Get the big picture with Adobe Express’s generative AI glossary
Done right, generative AI amplifies creativity and workflow intelligence without replacing the beauty and power of human savvy and ingenuity. Using everyday language to generate extraordinary new visual content, the applications and inspiration are endless for designers, marketers, small businesses, and solopreneuers.
This generative AI glossary will get you going with the basics and beyond that you’ll need to grasp all things AI and recognize the big picture when creating unique, stunning imagery and text using Adobe Firefly.
Summary/Overview
Demystifying artificial intelligence and generative AI
As a type of generative artificial intelligence technology, AI art and image generators work similarly to other types of AI, using a machine learning model and large datasets to produce a specific type of result. With the ability to generate images based on simple text inputs, AI image generators revolutionize the creative process by providing a quick and efficient way to bring your visual ideas to life with countless applications.
Referring to a class of machine learning algorithms, generative AI is designed to create new and original content like the scenic image examples later in this blog. The quality and diversity of the generated output depend on the quality and diversity of the training data, the architecture and parameters of the model, and the training process. Anybody can use everyday language and other inputs to produce images, videos, documents, digital experiences — and in the case of Adobe Firefly: beautiful images, transformed text, play with color, and so much more. Getting a grasp of the AI big picture is a great place to begin.
Generative AI glossary
The practical applications are many and the results can be fantastical, but AI is neither magic nor science fiction. The following glossary for this innovative and evolving field contains the fundamental terms and concepts you will need to get smart on artificial intelligence and generative AI tools.
1. Artificial Intelligence (AI)
AI refers to the field of computer science that aims to create intelligent systems capable of performing tasks requiring human-like intelligence. It encompasses various techniques and algorithms enabling machines to learn from data, reason, and adapt to new information. Artificial intelligence uses a machine learning model, large datasets, and pattern recognition to be able to produce a specific type of result, such as personalized recommendations, voice recognition, language translations, and much more.
2. Generative AI
Generative AI is a type of artificial intelligence that can translate ordinary words and other inputs into extraordinary results. While the conversation around this technology has centered on AI image and art generation, generative AI can do much more than generate static images from text prompts.
With a few simple words and the right AI generator, anyone can create videos, documents, and digital experiences, as well as rich images and art. AI art generators can also be useful for producing “creative building blocks” like brushes, vectors, and textures that can add to or form the foundation of pieces of content.
3. Text prompts
A text prompt is a specific input given to an AI language model to generate desired content or responses. It typically consists of a short sentence or phrase that provides context and cues the AI to generate text relevant to the given prompt. Text prompts are widely used in natural language processing and creative AI applications.
Writing text prompts involves crafting specific written instructions or questions to guide generative AI models, shaping their output according to desired content and style. Effective prompts play a crucial role in obtaining the desired results.
4. Adobe Firefly
Adobe Firefly is the new family of creative generative AI models coming to Adobe products, focusing initially on image and text effect generation. Firefly will offer new ways to ideate, create, and communicate while significantly improving creative workflows. Firefly is the natural extension of the technology Adobe has produced over the past 40 years, driven by the belief that people should be empowered to bring their ideas into the world precisely as they imagine them.
5. Large language models
Large language models, like the popular ChatGPT-3, contain billions of parameters and excel in processing and generating human-like language. They have proven remarkable in various natural language understanding and generation tasks.
6. Machine learning
Machine learning is a subset of AI that enables systems to learn from data and improve their performance over time without explicit programming. It includes supervised, unsupervised, and reinforcement learning techniques.
7. Neural networks
Neural networks are computational models inspired by the human brain's structure. They consist of interconnected nodes or neurons, organized in layers, and are widely used in various AI tasks.
8. Deep Learning
Deep learning is a subset of machine learning that utilizes neural networks with multiple hidden layers to process complex data and solve intricate problems, such as image recognition and natural language processing.
9. Image generation
Image generation in AI involves creating realistic images from scratch using generative models like GANs, VAEs, or transformers, revolutionizing creative applications and visual content generation. As a type of generative AI technology, AI art generators work similarly to other types of artificial intelligence, which use a machine learning model and large datasets to be able to produce a specific type of result.
10. Text generation
Text generation in AI refers to the process of producing coherent and contextually relevant written content using large language models or recurrent neural networks.
11. Unsupervised learning
Unsupervised learning is an AI machine learning approach where the model learns from unlabeled data, finding patterns and structures without explicit supervision.
12. Transfer learning
Transfer learning is an AI technique where knowledge gained from training on one task is applied to improve the learning and performance of another related task, reducing the need for extensive training data.
13. Data augmentation
Data augmentation involves artificially increasing the size of a dataset by applying various transformations to the original data, enhancing the model's generalization capabilities.
