelligence is the broad field of machines imitating human cognitive functions such as learning and problem solving.
Machine learning is the branch that learns patterns from data and makes decisions without being explicitly programmed for each case. Deep learning is the part of machine learning that uses many-layered neural networks, and a neural network is itself an arrangement of artificial nodes loosely modelled on biological neurons that pass information from layer to layer.
Generative AI is the family that creates new text, images, audio and video by learning the patterns in data. A large language model is a deep learning model trained on vast amounts of text, and a prompt is the instruction a user gives any of these systems.
Knowing which layer a tool sits in tells you what it can and cannot be asked to do.
Newer terms you will meet in the pipeline lectures
The lecture also lists the terms that arrived with the latest wave: multimodal models that handle text, images, video and audio together; retrieval that combines outside knowledge with generation; mixture-of-experts systems in which several specialised models cooperate; alignment, which is about keeping a system's goals consistent with human values; hallucination, a model producing inaccurate or false output; fine-tuning, which adapts a pretrained model to a specific task; and diffusion models, which generate an image by removing noise step by step.
The distinction between a chatbot and an agent matters for production planning: a chatbot answers one request at a time, while an agent carries out multi-step tasks with tools, memory and a goal.
AI changes tasks, not whole jobs
The most useful slide for a producer divides work by how far AI transforms it. Minimal transformation, covering about forty per cent of skills, includes patient care and construction.
Assisted transformation, about nineteen per cent, includes teaching and creative design. Mixed transformation, again about forty per cent, includes medical coding and travel planning. Full transformation is only about one per cent and covers tasks such as basic arithmetic and image classification.
Creative design sits in the assisted band, which is the course's whole premise: the tools speed up and extend the work of a designer or producer rather than replacing the judgement behind it.
Five levels of AI-assisted creation
The lecture then ranks creative practice by how much the human contributes.
At the bottom is passive automation, where a person types a simple prompt and the machine does almost everything. Above it are manual enhancement, where the creator actively adjusts inputs and style; hybrid enhancement, where AI output is combined with photography or drawing; and strategic automation, where AI handles repetitive and technical steps while the human keeps every creative decision.
At the top is symbiotic co-creation, in which human and machine become sources of inspiration for each other. A companion slide traces creative growth from imitation through variation, combination and transformation to original creation.
Both scales are a warning against the lowest rung: work generated from a bare prompt tends towards the average of its training data, and the lecture quotes the observation that such output reflects an echo chamber of collective taste.
What this chapter covers
- 01
Artificial intelligence, machine learning, deep learning and neural networks
- 02
Generative AI, large language models and the prompt
- 03
Multimodal models, retrieval, alignment, hallucination and fine-tuning
- 04
Why a chatbot and an agent need different production plans
- 05
The four bands of task transformation
- 06
Five levels of AI-assisted creation
- 07
From imitation to original creation
Worked example · free
Place a production task on the five-level scale
- 2Place the one-line background paintings.
- 2Place the line-art clean-up and the lighting variations.
- 2Say which use weakens a pitch and why.
Key terms
- Machine Learning
- A branch of AI in which a system learns patterns from data and uses them to make decisions or predictions, without being explicitly programmed for each case.
- Deep Learning
- A branch of machine learning that processes large, complex data through neural networks with many layers, loosely imitating how the brain works.
- Large Language Model
- A deep learning model trained on very large amounts of text that can understand, generate and process human language, and that sits behind many AI applications.
- Diffusion Model
- A generative model that creates an image or other data by starting from random noise and removing it step by step until a coherent output appears.
- Hallucination
- Output from a generative model that is inaccurate, meaningless or false but presented with the same confidence as correct output.
- AI Agent
- A system that carries out multi-step tasks on its own, using tools, memory and a goal, rather than answering a single request and stopping.
- Passive Automation
- The lowest level of AI-assisted creation, in which a person supplies only a simple prompt and the machine generates the work with almost no further human input.
AI for Creatives: Tools and Shifting Creative Roles FAQ
Will AI replace creative designers?
The lecture's evidence points the other way for now. Creative design is placed in the assisted band of transformation, the group of skills where AI speeds up and supports the work rather than taking it over, and only about one per cent of skills are shown as fully transformed. The same slides note that anyone hoping to replace a designer with a prompt must first describe exactly what they want, which is itself design work.
What is the difference between machine learning and deep learning?
Machine learning is the wider family of methods that learn patterns from data instead of following hand-written rules. Deep learning is one branch of it that uses neural networks with many layers, which lets it handle very large and complex data such as images, sound and video. Most of the generative tools used in this course are deep learning systems.
Why does the course care about the level of human input?
Because the pitch is graded on innovation and on a justified, specific workflow, and because copyright protection depends on human creative contribution. Work that stops at a bare prompt tends to look generic and gives the weakest authorship claim, while a workflow that shows human decisions before and after each AI step is both more original and easier to defend.
Assessment move
Take one scene from your planned pitch and write down every task needed to make it. Mark each task with the band of transformation it belongs to and the level of human input you intend, then circle any task that sits at passive automation. Those circled tasks are where your workflow is least specific and where a reviewer will push hardest.
Working through AI for Creatives: Tools and Shifting Creative Roles in AF7620? Sia is AskSia’s AI Arts and Humanities tutor — ask any AF7620 AI for Creatives: Tools and Shifting Creative Roles question and get a clear, step-by-step explanation grounded in how AF7620 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