14. Bias in AI
Bias in AI refers to the presence of unfair or unjust preferences within AI models, often reflecting human biases present in the training data, leading to discriminatory outputs.
15. Explainable AI
Explainable AI aims to make AI models' decision-making processes understandable and transparent, essential for building trust and understanding their behavior, particularly in critical applications.
16. Ethics in AI
Ethics in AI addresses the responsible development, deployment, and use of AI technologies, addressing concerns related to privacy, bias, transparency, and accountability.
17. AI art
AI art encompasses artwork created or co-created by AI systems using generative algorithms, reflecting the merging of human creativity and artificial intelligence capabilities. AI generators like Adobe Firefly can enhance creativity by giving people new ways to imagine, experiment, and bring their ideas to life.
For Firefly, the future vision is for creators to be able to use everyday language and other inputs to use AI to quickly be able to test out design variations, remove distractions from photos, add elements to an illustration, change the mood of a video, add texture to 3D objects, create digital experiences, and more.
18. Data privacy
Data privacy refers to the protection of personal and sensitive information used in AI training datasets to prevent unauthorized access and potential misuse.
19. Image-to-image translation
Image-to-image translation is a generative AI technique that converts images from one domain to another, enabling tasks like converting sketches to photorealistic images or changing day scenes to night.
20. Language models
Language models are AI systems designed to process and generate human language, crucial for various NLP tasks like machine translation and text summarization.
21. Pre-trained models
Pre-trained models are AI models trained on large-scale datasets and made available for further fine-tuning or transfer learning on specific tasks.
22. Natural language processing (NLP)
NLP is a field of AI focused on enabling computers to understand, interpret, and generate human language, underpinning applications like chatbots and sentiment analysis.
23. Style transfer
Style transfer in AI involves merging the style of one image with the content of another, creating novel and artistic visual outputs.
24. Inpainting
Inpainting is an AI image generation technique used to fill in missing parts of an image, often applied for restoration and enhancement purposes.
25. Hyperparameter tuning
Hyperparameter tuning is the process of optimizing the settings (hyperparameters) of AI models to achieve better performance.
26. Transferability
Transferability refers to the ability of AI models to apply knowledge gained from one domain to improve performance in another domain.
27. Multi-modal AI
Multi-modal AI handles and generates content from multiple data types, such as text, images, and audio, enabling more diverse and creative outputs.
28. Generative adversarial networks (GANs)
GANs are a class of generative AI models consisting of two neural networks, the generator and the discriminator, working in tandem to produce high-quality synthetic data.
29. Variational autoencoders (VAEs)
VAEs are generative models that utilize an encoder and a decoder to learn latent representations of data and generate new samples.
30. Transformers
Transformers are a type of deep learning architecture widely used in large language models, facilitating parallel processing of data and improving computational efficiency.
31. Collaborative AI
Collaborative AI involves AI systems designed to work in cooperation with humans, enhancing human capabilities and facilitating collaborative decision-making.
32. Ethical issues in generative AI
Ethical issues in generative AI include concerns about biased content generation, responsible AI deployment, and the impact of AI-generated content on society and the art industry.
33. Domain-specific language models
Domain-specific language models are AI models fine-tuned for specific industries or fields to provide tailored and accurate outputs.
34. AI-generated content licensing
AI-generated content licensing addresses legal considerations around ownership and usage rights of content produced by AI systems.
35. Human-AI collaboration in creativity
Human-AI collaboration involves integrating AI technologies alongside human input, fostering a symbiotic relationship to leverage the strengths of both parties in creative endeavors.
36. Bias mitigation in generative AI
Bias mitigation techniques aim to reduce bias in AI models, ensuring fair and equitable outcomes in content generation and decision-making.
37. Inference
Inference is the process of using a trained AI model to make predictions or generate content based on new input data.
38. In-domain data for AI training
In-domain data refers to training datasets that closely represent the specific domain or target task for optimal AI model performance.
39. Deep dream
Deep Dream is a neural network visualization technique used to enhance and modify images, creating surreal and dream-like visuals.
40. Data cleansing
Data cleansing involves identifying and correcting errors and inconsistencies in datasets to ensure accurate and reliable AI model training.
41. Autoencoders
Autoencoders are a class of neural networks used in unsupervised learning to compress and then reconstruct data, often used in generative AI tasks.
42. Stochasticity
Stochasticity refers to the element of randomness in AI models, contributing to the diversity of generated outputs.
43. Zero-shot learning
Zero-shot learning is a machine learning approach where a model can perform tasks it was not explicitly trained for, given only a textual description of the task
44. One-shot learning
One-shot learning is a machine learning approach where a model can learn from just a single example, mimicking human-like learning capabilities.
45. Self-supervised learning
Self-supervised learning is a learning paradigm where a model leverages the inherent structure of data to generate its own training labels, reducing the need for extensive human-labeled datasets.
46. Collaborative AI platforms
Collaborative AI platforms like Adobe Firefly provide tools for artists and creators to collaborate with AI, enabling innovative content generation and exploring new dimensions in creativity.
47. Content Authenticity Initiative
The Content Authenticity Initiative (CAI) is a collaborative effort among various organizations and technology companies aimed at establishing standards and technologies to verify the authenticity of digital media content. CAI seeks to combat misinformation and deepfakes by providing tools to trace the origin and alteration history of media files, ensuring content integrity and transparency.
48. Content Credentials
Content Credentials refer to metadata or digital certificates attached to digital content, verifying its origin, authorship, and modification history. These credentials are part of the Content Authenticity Initiative's framework and play a crucial role in establishing the legitimacy and trustworthiness of media files.
49. Outpainting
Outpainting is a technique in artificial intelligence and computer vision used to generate content beyond the boundaries of an existing image. Unlike inpainting, which fills in missing parts of an image, outpainting extends the content beyond its original borders, often producing creative and realistic extrapolations.
50. Beta
In the context of software and AI applications, Beta refers to a phase of testing where a product or service is made available to a select group of users before its official release. Beta testing allows developers to gather feedback, identify bugs, and make improvements based on real-world usage, ensuring a more stable and polished final version.
51. Commercial use/indemnification
Commercial use, in AI-related contexts, refers to the utilization of AI models, software, or generated content for profit-making purposes. Indemnification, on the other hand, pertains to the legal protection or compensation provided to users against any potential losses, damages, or liabilities resulting from using AI products or services.
52. Coalition for Content Provenance and Authenticity
The Coalition for Content Provenance and Authenticity (C2PA) is a collaborative alliance of organizations that work towards developing standards and technologies to certify the origin, authenticity, and trustworthiness of digital content. C2PA is closely related to the Content Authenticity Initiative and aims to combat the spread of disinformation and ensure media integrity.
53. Content Aware Fill
Content Aware Fill is a computer vision and image editing technique that allows AI-powered software to intelligently remove objects or fill in missing areas within an image. The AI analyzes surrounding pixels and textures to generate a seamless replacement, resulting in a visually consistent and natural-looking result.
54. Metering/credit
Metering or credit refers to the process of attributing proper recognition or acknowledgment to the creators or owners of AI-generated content. It involves providing appropriate accreditation to avoid plagiarism and uphold intellectual property rights.
55. Noise
In the context of artificial intelligence, noise refers to irrelevant or random data present in a dataset. This extraneous information can adversely affect the performance and accuracy of machine learning models, making noise reduction an essential preprocessing step in data preparation
56. Variants
Variants, in the context of AI models or algorithms, refer to different versions or iterations created by making specific modifications or adjustments to the original model. These changes may involve altering hyperparameters, training data, or architectural configurations to explore different approaches and improve performance.
57. Seeds
Seeds, in the context of AI model training, are the initial random values set to initialize the model's parameters. The use of different seed values can result in varied model outcomes during training, and it is a common technique to control randomness and ensure reproducibility.
58. ControlNet
ControlNet is a concept in AI research that involves training a separate neural network alongside the main model to monitor and regulate its outputs. This auxiliary network acts as a control mechanism, helping to enhance the model's stability, reliability, and adherence to desired behavior.
59. NSFW list, blocklist
NSFW stands for "Not Safe for Work," indicating content that may be inappropriate or explicit for certain settings. An NSFW list or blocklist is a compilation of keywords, phrases, or content descriptors used by AI systems to detect and filter out inappropriate or sensitive material from user-generated content or search results.
Create incredible content with Adobe Express, powered by Firefly
Unlike other generative AI tools, Adobe Firefly is the most differentiated generative AI service available today, designed to generate images that are safe to use in commercial settings — free from copyrighted materials like popular characters and branded content.
Firefly is trained on hundreds of millions of professional-grade, licensed images in Adobe Stock — among the highest quality licensed images available to marketers and creatives — along with openly licensed and public domain content where a copyright has expired. Use simple text prompts in over 100 languages to make beautiful images, transform text, play with color, and so much more.
Enterprise businesses will be able to train Firefly with their own creative collateral and generate content in an organization or company’s own brand language. From there, just post to your social channels directly from Adobe Express — wham, bam, complete marketing synergy! It can take just minutes to bring your creative vision from idea to design to done, and your creative production workflow will never be the same.
And there’s so much more on the horizon. We’re exploring generating custom vectors, brushes, and textures from text prompts, altering the weather in a video with a few words, or turning simple 3D designs into photorealistic images and quickly creating new styles and variations. We’re exploring all of these possibilities and more.
Unlock your imagination with Firefly, and experiment, imagine, and create an infinite range of images with Firefly, generative AI-powered content creation from Adobe.